Accounting, in the order it happens

Part I · Why accounting looks the way it does

Chapter 1

Two Questions, and Neither Is About Accounts

A business owner who calls the accountant at the start of the month never asks what the balance of account 4111 is. They ask a single thing, in two forms: "did I make money last month?" or, more briefly, "how are we doing?"

The question seems like one. It's actually two, and they have different answers, calculated differently and read from different documents. All of accounting — the chart of accounts, double-entry bookkeeping, the fourteen stages in this book's table of contents — exists to answer them. They deserve to be named from the very first chapter, since everything else serves them.

First, Why It's Not Enough to Look at the Bank Account

The most natural reflex is to open the banking app. If you have more today than a month ago, it was a good month. It's an honest tool and everyone uses it — but it answers something other than what you think.

Let's take a trading company, in an ordinary month. It started the month with 30,000 lei in the account. It bought goods worth 60,000, which the supplier extended on credit. It sold all those goods for 100,000. Of the 100,000 invoiced, customers had paid only 40,000 by the end of the month — the rest is due the following month. Of the 60,000 debt owed to the supplier, the company paid 55,000.

Looking only at the bank account, the month looks bad: you started with 30,000 and ended up with 15,000. And yet the company sold, for 100,000, goods that had cost it 60,000. It earned, that month, 40,000 lei.

Both statements are true at the same time. The money decreased by 15,000 and the earnings were 40,000. It's not a contradiction, nor an accounting subtlety: they are simply the answers to two different questions.

The First Question: How Much Did I Earn?

The earnings of a period are measured by matching what you sold against what it cost you to sell it — regardless of when the money comes in or goes out.

In our example: you sold 100,000 worth, those goods had cost you 60,000, so you earned 40,000. The fact that a customer pays you three weeks later doesn't change in any way the reality that the goods left the warehouse and the company has a right to the money. Nor does the fact that you still owe the supplier 5,000 change how much the goods cost you.

The document that answers this question is called the income statement. It's a list with two columns for a given period: what came in as value (revenues) and what was consumed to bring it in (expenses). The difference is the result.

Earnings are measured over a PERIOD — a month, a year. It makes no sense without stating for how long.

The Second Question: What Do I Have and to Whom Do I Owe It?

The second question isn't about a period, but about a moment. On the 31st of the month, what is in the company, and who has claims on what is there?

The company in the example has, at the end of the month, 15,000 lei in the account and 60,000 to collect from customers — so 75,000 lei in total. The goods are gone, they've been sold. Of the 75,000, five thousand are not actually the company's: they are owed to the supplier. The remaining 70,000 belong to the business — the 30,000 put in at the start plus the 40,000 earned during the month.

The document that answers this is called the balance sheet. It's not a report on how things went, but a snapshot: as of this date, this is what you have and this is what you owe.

Position is measured at a DATE. It makes no sense without stating which day.

The Two Questions Meet at a Single Point

Look again at the right-hand column of the table above. The month's earnings, the 40,000, appear there. It's not a coincidence of page layout: the earnings from a period increase, by exactly that amount, what remains the business owner's.

That's the link between the two documents, and it's the only one that matters at the start: the income statement explains why the company's wealth is different today than it was at the beginning of the period. One tells you how much, the other tells you why.

That's why you can't choose only one. A business owner who looks only at earnings can end up going bankrupt with profit on paper, because the money comes in later than it goes out. One who looks only at the bank account may think a month in which they invested went badly, or that a month in which they collected on old debts went well.

Why Accounts Come Only After This

A classic manual would have started with the chart of accounts. That's an order with its own logic, but it leaves the reader without an answer to the question "why." Accounts are a tool, not an end: drawers in which sums are gathered, so that at the end of the period you can compose the two documents without digging through every piece of paper.

An accounting program makes this invisible. You tell it what document you received — an invoice from a supplier, a receipt, a bank statement — and it chooses the drawers. Still, it remains useful to know what drawers exist and why there are so many, because when a figure looks strange, that's where you return.

Before the drawers, there's one more thing to clarify: why every sum is written in two places at once. This isn't a bureaucratic convention, but exactly the mechanism that makes the two columns in the second table come out equal every time. That's the subject of the next chapter.

The Same Month, With a Single Figure Changed

To see that the two questions are truly independent, it's worth changing a single thing in the example and following what moves. Let's assume everything stays the same, but the customer pays in full during the month: not 40,000, but all 100,000.

The line that doesn't move is the one that matters. Collection doesn't create earnings; it only moves a value already earned from one place to another — from receivables into cash. And if the second column looks better, it's because the company has the same wealth in a more liquid form, not because it produced more.

This exercise deserves to be repeated mentally every time someone says "this month went badly." The first question to ask is: are we talking about earnings or about money? Almost always it's the second that's meant, while the decision is made as if it were the first.

Part I · Why accounting looks the way it does

Chapter 2

Why every amount is written twice

The previous chapter ended with a promise. In the table showing where the business stood on the 31st of the month, the two columns came out equal — 75,000 on one side, 75,000 on the other. I said then that this was no accident. Now we'll see why.

The answer is a single idea, and from it flows almost the rest of accounting: nothing appears out of nothing. Any amount that enters somewhere comes from somewhere specific, and accounting writes down both ends.

Money doesn't appear, it moves

When a business buys 60,000 lei worth of goods on credit, two things happened at once, not one. The business has more goods, worth 60,000. And it has an additional liability, also 60,000. You can't have the goods without also having the liability; it came from somewhere.

When the client pays 40,000, again two things happen: the money in the account increases by 40,000, and the right to still claim it from the client decreases by exactly that amount. The business hasn't gotten richer at that moment — it has exchanged one form of wealth (a receivable) for another (cash).

That's all there is to it. Every transaction has a "from where" and a "to where," and the two are equal by the nature of things, not by accounting convention. Accounting doesn't invent the rule; it merely refuses to write down only half of it.

Our business's month, unfolded

Let's take the exact same four transactions from the previous chapter and write them on both ends. No new numbers appear — they're the same ones, looked at more closely.

Notice the second and third rows. A single sale required two entries, because two distinct things happened: the business gained the right to 100,000, and, separately, it lost the goods that had cost it 60,000. The difference between the two — 40,000 — is the gain the previous chapter spoke of. It isn't written down anywhere directly; it emerges from the subtraction.

The equality of the totals is not a check added at the end. It's the consequence of how each line was written.

Debit and credit: two words that don't mean what they seem to

So far we've used "what it obtained" and "where it came from." Accounting has two old names for these, and the only real difficulty of this chapter is that they sound like something other than what they mean.

The left-hand column of any account record is called debit. The right-hand one, credit. That's it. They don't mean "good" and "bad," nor "plus" and "minus," and they have nothing to do with a debit card or being credited by a bank.

What they do depends on the drawer you're writing into. For a cash or inventory account, what accumulates goes through the left: it increases on the debit side. For a liability or a revenue account, the opposite: it increases on the right, on the credit side. It seems arbitrary, and in part it is — it's a five-hundred-year-old convention. What matters is that it's the same for everyone, which makes it possible to compare two different sets of accounting records.

A business owner doesn't need to memorize this rule. An accounting program applies it on its own: you tell it what document you received, it chooses the columns. Still, it's worth knowing that when someone says "I debited account 5121," they're simply saying "I wrote the amount in the left column of the bank's record" — meaning "money came in."

The check that comes for free

From the fact that each line has equal amounts on both ends, something very practical follows: if you add up all the amounts written on the left, across the whole business, and all the amounts written on the right, the two totals must be identical. For our business, 315,000 and 315,000.

If they're not, somewhere only half was written, or an amount was entered incorrectly on one end. You don't yet know where, but you know for certain there is an error. The tool that performs this comparison is called the trial balance, and it has its own chapter later in the book.

And, more importantly, what the check does NOT catch

Here it must be said, from the first technical chapter of the book, something the manuals postpone for too long: the equality of the totals does not prove the figures are correct. It only proves they are symmetrical.

If you record a 1,000-lei invoice as 10,000, but enter it wrong on both ends, the totals still come out equal and the check stays silent. If you record a correct expense in the wrong drawer, the same applies. If you forget a document entirely, even more so: what was never written at all throws nothing out of balance.

So double-entry bookkeeping catches one class of errors — the missing halves — and is blind to the rest. It's a very good tool, not an all-powerful one, and the difference between the two is paid for dearly when confused. Chapter 33 of this book is dedicated exactly to this confusion.

Where this leads

If every amount is written in two places, the natural question follows: exactly where? The goods go to one place, the liability to the supplier to another, the revenue to a third. These places are the accounts, and their complete list is the chart of accounts.

It's not a list to memorize — it's a vocabulary, and, as with any vocabulary, it's learned by using it. The next chapter shows how it's organized and why the numbers that make it up aren't chosen at random.

The four types of movement

Once it's accepted that every transaction has two ends, a practical question follows: how many kinds of movements actually exist? The answer is reassuring — four, and you can count them on your fingers.

The usefulness of the table is not theoretical. When an entry seems odd, the first check is to fit it into one of the four rows. If it doesn't fit into any of them, almost certainly a line is missing — or an account was used that doesn't describe what happened.

Why the system was invented, and why it has endured

Double-entry bookkeeping is not a modern convention. It was described in print in 1494, by the Franciscan friar Luca Pacioli, in a mathematics treatise — meaning it was already the current practice of Venetian merchants. Five centuries later, it remains essentially unchanged.

The longevity has a simple explanation worth understanding: the system was not designed to calculate, but to make errors visible. In an age without computers, a method that signals on its own the missing half was worth more than a fast one. Today calculation is free, and that property remains the sole reason the system is still here.

Part I · Why accounting looks the way it does

Chapter 3

The Drawers: The Chart of Accounts

The previous chapter ended with a practical question. If every amount is written in two places, which two? The goods in one place, the payable to the supplier in another, the revenue in a third. These places are called accounts, and their complete list is the chart of accounts.

This is probably the page that looks scariest at first sight: a few hundred lines of numbers. The good news is that you don't need to memorize it, and the even better news is that the logic behind it can be explained in fifteen minutes. Once you understand that logic, the chart is something you read, not something you memorize.

What an Account Actually Is

An account is a record that gathers together all movements of the same kind. The record has exactly two columns — the ones from the previous chapter: left, debit; right, credit. That's all.

The record for account "Cash in lei" gathers everything that came in and went out of the cash register. The record for account "Suppliers" gathers everything you came to owe and everything you paid. At any moment, the difference between the two columns is called the balance and answers a precise question: how much do I currently have in cash, how much do I still owe suppliers.

An account is not, therefore, an administrative file. It's a question you want to be able to answer at any time, without digging through invoices. The number of accounts in a chart is the number of questions the legislator considered worth being able to ask separately.

Eight Families, in a Logical Order

The Romanian chart of accounts isn't an alphabetical list, it's a classification. The first digit of any account tells you which family it belongs to, and the families are arranged in an order that isn't accidental.

The order of classes 1–5 follows a principle: from what is most stable to what is most liquid. Capital stays for years; a building, for many years; goods, for a few months; a receivable, a few weeks; money in the bank account is already money. Going through classes 1 to 5 means, in effect, descending the slope of time.

Classes 1–5 are the balance sheet. Classes 6 and 7 are the income statement. The two documents from Chapter 1 are, literally, two slices of the same chart.

The Difference That Only Shows at Year-End

Between the first five classes and the following two there is a difference in nature, not in content, and it's worth understanding early, because it explains half of the operations at year-end closing.

An account from classes 1–5 has memory. The cash balance on December 31 becomes the cash balance on January 1; money doesn't disappear because the year changed. The same is true for receivables, payables, and inventory.

An account from classes 6 and 7 has no memory, nor should it. The question "how much have I spent" only makes sense if you specify for how long — otherwise you'd keep adding forever. That's why, at the end of every year, all revenue and expense accounts are emptied into a single account, 121 Profit or Loss, and the following year they start from zero. This operation has its own chapter further on; for now it's enough to note that emptying isn't a formality, but the natural consequence of the fact that these accounts measure a period.

How to Read a Code

The digits aren't inventory numbers — they're an address read from left to right, narrowing the meaning at each step.

The uniformity of the first levels is why the accounting of two companies can be compared, and why a trial balance can be read by anyone in the profession. It's not bureaucracy: it's like postal addresses, where the street name is written the same way for everyone, otherwise the letter never arrives.

On Which Side an Account Grows

The previous chapter said that debit and credit don't mean plus and minus, but left and right, and that the effect depends on the account. Now we can say precisely what it depends on: the class.

The rule is easier to remember from a single observation: accounting looks at things from the company's perspective, not yours. When the bank tells you "we've credited your account," it's speaking from its own perspective — to the bank, you are a liability. In your company's accounting, the same money coming in increases the bank account on debit.

Analytical Accounts: The Only Part You Choose

Account 4111 gathers all customers together. It's enough to know how much you have to collect in total, but it doesn't tell you from whom. For that, analytical accounts are opened — 4111.01, 4111.02, and so on, one per customer.

The practical rule is simple: open an analytical account only where you really need a separate answer. A chart with a thousand useless analytical accounts is harder to read than one with twenty well-chosen ones, and at month-end someone still has to review them all.

In practice, an accounting program tracks records by partner without requiring you to open anything; analytical accounts remain useful for cases where you want the separation to actually appear in the trial balance.

Why You Don't Need to Memorize It

A business owner never opens the chart of accounts. In an accounting program, he states what document he has — a supplier invoice, a receipt, a bank statement — and the program chooses the drawers. That choice is called an accounting mapping ("monografie contabilă") and is, essentially, a dictionary between everyday language and the language of accounts.

An accountant does open it, but not to memorize it: to verify. When a figure looks odd in a report, the path back invariably passes through an account's record, and from there through the document that produced it. The chart is the map on which that path is drawn.

With this, the tools are on the table: we know what questions accounting answers, why every amount is written twice, and exactly where it is written. What remains is a question that isn't technical, but on which everything depends: who is responsible for these figures. That's the subject of the next chapter.

How to Read an Account You've Never Seen Before

Nobody memorizes the chart of accounts. It's read, and reading follows three steps that work for any symbol, including one encountered for the first time.

The three levels have names in the regulation: first-degree synthetic account (three digits), second-degree synthetic account (four digits), and analytical account (whatever is added after, at the company's choice). An analytical account isn't a new account — it's a breakdown of the synthetic one, and the sum of all analytical accounts must exactly equal the synthetic balance.

The Second Digit, Which Tells You the Meaning

There's a pattern that shortens learning considerably, and textbooks often pass over it in silence. In fixed asset and inventory accounts, an 8 in the second position means depreciation, and a 9 means impairment adjustment. These are contra accounts: they don't add value, they subtract it.

Part I · Why accounting looks the way it does

Chapter 4

Who Is Responsible for the Numbers

Up to this point, the book has discussed mechanics: what questions accounting answers, why every amount is written twice, and in which drawers. This chapter has no formula and no accounting scheme. Instead, it has the one question where a wrong answer costs real money: who is responsible for what's written there.

It's a question no one asks at the start of a collaboration, because the answer seems obvious to each of the two parties — and, usually, differently obvious. The business owner believes they paid an accountant to have peace of mind. The accountant knows they work with the documents someone gives them. Both are half right, and the missing half only becomes visible during an audit.

What the Law Says, Briefly and Directly

Responsibility for organizing and running the accounting rests with the company's administrator. Not the accountant, not the software, not the accounting firm. The fact that the work is outsourced doesn't move the responsibility outside: it can be outsourced as execution, not as responsibility.

At the same time, the annual financial statements cannot be signed by the administrator alone. The Accounting Law requires a second, qualified signature — an accountant with higher economic studies or a CECCAR member (art. 10 and 28). The current monthly declarations, however, can be filed by the administrator alone.

Execution can be outsourced. Responsibility cannot.

What Only One Knows and What Only the Other Knows

The division of work between business owner and accountant is not a matter of internal organization, but stems from something deeper: each of them holds information that the other cannot possibly have.

Only the business owner knows whether the transaction actually happened. Whether the goods on the invoice entered the warehouse. Whether the consulting service was actually performed. Whether the car registered to the company is also used for personal purposes, and to what extent. An accountant, however good, cannot learn these things from paper: paper says what's written on it.

Only the accountant knows how the transaction translates into accounting and what tax consequences it has. Whether an expense is deductible or not. Whether the VAT is fully deductible, half deductible, or not at all. Whether an asset is expensed now or depreciated over five years. A business owner, however smart, has no way of knowing this without doing the job themselves.

That's why the collaboration isn't a luxury, but a structural necessity. The document is where the two kinds of knowledge meet: the business owner affirms that the transaction behind it is real, the accountant gives it the correct accounting form.

The Two Ways of Getting It Wrong

When one of the two does the other's job, the result is predictable, and both variants are seen often.

The business owner who "manages alone" keeps accounting that looks formally correct but says something different from the reality of the business: personal expenses charged to the company because "that's how it's done," unrecorded revenue because "it was a small amount," a trial balance that closes impeccably on figures that describe nothing. Double-entry bookkeeping, as seen in Chapter 2, is perfectly happy with false data, as long as it's symmetrical.

The accountant who receives a stack of papers at year-end and records everything given to them, without asking any questions, is also only doing half the job. They cannot know what's missing from the stack. And what was never recorded at all — a lost invoice, an unreported receipt — doesn't unbalance anything and doesn't appear in any report.

Both failures have the same cause: someone assumed that the other's responsibility also covered them. It doesn't. In an audit, the administrator is responsible for the company's accounting, and the accountant is professionally responsible for what they signed.

What a Working Collaboration Actually Means

No complicated procedures are needed. Three habits cover almost everything.

And Where Does the Software Fit Into All This

An accounting program calculates and verifies. It applies rules without tiring, keeps track of deadlines, refuses an entry that doesn't balance, shows what documents are missing from a month. All of this is real and belongs to the execution side.

However, it does not decide whether an expense belongs to the company or to you. It doesn't know whether the invoice corresponds to a delivery that actually took place. It doesn't sign the balance sheet and isn't accountable to anyone. A program claiming otherwise would be selling a peace of mind it cannot back up.

With that, the roles are clear: the business owner brings the reality, the accountant brings the rule, the software does the work between them. One thing remains before moving into practice — the order in which all this happens, month by month. That's the map of the book, and it's the next chapter.

Three Situations Where Responsibility Is Determined in Practice

The principles in this chapter seem abstract until the day a problem appears. It's worth, therefore, running them through three cases that happen often.

The second row deserves nuance, since it's the most often wrongly invoked. The accountant is responsible for the classification if they had all the information. If the document didn't state what was purchased, and no one clarified it, responsibility shifts to whoever should have clarified it — which is why a question asked in writing is worth more than any explanation given afterward.

Part I · Why accounting looks the way it does

Chapter 5

The Accounting Cycle: The Book's Map

The four chapters so far have laid the tools on the table: what questions accounting answers, why every amount is recorded twice, where exactly it is recorded, and who is responsible for what. What remains is order — and, unlike the rest, order is not a didactic convention.

A company goes through the same sequence of operations every month, and each step assumes the previous one has been completed. You cannot settle VAT before recording the month's invoices. You cannot close the month without reconciling the bank statement. You cannot prepare the balance sheet until all the months are closed. This sequence is called the accounting cycle, and the book follows it step by step.

Why Not by Topic, Like Textbooks

A classic textbook divides the material into thematic chapters: accounts, then inventory, then payroll, then VAT. That's a good order for someone taking an exam and a poor one for someone running a company, for two reasons.

First, because it hides dependencies. In reality, an inventory mistake shows up in VAT, and a payroll mistake shows up in the balance sheet. Presented separately, the topics seem independent; in practice they rest on one another.

Second, because it answers questions at a different moment than when they arise. Someone keeping the books doesn't ask "what do I know about depreciation," but "what do I need to do now, since it's the 25th of the month." This book is organized around the second kind of question.

The order of the phases is not an author's choice. It's a real dependency: each step consumes the result of the one before it.

The Fourteen Phases

The complete cycle looks like this. The right-hand column shows the pace at which each phase runs — and this is where the chapter's second important idea lies: not everything is done monthly.

Three Paces, Not One

The costliest confusion at small companies is thinking that everything is done "in accounting, once a month." In reality, the cycle has three overlapping paces, and mixing them up is the classic source of chaos in March, when documents from September are being searched for.

The daily pace belongs to documents and money. An invoice received today and recorded today costs a minute; the same invoice sought four months later costs an afternoon, and sometimes it is never found at all.

The monthly pace belongs to checks: payroll, depreciation, adjustments, trial balance, VAT, closing. It's a list gone through in order, and skipping a step doesn't show immediately — it shows at the next one that depends on it.

The annual pace belongs to the balance sheet: the result, the tax, the distribution, the financial statements. It's done only once, but only after the twelve months are closed. That's why a month left open in July becomes a problem in May of the following year.

Why the Book Groups Two Distant Phases

One deviation from the strict order deserves explanation, since it's visible in the table of contents. Part Eight treats VAT (phase 9) and declarations (phase 13) together, even though four phases lie between them.

The reason is that, from the company's point of view, they are the same thing: what you owe the state and how you prove you filed. The VAT return is the first declaration a company encounters and the most frequent one, and separating it from the rest of the declarations would have split a single idea into two places. The rest of the book follows the order of the phases without exception.

How to Read the Rest of the Book

Each part that follows is a phase. The chapters within it answer the questions that arise exactly at that moment, in the order in which they arise. A business owner can read the book once from cover to cover, then return to the part they need on any given day.

The main text is written for someone who hasn't done accounting. Where the subject requires it, the chapter continues with a section for the accountant — the journal entries, the legal basis, the verification — marked distinctly, so the two readings don't get mixed up. A business owner can skip them without losing the thread; an accountant will find them exactly where expected.

The introductory part ends here. The next chapter begins the cycle proper, with the phase that is done only once and on which the whole rest of the year depends: opening the fiscal year.

The Map, on an Actual Month

The phases make sense only once you see them dated. Here is the same sequence, laid out over the days of an ordinary month at a small company, a VAT payer with two employees.

Note that filing takes up a single day, while everything before it takes up three weeks. This ratio is precisely the book's thesis: the declaration is not a task in itself but the consequence of a month already completed. Whoever treats it as a task fills it in on the 24th and amends it on the 3rd.

Part II · Opening the fiscal year · phase 1

Chapter 6

The Company and Its Regimes

The first phase of the cycle happens only once and is rarely revisited. That is exactly why it is dangerous: you don't get the chance to get used to it, and if it's wrong, the mistake doesn't show up that day — it shows up in every declaration of the year.

Opening the fiscal year practically means answering two questions about the company. They are short questions with answers that seem administrative, but everything else follows from them: what declarations you owe, by what deadlines, how much tax you pay, and whether or not you add VAT to invoices.

The two questions are independent. A company can be a VAT payer and a micro-enterprise, or a non-payer and subject to corporate income tax, in any combination.

The first question: do you add VAT to the invoice or not?

A new company is not automatically a VAT payer. It can remain a non-payer as long as its annual turnover stays below the exemption threshold — 395,000 lei — and it can request registration earlier if it wants.

The practical difference is easy to describe and hard to intuit. The non-payer doesn't add VAT to its prices, but also doesn't deduct the VAT on what it buys: the tax paid to suppliers becomes, for it, a cost. The payer adds VAT to sales, deducts the VAT on purchases, and remits to the state only the difference.

From this comes the practical rule every business owner should know: payer status benefits you if you sell mainly to other VAT-paying companies and have large purchases with VAT. It disadvantages you if you sell to individuals and have low expenses — there, VAT gets added to the final price and pushes you out of the market compared to a non-paying competitor.

Above the threshold, the choice disappears: registration becomes mandatory. And the threshold is monitored throughout the year, not established at the end — whoever exceeds it in August doesn't have until December.

The second question: how is the gain taxed?

Here too there are two options, and the difference between them is deeper than it appears: it's not just the rate that changes, but what it's applied to.

The micro-enterprise pays, in 2026, the flat rate of 1% on the tax base formed from accounting-recognized revenues, with the deductions and additions provided by the Tax Code. It doesn't matter whether it had a profit or a loss, nor whether all invoices were collected: the tax follows accounting revenue, not the bank statement. It's simple to estimate, but its base is not the same as either collections or the total of class 7.

Corporate income tax is 16% of the gain, i.e., of the difference between revenues and expenses, fiscally adjusted. If the year was a loss, the tax is zero. It's more complicated to calculate, but it follows the reality of the business.

The micro regime is not available at will at any time: eligibility requires cumulative fulfillment of the legal conditions, including the €100,000 threshold, having at least one employee, the rule on holdings above 25%, timely filing of financial statements, and the absence of excluded activities. From 2026, one can return to micro status if all conditions are met, but the option and exit are declared within their own deadlines; simply being below the threshold is not sufficient.

Which is cheaper: arithmetic, not opinion

The answer doesn't depend on the company's size, but on its margin. The two taxes become equal at a point that can be calculated exactly: 1% of revenue equals 16% of profit when profit represents 6.25% of revenue.

The conclusion is easy to remember: at high margins, micro is advantageous; at low margins, corporate income tax is. And if the company has loss-making years, micro becomes downright expensive — it pays even when it doesn't earn.

In 2026 there is no longer a 3% micro rate, nor the €60,000 threshold that used to change the rate. The economic comparison is made exclusively between 1% on the micro base and 16% on the taxable result, after prior verification of eligibility.

What changes during the year

Both answers can be overturned by the company's growth, and in both cases the change doesn't wait for the end of the year.

For VAT, exceeding the threshold requires registration within a short deadline, and from the registration date invoices are issued differently. For micro, exceeding the €100,000 threshold triggers corporate income tax starting with the quarter in which it was exceeded. The threshold is also checked together with the revenues of related enterprises, under the rules of Articles 47, 52, and 54 of the Tax Code; for the 2026 threshold test, the turnover as defined by accounting regulations is used, with special rules for repeated asset transfers.

That's why monitoring these thresholds isn't a December concern, but a permanent one. Accounting software can calculate them continuously and warn in advance; this is, in fact, one of the few places where automation actually prevents a fine, not just saves time.

What gets filled in, concretely

In the application, this phase means a few fields on a single page: the company's identification data, the VAT code if any, the tax regime, the fiscal period, the bank accounts, the document series. It's filled in once, at the start, and revisited only when something real changes.

From that moment on, the software knows what declarations you owe and by what deadlines, what rates to apply on invoices, and what reports are required at year-end. Nothing that follows in this book works correctly if the answers here are wrong — and correcting them six months later usually means redoing the declarations filed in the meantime.

One more thing remains before the first documents: if the company is not in its first year, the bookkeeping doesn't start from zero. Balances must be carried forward. That's the next chapter.

The four decisions made at incorporation, and what each one costs

A new business owner thinks the only decision is "LLC or sole proprietorship." In reality there are four, they're made almost simultaneously, and three of them are hard to change.

It's worth saying clearly something nobody says at incorporation: these choices are not made once and for all — they are REVIEWED annually, based on actual figures. A company that chose micro at incorporation because it seemed cheaper, and that ends up with low margins, pays tax on revenue even in loss-making years.

Part II · Opening the fiscal year · phase 1

Chapter 7

The Opening Trial Balance

The previous chapter established the company's tax regimes. Only one question remains before the first document: what figures do we start from?

For a newly established company the answer is almost blank — but not quite. For one that has been operating for years and is only now entering an accounting program, the answer is an entire trial balance, and everything calculated from here on depends on its accuracy.

What Carries Forward and What Doesn't

Chapter 3 divided accounts into two kinds, and now the distinction becomes practical. Accounts in classes 1–5 have memory: their balance on December 31 becomes the opening balance on January 1. The money in the bank account doesn't disappear because the year changed, nor do the receivables, nor the payables, nor the goods in the warehouse.

Revenue and expense accounts, classes 6 and 7, never carry forward. They measure a period, and the period has ended; the new year starts from zero. Anyone who "carries forward" revenue or expense balances makes a mistake that shows up immediately: the new year's result starts inflated.

The opening trial balance contains only balance-sheet accounts. If you see a class 6 or 7 account in it, something is wrong.

The Same Company, One Year Later

Let's return to the company from the first chapters. It had closed the month with 15,000 lei in the bank, 60,000 receivable from customers, 5,000 payable to the supplier, the initial 30,000 contributed, and a gain of 40,000. Assume the year ended there and that the profit was left in the company, not distributed.

Here, in a five-row table, is everything the first part of the book said. The two columns are equal because every transaction over the year was recorded on two sides. The revenue and expense accounts have disappeared, because they measured the period that has ended. And the gain hasn't evaporated: it has changed its name.

A New Company Doesn't Really Start from Zero

A company set up last month has no receivables, payables, or inventory, but it does have share capital. The amount contributed by the shareholders at incorporation is the first entry in its life, and the opening trial balance contains exactly those two lines: the money on one side, and the capital on the other.

It's the simplest case possible and, precisely for that reason, the best example to read: the company has something (money) and that something belongs to someone (the shareholders). Nothing appears out of nothing — not even on the first day.

Three Levels of Takeover, Not One

When taking over a company with history, "I entered the balances" actually means three distinct operations, and skipping one of them produces bookkeeping that balances in total but is unusable in practice.

There is only one rule linking the levels, but it's the most frequently violated one at takeovers: the sum of the analytical balances must equal the synthetic balance. If the customers account shows 60,000, and the customers taken one by one add up to 58,400, the difference of 1,600 isn't a rounding error — it's a forgotten customer or an amount entered twice.

Why an Unbalanced Trial Balance Never Heals Itself

The equality between the total debit and total credit isn't a formality to fill in at takeover, but the starting point of all future verifications. Chapter 2 showed why: every transaction adds equal amounts to both columns, so the difference between them never changes.

The consequence is unpleasant and worth stating clearly. If you start with a difference of 1,200 lei, you will have that same difference of 1,200 lei in every trial balance throughout the year, no matter how many correct transactions you record in the meantime. It doesn't dilute, doesn't get lost, doesn't get offset. The only fix is to go back to the takeover trial balance and find what's missing.

That's why a serious program refuses an unbalanced takeover from the start, instead of accepting it and complaining later — and refuses it entirely, not halfway, because a partial takeover is harder to fix than one that never happened. For the same reason it also refuses a balance entered with a minus sign in the wrong column: an account doesn't have a "negative debit balance," it has a credit balance.

How It's Done, in Practice

In the application, the takeover is done from a single place: either the file exported from the old program is imported, or the balances are entered account by account. The program checks the equality before writing anything and refuses the takeover if it doesn't balance, stating exactly by how much it differs.

Balances for each customer and supplier are detailed separately, and the inventory has its own quantitative-value takeover. There are three steps, not one, exactly as in the table above — and their order doesn't matter, as long as in the end all three match each other.

With this, phase one is complete: the company knows who it is, what tax regime it has, and where it starts from. From the next chapter begins the day-to-day work, with the single door through which anything enters the accounting — the document.

Where the Opening Figures Come From, by Case

"Opening trial balance" means different things depending on whether the company is new, established, or taken over from another accountant — and confusing the three is the source of most beginning-of-year errors.

When taking over from another accountant, there is a practical rule worth adopting without exception: balances are taken over, but they are also verified. A customer balance that cannot be broken down by partner, or a cash balance that doesn't match the register, are inherited problems that become yours from the first day you recorded them.

Part II · Opening the fiscal year · phase 1

Chapter 8

The equality that must hold from day one

The previous chapter said that an opening trial balance must balance, and that if it doesn't, the difference never simply disappears. This chapter answers the question that naturally follows: why must it, in fact, balance? And, more practically, what do you do when it doesn't.

It's worth clarifying now, at the start of the cycle, since this is the last chance to examine the equality calmly. From here on it will be assumed at every step.

It's not an imposed rule, it's a definition

The equality in accounting is not a requirement that the legislator added to make life difficult. It's a consequence of the way one of the terms is defined.

What the business owns is called an asset. What it owes to others is called liabilities. And the difference between them — what would remain if all creditors were paid — is called equity. It's not a sum set aside somewhere, but a remainder: exactly what's left over after everyone takes what's owed to them.

Assets = Liabilities + Equity. The equality holds because the last term is defined as the difference between the first two.

For the business from previous chapters: it has 75,000 lei (cash and receivables), owes 5,000 to the supplier, so equity is 70,000 — the 30,000 contributed by the owner plus the 40,000 earned. It couldn't be otherwise: if tomorrow it collected everything and paid everything, exactly 70,000 would remain in the business.

That's why a trial balance that doesn't balance doesn't signal a "calculation mismatch." It signals that somewhere something was written that cannot be true about the world: either the business has an asset that comes from nowhere, or it owes someone something it never received.

The same equality, seen twice

A beginner has the impression that these are two different checks: one on the balance sheet, where assets must equal liabilities plus equity, and another on the trial balance, where the debit total must equal the credit total. They are the same equality, viewed from two places.

The balance sheet shows it in balances, at a date. The trial balance shows it in movements, over a period. The first follows from the second: if each transaction added equal amounts to both columns — chapter 2 — then the resulting balances will also be equal in total.

The four equalities of the trial balance

The Romanian trial balance (balanță de verificare) does not compare a single pair of totals, but four. This is not redundancy: each pair catches something different, and the point at which the first equality breaks tells you where to look.

The practical use is exactly this: before searching through three hundred transactions of the month, look at the first pair. If the opening balances don't reconcile, the current month is innocent.

When it doesn't balance: how to search

Here an old craft comes in, predating calculators, and still valid. The size of the difference often tells you what kind of error was made — and shortens the search from hours to minutes.

The double-amount rule deserves to be understood, since it's the most commonly used. If an amount of 1,210 lei was posted to credit instead of debit, the debit column loses 1,210 and the credit column gains 1,210: the difference is not 1,210, but 2,420. So, when you see an imbalance, divide it by two and look for that number among the period's transactions — most of the time you'll find it right away.

What the equality does not prove

Chapter 2 already said this, but it must be repeated here, since we've just spent a page praising the equality: the fact that the trial balance balances does not mean the figures are correct. It only means they are symmetrical.

A wrong amount on both ends, a transaction recorded in the wrong account, a document forgotten entirely — none of these disturb the equality. The trial balance is a detector of asymmetry, not of truth, and chapter 33 shows what else must be verified.

What comes next

Phase one is complete. The business knows who it is, what tax regime it has, what figures it starts from, and those figures balance. Nothing that follows requires returning to these pages unless something real changes in the business.

From here begins the day-to-day work, and its first rule is also the strictest in the whole book: nothing enters the accounting records without a document. That's phase two, and the next part.

What "doesn't balance" means on day one

The equality must hold from day one, and the causes for it not holding at opening are few and identifiable. They're worth listing, since each has a different remedy.

The rule of a difference divisible by nine deserves to be understood, not memorized. If you transpose two adjacent digits — writing 5,480 instead of 4,580 — the difference is 900, i.e., 9 × 100. Any transposition produces a multiple of nine, regardless of the size of the number. So a difference of 6,300 in a trial balance that won't balance is, almost certainly, a pair of digits swapped with each other.

Part III · The document · phase 2

Chapter 9

Nothing goes in without a document

This is where day-to-day work begins, and this is also where the strictest rule in the entire book lies. It's not a recommendation of good practice, but the condition of existence of accounting itself: any amount recorded must have behind it a document attesting that the transaction took place.

Stated in reverse, the rule sounds unpleasant and is just as true: what has no document does not get recorded. Not even if you're certain it happened. Not even if you paid by card and see the amount on the statement. Not even if the supplier promises the invoice next week.

Why the rule is so rigid

An accounting register is, in essence, a statement about the past: on this date, this thing happened, for this amount. A statement about the past that cannot be verified has no power — not before an inspector, not before a bank, not in a dispute with a partner.

The document is what turns the statement into evidence. That's why the law doesn't require it as a formality, but defines it: the recording of the transaction is made at the moment it is carried out, in a written instrument that underlies the accounting entry.

Accounting does not record what happened. It records what can be proven to have happened.

What makes a piece of paper a supporting document

Not every piece of paper with figures on it serves the purpose. A supporting document must allow anyone reading it five years from now to reconstruct the transaction without asking anyone. Hence the elements it must contain.

It's worth noting what does not appear on the list: the accounting account. The document states what happened in the real world, not how it is recorded in accounting. The translation is the job of the next chapter, and it belongs to the accountant — not to whoever issued the paper.

The bank statement is not a supporting document

This is the most frequent confusion at small companies, and it deserves separate clarification, because it seems like pedantry and it isn't.

The bank statement proves that money moved. It does not prove why it moved. A payment of 5,000 lei to a supplier could be the settlement of an invoice, an advance, the repayment of a loan, or an error — the statement looks the same in all cases.

That's why the payment is recorded based on the statement, but the expense is recorded based on the invoice. These are two different transactions, with two different documents, and confusing them is the classic cause of expenses recorded twice: once after the invoice and again after its payment.

When the document is missing

It happens: a lost invoice, a faded receipt, a supplier who no longer answers. The rule isn't suspended, but there is a procedure — reconstruction of the document, with a declaration and proof of the steps taken to obtain it.

What isn't permitted is bypassing it: an entry made "based on" an email, a phone photo, or someone's memory. Not because it would necessarily be untrue, but because in three years no one can say anymore whether it was.

What changes when documents are electronic

The rule stays the same; only the medium changes. A file has the same value as a paper document if it meets the content requirements and can be kept and read throughout the legal retention period.

An accounting program helps here in a way paper cannot: it keeps the document attached to the entry, not just the figures from it. When, four years later, someone asks where the amount in a report comes from, the trail back to the original file takes two clicks, not a search through an attic archive.

What remains the hard part, which no program does for you: the document has to make it into the application. The next chapter shows what happens to it once it arrives — how it becomes an accounting entry.

Part III · The document · phase 2

Chapter 10

From paper to journal entry

The document tells you what happened in the real world. Accounting needs the same information in a specific form: which account is debited, which account is credited, and for what amount. The translation between the two is called a journal entry, and this chapter shows what it looks like.

The good news for a business owner: they don't have to do the translation themselves. The useful news: it's worth knowing how to read one, because it's the only place where you can check whether the software — or the accountant — has understood the document correctly.

Anatomy of an entry

A journal entry has five elements, none of them superfluous.

The short form, used by everyone in the profession, places the debit account before the credit account, separated by an equals sign: 371 = 401 means "goods came in, the payable to the supplier increased." It's read from left to right: what we obtained, where it came from.

One invoice, three lines

The most frequent document in a company's life, the purchase invoice for goods with VAT, produces a three-line entry. Let's take an invoice of 1,000 lei plus 210 VAT, total 1,210.

This shows why VAT is neither goods nor an expense. The 210 lei cost the company nothing: it will be deducted from the VAT payable. That's why it sits in its own account, separate from the value of the goods — and if the company were not VAT-registered, the 210 would be added to the cost of the goods, because it would remain there permanently.

The same document produces different entries depending on the company's tax regime. The document is the same; its accounting is not.

Simple entry and compound entry

When a transaction links a single account to a single account, the entry is simple: collection from a customer, 5121 = 4111. When one of the two sides splits into several accounts, as in the invoice above, the entry is compound.

What is not allowed, though it may seem convenient, is an entry with multiple accounts on both sides at once, mixing unrelated transactions. Such an entry balances on the total, but can no longer be read: a year later, no one can say which amount on the left corresponds to which amount on the right.

The preview: how to check without knowing accounting

Here lies, for a business owner, the most useful habit in the whole book. Before saving a document, the software shows exactly what entry it will record. You don't need to know the formula to do the check; you just need to look at three things.

A detail that seems technical but is important: the preview must come from the same engine that performs the actual recording. If the software calculates the preview in one place and the recording in another, the two can silently drift apart, and what you checked is no longer what got saved.

Who does the translation

In practice, no one composes entries by hand for routine documents. The software has the translation prepared for every typical situation: you choose which document you have, fill in the amounts from it, and the entry composes itself.

This set of ready-made translations is called a chart of accounting treatments (monografie), and its richness determines how often you'll need to improvise. That's the subject of the next chapter.

The translation, on the same document, at three different companies

The same piece of paper doesn't produce the same entry at every company. It's worth seeing on a case, because it clarifies what actually decides the translation: not the document, but WHAT the company does with what it bought.

From this follows a surprising consequence: the accountant cannot always do the translation alone, from the document, for every transaction. For routine ones, the pattern is clear. For ambiguous ones, information is needed that is written nowhere — and the question asked at that moment is cheaper than the correction made a year later.

Part III · The document · phase 2

Chapter 11

The Vocabulary of Situations

The previous chapter ended with a practical promise: no one composes journal entries by hand for ordinary documents. The software has the translation ready for every typical situation, and the set of these translations is called the accounting monograph (monografie contabilă).

It deserves to be seen as a vocabulary, not a list. Each entry links a real-world situation — "I bought goods on credit," "I paid wages," "a piece of equipment broke down and I scrapped it" — to the accounting entry that corresponds to it. Whoever knows the vocabulary no longer needs to compose; they need to choose.

How Many Situations Does a Small Business Actually Have

The answer usually surprises in both directions. There are fewer than a beginner fears and more than someone who has run a simple business for a year believes.

In the application from which this book was born, the vocabulary has over one hundred and twenty entries, grouped into twenty-one families. The richest are purchases and sales, which is natural: that's where most of a company's economic life takes place.

The useful observation is that an ordinary business uses, in practice, between fifteen and thirty entries. The rest sit there for the day the situation appears — and when it does appear, it's better for it to already be prepared than improvised under pressure.

Why You Choose a Type Instead of Writing the Accounts Directly

A program could, in principle, let you write debit and credit directly. Some do. Choosing a document type, however, has three advantages that you lose as soon as you start improvising.

First, the type brings with it all the lines, not just the obvious one. In a sale of goods, a beginner records the revenue and forgets the inventory write-off; the document type doesn't forget. We saw in Chapter 2 that this omission leaves the trial balance perfectly balanced and the result false.

Then, the type carries the tax information: whether the transaction enters the VAT return and on which line, whether the document must be sent to the e-Factura system, whether the value is reported in informative declarations. The accounts alone say nothing of all this.

Finally, the type makes the document retrievable. A year from now, the question won't be "where's the entry with 371 = 401," but "where's the invoice from that supplier."

Choosing the right type isn't a formality to fill in: it's the decision that determines what gets recorded, what gets declared, and what can be found.

When the Wrong Type Is Chosen

The most instructive case is cancelling an invoice. Many people's reflex is to enter the same invoice with negative amounts — it seems clean: minus one thousand, minus two hundred ten, done.

The result, however, is an incomplete entry that closes perfectly. And the consequence only shows up weeks later: the customer is left with a receivable that doesn't exist, and the collected VAT remains inflated and goes out that way in the return.

The vocabulary has, for exactly this situation, its own entry — sales credit note, "in red" — which reverses all the lines and, in addition, keeps the reference to the corrected invoice. The difference between using the right tool and improvising isn't one of elegance, but of the correctness of the declaration.

A well-made program doesn't just offer the correct tool: it refuses improvisation and tells you which tool to use. Chapter 29 returns at length to corrections.

How It's Chosen in Practice

The entries are written in plain language, not accounting terms: "purchase invoice for goods," "supplier payment receipt," "materials consumption slip." You choose based on what's written on the paper in your hand, not on which account you think should be used.

When the document is read automatically from a file, the program proposes the type itself, and you confirm it. Confirmation isn't a formality: it's exactly the moment when a human checks a machine's assumption, and Chapter 13 shows what's being checked then and why.

Before that, though, there's another question to clarify: who exactly presses the save button, and what it actually means for a document to have "entered the accounting records." That's the next chapter.

Part III · The document · phase 2

Chapter 12

Draft, validated, approved, posted

A document entered into the program does not, by that simple fact, become part of the accounting records. Between "I typed it" and "it's recorded" there are several steps, and they are not the program's bureaucracy: they are exactly the separation discussed in chapter 4, between who enters and who takes responsibility.

In a small business, where the same person does both things, the steps seem like a fuss. But they become useful on the first day someone makes a mistake — and they are indispensable the moment someone other than the one who signs the balance sheet works on entering documents.

The four states

The threshold that matters is the last one. Until posting, an entry is an intention: it can be corrected, completed, or deleted without anyone ever knowing it existed. After posting it becomes a recorded fact, and recorded facts are not deleted — they are corrected through another entry, which also remains.

Before posting, a mistake is a correction. After posting, it is a new operation.

Why what has been posted is not deleted

The question comes up right away, and it is legitimate: if I made a mistake, why can't I just delete it and write it correctly? The answer has two parts, one practical and one of principle.

Practical: other things have already been calculated based on the posted entry. A VAT return already filed, a trial balance sent to the bank, a statement handed to a partner. Deletion would make the same documents say something different today than they said yesterday, with nothing explaining the difference.

Of principle: the trail of the correction is part of the evidence. Accounting records in which mistakes disappear without a trace are, to an outside observer, indistinguishable from records where figures are adjusted as needed. Precisely because errors are normal and inevitable, the way they are corrected says everything about the seriousness of the records.

Who can do what

The steps make full sense when linked to different people. The usual model at a business with an external accountant looks like this: someone in the business enters documents as they appear, the accountant verifies and posts.

The advantage is not control, but the correct division of labor. The person in the business knows what happened — they saw the goods, they signed for receipt. The accountant knows how it should be recorded. Each does the part the other cannot do, and the state of the document shows, at any moment, where things stand.

In a business where the owner does everything alone, the steps remain useful in a more modest way: they separate "I set the document aside so I wouldn't lose it" from "I checked it and it's good." These are two different things, and mixing them is the usual cause of invoices recorded twice.

What is visible from the outside

A side effect of the states, often more valuable than the control itself: at any moment one can say what is ready and what is not. How many documents for the month are still drafts, how many are awaiting verification, how many have actually entered the accounting records.

At month-end, the question "are we done?" is no longer answered from memory, but read off. And when the answer is no, one can see exactly what is missing — which turns the month-end close from a hunch into a list. Chapter 39 returns to this.

The question remains of what is actually verified before a document is allowed to pass through. That is the next chapter.

What can be done in each state

The four states are not decorative labels: each one changes what can be done with the item and what the rest of the application sees.

From this comes a practical consequence that beginners discover late: an item entered and left in draft is NOT a recorded entry. It does not appear in the trial balance, does not enter the VAT return, and is not visible in reports. "Entered" documents that cannot be found anywhere are, almost always, forgotten drafts.

Why three states before posting, and not one

In a small business, the three states before posting seem like a luxury: the same person enters and posts. But they describe three different questions, and in a business with a team each has a different answer.

Posting is not a save button. It is the signature on a statement about the past — and that is why, after it, correction is done by reversal (storno), not by modification.

What happens to a draft from a closed month

The case comes up often and deserves clarification: a draft created in June, left unposted, and in the meantime June has been closed.

The draft still exists — nothing has been deleted — but it can no longer be posted there, because the period is locked. It is not a loss of data, but a question: did that document belong to June or not? If yes, it is recorded in the current month and it is assessed whether an amending declaration is needed. If not, the date is corrected.

This situation is, in fact, exactly why the period guard checks at POSTING, not at creation: a draft can be composed at any time, but entry into the records is subject to the calendar.

Part III · The document · phase 2

Chapter 13

The checks at the door

A document can be entered by a person who has it in front of them, or read automatically from a file — a scanned invoice, a PDF received by email. In the second case a new question arises: how much can you trust what the machine has read?

The serious answer is not "a lot" or "a little", but a list of checks performed before the document is allowed to pass through. This chapter shows what is checked, why each check exists and — the most important part — how their results are combined.

The nine checks

It is worth noting that only two of them concern the quality of the reading. The remaining seven check things that an attentive person would also check: that the figures add up, that the date is valid, that you don't already have the same invoice. Automation does not invent new checks — it performs the well-known ones every single time, without getting tired.

Why all of them, not their average

Here lies the chapter's most important design decision, and it is one that any accountant should demand from the software they work with.

The natural temptation is to calculate a score: eight checks out of nine passed means 89%, i.e. "almost certainly good". The temptation is wrong. If the only failed check is the duplicate check, the document is not "89% good" — it is an invoice recorded a second time, and the other eight passes change nothing.

That is why the decision to post automatically is a conjunction, not an average: it posts by itself only if all checks pass. The score exists, but it serves reporting purposes — it shows how well automatic reading is performing overall — and it decides nothing about any particular document.

A high score conceals precisely the check that matters. Blocking checks are not averaged, they are conjoined.

The confidence threshold

The check concerning reading certainty has a numeric threshold: below it, the document does not pass by itself, no matter how good the rest looks. The threshold is deliberately set high, higher than the one at which the system merely flags a person, because letting a document in without any human review requires more certainty than merely flagging it.

A detail that seems technical and is a real trap: the threshold is not a property of the world, but of the model doing the reading. A new model may systematically report lower confidence at the same accuracy — in which case a threshold left unchanged stops perfectly good documents. Thresholds of this kind must be recalibrated whenever the underlying engine changes, not inherited.

What happens when a check fails

Nothing dramatic: the document stays with the human, with the reason written down. It is not rejected, it is not deleted — it is simply stopped before it produces effects, exactly like a document brought in by someone without approval rights.

And when the person intervenes and corrects what the machine read wrongly, the difference between the two records itself automatically. Not because someone is being monitored, but because the information is useful: the places where a person corrects often are exactly the places where automatic reading needs to be improved. An intervention that had to be requested through the interface would be forgotten; one that records itself would not.

What no check can verify

The list above is complete for what can be verified from the document. It does not contain, and cannot contain, the one thing that truly matters: whether the transaction actually took place.

A perfectly drawn-up invoice, with the arithmetic balanced, the correct rate, a known partner and no duplicate, may correspond to no delivery at all. All nine checks let it through, and they are right to do so: nothing that can be read on paper says otherwise.

That is why automatic checks shift the threshold, they do not eliminate it. They ensure that the human no longer wastes attention on arithmetic and duplicates, so that it can be kept for the question only they can answer. This concludes the section on incoming documents; the next chapter moves on to those you issue.

Part III · The document · phase 2

Chapter 14

The Invoice You Issue

Until now, documents have been coming into the company. This chapter looks at the ones that leave it, and the difference isn't just semantic: for received documents you're the one who checks, for issued ones you're the one who answers for the content.

The invoice you issue is, at the same time, proof of your right to collect payment, the basis on which the client will deduct the expense and the VAT, and a line in your own VAT return. Three roles in a single piece of paper — which is why it carries the most formal requirements in all of accounting.

Numbering: more important than it seems

An invoice must have a unique number, from a continuous series, assigned in chronological order. The rule seems administrative, and in fact it's what keeps sales from being hidden: if number 101 is missing between invoice 100 and 102, the question asks itself.

Two things follow from this that many small businesses break. Numbers aren't skipped — a wrong invoice is reversed, not thrown away. And invoices aren't issued "dated last month" after others with the current date have already been issued, because that puts chronology and numbering in contradiction.

A program assigns numbers on its own, from series configured once, precisely so these two rules don't depend on someone's attention on a busy day.

What Must Be Written on It

The mandatory content of the invoice is set by the Tax Code, and missing an element isn't a formal slip: it can leave the client without the right of deduction, which instantly turns your negligence into a commercial problem.

The Journal Entry It Produces

An invoice for the sale of goods produces, like any sale, two distinct operations, which chapter 2 already showed: earning the right to collect payment and the goods leaving the warehouse.

This is perhaps where it's clearer than anywhere else why VAT collected is not revenue. The 210 lei enter the company along with the rest of the amount, but they don't belong to it: they're collected for the state and will go to it. A business owner who calculates their gain from the total invoiced systematically overestimates their business.

The Electronic Invoice

From the same invoice comes the electronic file that goes to the ANAF system. It's not a second invoice and isn't filled out separately: it's the same information, in a standardized format, readable by machines.

Not all issued documents have to go there, and the distinction isn't made by preference but by the type of operation. An invoice to a company has a different regime than a receipt to an individual, and a document that isn't an invoice has no business being in the system.

A well-made program deduces the obligation from the type of document chosen — chapter 11 — and not from a hand-written list that would lag behind legislation. And when a document should go there and can't be built correctly, it refuses to issue it empty, rather than sending an invoice with a zero base.

Which Invoice Is the Real One

The question seems scholastic and has a very practical consequence. From the moment the invoice has passed through the RO e-Factura system, the original is no longer the paper you printed nor the PDF you sent by email: it's the electronic file, with the seal applied by the ministry. What you send to the client is a courtesy copy, made so a human can read it.

Two things follow from this, usually learned during an audit. First: if the PDF and the electronic file say different things, the version that counts is the one in the system. Second: an "issued" invoice that hasn't reached the system isn't half-sent — it's an unfulfilled obligation, with its own deadline and its own penalty.

The original is the file sealed by ANAF. The paper and the PDF are copies for people — useful, but they aren't the document.

Who Falls Under the Obligation and Who Doesn't

The scope isn't ticked off, it's read from the data. Three situations, with different regimes, and the distinction between them is made based on who the beneficiary is and where they are established.

A technical detail that's actually a matter of common sense: a CNP has thirteen digits, while a Romanian tax code has at most ten. The two can't be confused, so the program can deduce on its own which situation you're in, without asking you. And when the code is entirely missing and the client is from another country, the invoice remains valid and outside the obligation — it doesn't become overdue.

Five Days, and What "Sent" Means

The current deadline is five BUSINESS days from the issuance of the invoice, but no later than five business days from the legal deadline for issuing it. The two reference points must be tracked together: delaying issuance until the maximum deadline doesn't create an extra five days beyond the limit provided by law. The calendar must know legal holidays, not just weekends.

The second trap is that "I sent it" isn't one state, but three. The file is uploaded and receives a registration number; the system processes it; only then does it return either confirmation or a list of errors. Between the second and third moments, the invoice is neither accepted nor rejected — and that's exactly where programs that say "sent" immediately after upload usually stop.

Rejection errors are almost always about missing data, not amounts: a wrong tax code, an unspecified county, a unit of measure the nomenclature doesn't recognize. That's why they're worth filling in correctly in the partner's record just once, not fixed on every invoice.

Changes Applicable from June 1, 2026

Law 88/2026 redesigned the registers and exceptions of RO e-Factura. Suppliers or service providers who identify for tax purposes by CNP are no longer required to transmit, but may opt in, in which case they are entered in the optional register. For individual farmers under the special regime, certain non-profit entities, religious denominations, homeowners' associations, and cultural institutes, there are separate rules or registers. Before marking an invoice as "outside the obligation," the professional checks both the issuer's status and the beneficiary's entry in the relevant public register.

The scope of RO e-Factura isn't decided solely by the B2B/B2C label. In 2026, what matters is the issuer's tax identification, the place of the operation, the beneficiary's category, and the entry in the public registers.

A rejected invoice is not a sent invoice. The upload number proves you tried, not that you succeeded — the proof is the sealed response.

What Cannot Be Done

An invoice that has reached the system can no longer be modified and can no longer be withdrawn. Correction is done the way it was done on paper, before any electronic system: through a new document, which reverses or adjusts it, and which in turn goes into the system. The rule from this chapter still holds — a wrong invoice isn't deleted, it's corrected with another invoice.

And the invoices you receive travel the same road, in reverse: the supplier uploads them, and you retrieve them from your own space. The practical consequence is that, for purchases, "I received the invoice by email" has stopped being the main source — the document waits for you there regardless, whether or not the supplier also sent it to you separately.

What Comes Next

With this, phase two is complete. Documents come in and go out, each with its own journal entry, checked before they produce effects. The company's accounting now has content.

The next thing that moves, usually as a result of these documents, is money. Phase three, and the next part: treasury — the only place in accounting that can be confronted with an external reality.

Series and Number: Two Things, Not One

Numbering issued invoices seems like a formality until the first audit, when it becomes the first check performed. The reason is simple: from a continuous numbering, one can deduce whether an invoice is missing; from a discontinuous one, nothing can be deduced.

The company sets its own series and number ranges through an internal decision. What it cannot choose is continuity: within a series, numbers follow one another in sequence, and a number once assigned is never reused — not even if the invoice was cancelled.

What turns a document into an invoice

The mandatory elements are not a checklist: each one answers a question that someone else will ask years later. They deserve to be read that way, because that's when it becomes clear why the absence of one leads to the buyer losing the right to deduct.

The invoice is not written for the client, but for whoever will read it four years from now knowing nothing about the transaction. If they cannot reconstruct what happened, the document has not done its job.

In electronic form, the same elements become fields in a structured file, and the validator checks all of them before accepting the invoice. The side effect is useful: formal errors that used to be discovered during an audit are now discovered at issuance, within seconds.

Part IV · The money · phase 3

Chapter 15

Bank and cash

Treasury accounts have a property that no other account in the accounting has: they can be checked against a reality outside the company. The balance of the bank account in the accounting must equal what the bank statement shows — a figure issued by someone else, which cannot be adjusted.

That is why treasury is, at the same time, the simplest part of accounting and the best error detector in the whole company. Simple, because the operations are few and obvious. A good detector, because any mistake shows up immediately, by comparison.

The four accounts and the difference between them

The practical difference between bank and cash isn't the size of the amounts, but verifiability. The bank statement comes from the bank, so it's external evidence. You draw up the cash register yourself, so no external evidence exists — the only possible check is physically counting the money in the drawer.

Collection and payment: two simple entries

Treasury operations are, in the overwhelming majority of cases, simple entries: a cash/bank account and a third-party account.

Payment is not an expense, and collection is not revenue. They extinguish payables and receivables that arose earlier, with another document.

It's the same idea the book started with, in chapter 1: profit and money are different things. Here it takes accounting form. The expense was recorded when the invoice arrived; the payment only extinguishes the debt. Whoever records the expense both at the invoice and at the payment counts it twice — the classic mistake of companies that work "by bank statement."

The cash balance cannot be negative

It's the only statement in accounting that anyone can verify, without any training: you cannot take more money out of a drawer than you put into it.

A negative cash balance is not a treasury problem, but a recording error — every time, without exception. The causes are almost always two: a collection that wasn't recorded, or a payment recorded with a date earlier than the actual one.

It deserves to be treated as an alarm, not an inconvenience. Software that calmly displays "−1,240 lei" in the cash register is, in reality, hiding a missing document — and that document is also missing from VAT and from the result.

The internal transfer trap

When you move money from the bank into the cash register, or the other way around, you have an operation that appears twice: once as an outflow from the bank, in the statement, and once as an inflow into cash, in the cash register. The two documents often arrive separately, and if each is recorded on its own account, the amount gets doubled.

That's what account 581 Internal transfers exists for, a transit account: the outflow from the bank credits it, the inflow into cash debits it, and at the end its balance must be zero. If it isn't zero, one end of the transfer is missing.

It's a good example of an account that represents nothing in the real world — there is no safe called "internal transfers." It exists only to make visible a movement that would otherwise risk getting lost between two documents.

What comes next

The money is recorded. What remains is the part that gives treasury its true value: reconciliation with the bank statement, that is, the only check in accounting that relies on evidence coming from outside the company.

That's the next chapter, and it's the best error-finding tool in the whole book.

The cash register: the required form

The cashier's office has its own register, with a prescribed form, and its rigor comes from a simple fact: it is the only patrimony element that can be counted at any time and for which no external evidence exists.

Hence an organizational consequence: the cash register belongs to the company, not to the person. The money in the drawer isn't the cashier's, and mixing it with personal money — even temporarily, even with the intention of putting it back — makes the balance impossible to reconcile and turns an administrative operation into a matter of liability.

What gets checked automatically, every month

Treasury is the area where most checks can be done without anyone's intervention, because they are all calculated from already recorded data. A program can run them continuously.

It's worth noting that four out of five need no new document: they are questions asked of existing data. This is the typical case where automation doesn't save time but catches something a person wouldn't check every day — and ceilings are usually exceeded on ordinary days, not on the ones when someone is watching closely.

Part IV · The money · phase 3

Chapter 16

Reconciliation, line by line

Bank reconciliation is the confrontation between what your accounting records show and what the bank statement shows. Many treat it as a monthly formality, but it is actually the only check in the entire accounting process that is backed by evidence coming from outside the company.

An experienced accountant does it first, not last. The reason is practical: if the bank and the accounting records match, an entire class of errors is already ruled out, and the remaining checks start from solid ground.

Why the two never match right away

The beginner's first surprise is that balances differ even when everything is correct. This is not a problem: it is the natural gap between the moment an operation occurs and the moment the bank records it.

The distinction that matters is between timing differences and substantive differences. The former disappear on their own; the latter never disappear, no matter how long you wait, because behind them lies an operation that was not recorded, or was recorded incorrectly.

The method: match, then look at what remains

Reconciliation is not done by comparing balances, but by matching operations against each other, one by one. Every line in the statement is looked for in the accounting records, and vice versa.

What matches no longer needs attention. The value of the method lies precisely in what remains unmatched — usually a short list, in which each line is either a timing difference or an error to be found.

You don't compare balances, you compare operations. The balance tells you that a difference exists; the lines tell you what it is.

A program can perform the matching automatically, by amount, date, and party, and the result is that the person receives the short list directly. But their role remains unchanged: to answer the question the machine cannot ask — why this line is missing.

What each remaining line means

The unmatched list reads in two columns, and each side has its own meaning.

The third line deserves emphasis. When two figures differ, the temptation is to pick the one from the statement, because "the bank is right." Most of the time it is, but not always — and the reflex of aligning the accounting records to the statement without looking at the document replaces one error with another, this time invisible one.

What reconciliation catches and what it does not

Reconciliation guarantees that all cash movements are recorded and that their amounts are correct. That is a lot, and it is limited.

It says nothing about the reason for the payment. An amount transferred correctly, matching the statement perfectly, can be recorded against the wrong supplier or as an expense that is not the company's. Reconciliation confirms that the money left, not that it left for a legitimate reason.

This is, again, the distinction from Chapter 4: form can be checked automatically, substance cannot. Reconciliation is the best form-check in accounting — and precisely for that reason, it should not be asked for more than it can give.

How often

The classic answer is "monthly, at closing." The better answer is "whenever the statement arrives," because the effort grows non-linearly with delay: a mismatch from three days ago is cleared up with a phone call, one from four months ago requires searching among hundreds of similar operations.

With cash reconciled, only one topic remains in this part: the thresholds the law places on cash — and why exceeding them is one of the few accounting mistakes that is directly sanctioned. That is the next chapter.

What a machine can match, and where it guesses

Automatic matching of payments to invoices looks like an arithmetic problem and is, in reality, one of ambiguity. It is worth seeing exactly which cases it solves and where it chooses — because that is where the human must step in.

The choice on the third line follows the convention from Chapter 34 — the oldest is settled first — and it is prudent, because it errs by showing the balance as older than it is, not more recent. But it remains a choice: if the customer actually paid the second invoice, the aging schedule will show incorrectly until someone looks into it.

The fourth line deserves defending, because it looks like a gap: a program COULD try all combinations of invoices that add up to 3,500. With ten open invoices there are over a thousand combinations, and with thirty — a billion. More importantly, a combination found is not proof: two invoices that add up to the amount paid does not mean they were the ones paid.

A good automatic matching system is not the one that matches the most, but the one that leaves the human exactly the cases where it would have had to guess. The rest is a short list, not a long report.

What is done with what remains

The unmatched list is not closed "by difference." Each line has a treatment, and the wrong choice moves the problem rather than solving it.

Part IV · The money · phase 3

Chapter 17

Legal ceilings and limits

Cash is the only form of payment that leaves no trace with anyone else. A bank transfer passes through a bank and is recorded by a third party; a banknote handed from one person to another is not. Hence, legislation that restricts cash more strictly than any other accounting operation.

This chapter is short and worth reading even if your business works almost entirely through the bank. The reason is that exceeding the ceilings is sanctioned directly, and the sanction takes no account of good intentions: it doesn't matter that the transaction was real and correctly recorded.

The four thresholds

The last row of the note is the part small businesses miss most often. The ceiling applies per day and per counterparty, not per document. Splitting a payment into several receipts issued on the same day doesn't get around the limit — on the contrary, it's exactly the pattern an audit looks for.

Why the balance ceiling is different from the others

The first three limits concern flows: how much can go in or out in one day. The fourth concerns the stock: how much can remain overnight in the cash register.

This difference has a practical consequence. A business can perfectly respect the payment limits and still exceed the balance ceiling, simply because it collected a lot and didn't deposit the cash at the bank. The check is done at the end of each day, not at the end of the month — and the solution is depositing the surplus, on the same day.

Three ceilings are checked per transaction, one on the end-of-day balance. They can be breached independently.

The check that verifies itself

All four limits have a rare property in accounting: they can be fully calculated from data already recorded. They require no additional information, no judgment, no interpretation.

That's why they have no business being kept in anyone's memory. Software can track them continuously and warn on the day they approach the threshold, not the following month, when the breach has already become a fact. Along with tracking the fiscal thresholds in chapter 6, this is one of the few places where automation prevents a fine, not just saves time.

It's worth requiring of any accounting software that the thresholds be kept in a single, dated place, not scattered through the code. Law 70/2015 has been amended several times since its publication, and a business that has its limits written in one place can update them in a minute.

Cash advances (treasury advances)

When you give money to an employee to buy something for the business, the money hasn't yet left the company's assets: it has merely moved from one pocket to another. Until it is justified with documents, it remains a receivable from the employee, in account 542 Treasury advances.

The common mistake is recording it directly as an expense at the moment the money is handed over. This is wrong for two reasons: the expense doesn't yet have a supporting document, and if the employee brings back change, the change has nowhere to return to.

The balance of account 542 at the end of the month is a list of questions: who is holding unjustified company money, and since when. An old balance there isn't an accounting problem — it's a discipline problem.

What comes next

Phase three is complete. Documents are recorded, cash is tracked and reconciled against the bank statement, and the legal limits are under observation.

Next comes the part where accounting stops being merely a record of paper and starts measuring physical things: the goods in the warehouse, the hours people worked, the wear on a piece of equipment. These are phases four, five, and six, and the first of them is inventory.

The unjustified advance: how it becomes a tax problem

Account 542 seems the most harmless in treasury: money given to an employee, justified a few days later. Over time, however, it becomes one of the most expensive positions in the trial balance — and not because of the amount.

Justification is done with documents, within a deadline set by internal decision. Amounts not justified by the deadline don't remain a convenient receivable from the employee: from a tax perspective they become salary-type benefits, with everything that follows — income tax and contributions, calculated as for a salary.

An old balance in 542 isn't an administrative oversight. It's a quantifiable tax risk that grows with every month no one asks about it.

The check is simple and worth doing monthly, along with the reconciliation: the balance of account 542, broken down by person, with the date each amount was granted. Any position older than the internal deadline is a question worth asking — and a question asked in time costs a phone call, not a recalculation of contributions.

Part V · What you have, what you produce, who works · phases 4–6

Chapter 18

Inventory: in quantity and in value

So far, bookkeeping has tracked paperwork and money. Inventory is the first thing it tracks in two ways at once: how many pieces are in the warehouse and how much they are worth. The two records run in parallel and must match — and when they don't, the truth lies in the physical count, not in the software.

The difference from everything before is that goods move even without a document. A crate can break, a box can go missing, someone can take something "for a moment." That's why inventory is the only part of accounting that has, by law, a mandatory periodic confrontation with physical reality.

Receipt is simple. Issue is not.

When you buy goods, their value is written on the invoice: that's what it cost, no ambiguity. But when you sell, a question arises that the sales invoice does not answer: how much did the piece that left cost YOU?

If you bought the same product three times, at different prices — 10 lei, 12 lei, 11 lei — and you now sell one unit, which of the three left? Physically, you may no longer be able to tell. In accounting terms, you must choose a convention, and the choice changes both the recorded cost and the value of the remaining inventory.

At stable prices, the difference is negligible. At rising prices, FIFO gives a lower cost of issues — hence a higher profit and an inventory valued closer to the current price. Weighted average cost smooths both. There is no "more correct" method: there is one chosen and applied consistently.

The method is chosen once and applied year after year. Consistency is what makes the figures comparable — not the method itself.

Inventory locations (gestiuni): where exactly the goods are

A company with a warehouse and a shop does not have one stock, but two. The same goods can exist in both places, and the transfer between them is neither a sale nor a purchase — it is an internal movement, which affects neither revenues nor expenses.

The division into inventory locations (gestiuni) also has a consequence for accountability: each location has a custodian (gestionar), and at stocktaking, differences are established per location, not per company. A surplus in the warehouse does not offset a shortage in the shop, even if the company nets to zero overall.

The two entries of a sale

Chapter 14 already showed the complete entry for a sale, and it is worth revisiting from the inventory side here. The sale produces the revenue, but the goods leave separately, at their cost — not at the sale price.

Omitting the third line is the classic mistake, and it's dangerous precisely because it's invisible: the trial balance stays balanced, but the inventory appears inflated by 600 lei that no longer exist physically, and the profit appears higher by exactly the same amount.

What follows

Inventory bought and sold is the simple case. When the company doesn't buy goods but produces them, an additional question arises: how much does what you made actually cost? That's the next chapter.

Two records for the same thing

Inventory is the only asset item tracked simultaneously in two places, with different units of measure: the stock card counts pieces, accounting counts lei. They describe the same goods and must say the same thing — but they don't automatically check each other, because they don't speak the same language.

From this comes a practical rule: any transaction that moves inventory must move both records at once. That's why correcting an entry with an impact on inventory is not done through the usual accounting reversal — it would leave the accounting right and the stock card wrong, and the two would silently diverge until stocktaking.

The inventory location (gestiune): where the goods are

A company with a single warehouse can ignore the concept. One with two — warehouse and shop, or central warehouse and points of sale — cannot, because the same goods can exist in both, and the total quantity doesn't say where.

The inventory location also has a purpose that isn't about record-keeping but about accountability: each has a custodian (gestionar), and shortages found at stocktaking are charged to a specific person. Without separate locations, a shortage in one warehouse gets lost in the company total.

Goods valued at sale price

In retail trade, a complication arises that's worth understanding, because it produces three accounts where you'd expect one. A shop cannot keep records at cost: on the shelf is the price with VAT, and the cash register sells at that price.

The solution is to value the inventory right at the sale price, and keep the difference from cost separately, in two adjusting accounts: the trade markup in 378 and the deferred VAT in 4428.

At sale price, account 371 no longer shows how much the goods are worth, but how much is being charged for them. The balance sheet value is obtained by deducting 378 and 4428 — just as fixed assets are read by deducting depreciation.

The consequence that surprises at the first balance sheet: account 371 can show 1,554 lei for goods that cost 1,000, without anything being wrong. Whoever reads the gross balance and draws conclusions about inventory is wrong by the markup plus the tax.

Part V · What you have, what you produce, who works · phases 4–6

Chapter 19

Production: from raw materials to finished product

A company that buys goods and sells them knows exactly what each unit cost it: it's written on the invoice. A company that produces has nowhere to read that figure — it must calculate it, and the calculation is the only operation in accounting where the value of a good is built up, rather than taken over.

Hence the difficulty: the production cost is not a figure found, but one decided. What exactly goes into it is a choice governed by rules, and on it depends both the value of the inventory on the balance sheet and the result for the period.

What goes into the cost and what does not

The distinction has immediate consequences. An expense included in the cost remains in inventory until sale, so it does not affect the current month's result. The same expense left out hits the result now. It's not a matter of preference: inflating the production cost with administrative expenses is one of the oldest ways to show a profit that doesn't exist.

What you put into the cost you defer; what you leave out you pay for now. That's why the rule is not up to the company's discretion.

The recipe: consumption known in advance

For repetitive production, the quantities of materials needed for a product are known: so much wood, so much paint, so many screws. This list is called a recipe or specific consumption, and its practical usefulness is twofold.

First, it eliminates the need to manually enter consumption at each production run: you declare how much was produced, and the materials are automatically deducted from inventory, in the proportions from the recipe. Then — and this is the more valuable part — it makes visible the difference between theoretical and actual consumption. When ten meters of material yield eight products instead of ten, the difference is not an accounting error: it's a process loss, and accounting is the place where it becomes measurable.

The two movements of a production run

A completed production generates two opposite entries: materials leave inventory, the finished product enters.

This pair explains something puzzling at first glance: why production, taken alone, produces neither profit nor loss. The consumption of 600 hit the result, and the entry of the product offset it by exactly that amount. The gain appears only at sale, when the product goes out at a price higher than its cost.

What's next

Purchased inventory and produced inventory both sit in the warehouse, valued according to rules. The question remains that no recorded evidence can answer: are they still there? That's the next chapter.

What happens with what wasn't finished

Production rarely stops exactly on the 31st of the month. At closing there are almost always things started and unfinished — parts on the line, an order half done, a service in progress. They have consumed materials and hours, so they have value, but they are not yet products.

For this, work-in-progress accounts exist: 331 for products and 332 for services. They accumulate the cost incurred up to the closing date, and in the following month they are resumed, when the work continues.

Without them, the month in which a large job begins shows a loss that doesn't exist — all expenses, no revenue — and the month in which it ends shows inflated profit. Work in progress is, therefore, also a matter of timing delimitation, like the one in chapter 28, only it is resolved in inventory, not in accrual accounts.

The recipe: from a product to its consumptions

A company that produces the same thing multiple times does not recalculate what goes into it each time. It's written once — the quantities of raw materials for one unit or for a reference batch — and it's called a recipe, or, in production language, a bill of materials.

Its usefulness is twofold. Practically: when launching an order, consumption is calculated by multiplication, not by memory. From an accounting standpoint: the theoretical cost of the batch can be compared with the actual one, and the difference between them is management information that nothing else provides.

It's worth noting what the recipe does NOT do: it does not impose. Actual consumption may differ — a broken piece, an unused remainder — and the consumption slip records what was actually consumed, not what was planned. The recipe is a calculated proposal, not a record.

The difference between recipe consumption and actual consumption is not an accounting error: it's precisely the information for which it's worth having a recipe.

Part V · What you have, what you produce, who works · phases 4–6

Chapter 20

Stocktaking

Stocktaking is the moment when accounting admits it might be wrong. All the other checks in this book compare figures with other figures; this one compares figures with the world — actually counting what is in the warehouse, in the cash register, in the business.

It's also the only accounting operation that cannot be done from an office. Someone has to go, count, and write down what they found. That's precisely why it's mandatory by law, at least once a year, before the financial statements are prepared.

Why differences appear even when no one is stealing

The first reaction to an inventory shortfall is suspicion. Most of the time, the explanation is more mundane and worth knowing, because it completely changes how it should be treated.

That's why the correct order when dealing with a stocktaking result is: first look for recording errors, then confirm the shortage. A shortfall recorded as a loss, when in fact a delivery note was missing, "resolves" the difference and leaves behind an unbilled sale.

How the result is recorded

Once the real differences have been established, they are recorded, and the accounting treatment is brief and worth knowing.

Charging is not automatic and not mandatory. It presupposes that liability has been established — which requires a procedure, not just a difference in a table. A shortfall charged without grounds quickly turns into a labor dispute that the company loses.

Stocktaking does not end with a figure, but with a minutes report: what was counted, what came out differently, and what was decided for each difference.

What is actually inventoried

The word brings to mind the warehouse, but stocktaking concerns the entire patrimony. Goods and materials are counted, but the cash register and balances with partners are also checked — through confirmation requests — and the physical existence of fixed assets is verified.

The last part is the one most often skipped and the most instructive. A company that inventories its fixed assets often discovers equipment scrapped long ago that still appears in the records and on which depreciation continues to be calculated — or, conversely, equipment used daily and never recorded.

What comes next

With stocktaking, the part about things comes to an end. Next comes the part about people: the payroll, the most heavily regulated monthly operation in all of accounting.

The procedure, because without it the result cannot be defended

Stocktaking has a form prescribed by regulations, and the form is not bureaucracy: it's the only thing that turns a count into evidence. A shortfall found without procedure cannot be charged to anyone, and a surplus cannot be justified during an audit.

The order has a logic worth seeing: the first three stages produce facts, the fourth produces differences, the fifth produces explanations, and only the sixth touches the figures. Whoever jumps straight to the last one ends up with entries lacking a written reason — exactly what cannot be defended two years later.

Offsetting and shrinkage allowances

Two mechanisms reduce differences before they are recorded, and both apply restrictively, not as a method for making the total come out even.

Offsetting surpluses against shortfalls is allowed only between confusable item types, from the same group, within the limit of the smaller quantities, and only if the commission justifies the confusion. It is not an equalization of the total: two different products that cancel out in value remain two distinct problems.

Shrinkage allowances cover natural losses from handling, transport, or storage — evaporation, breakage, drying — within limits established by Government decision, by category of goods. They are granted only when an actual shortage is found and only up to the legal limit: beyond it, the rest remains a chargeable or non-chargeable shortfall.

Neither offsetting nor shrinkage allowances explain a shortage. They only acknowledge that part of it was foreseeable. The rest requires an answer.

Not just inventory

The word "inventory" brings to mind the warehouse, but the obligation concerns the entire patrimony: fixed assets, receivables, liabilities, cash and cash equivalents, even off-balance-sheet items.

Stocktaking of receivables is not done by counting, but by confirmation: the partner is asked to confirm the balance. It's the procedure that uncovers invoices they never received, payments allocated incorrectly, and receivables that no longer exist. For fixed assets, the count confirms that the item in the register still physically exists — and, in the case of undeclared disposals, that's where it shows up.

The result of the entire operation is recorded in the inventory register, which is a mandatory register, alongside the journal register and the general ledger. Without it, financial statements are prepared on figures that were never checked against reality.

Part V · What you have, what you produce, who works · phases 4–6

Chapter 21

The Payroll

Payroll is the most regulated monthly operation in accounting and, for a business owner, the most counterintuitive. The reason is simple: between what an employee costs the company and what they actually receive, there is a gap that no one gets right on the first try.

This chapter traces the path from gross to net and, more importantly, shows where the difference goes — because part of it is not deducted from the salary but added on top of it, as a burden on the company.

The path from gross to net

Gross salary is the starting point. From it, the employee's contributions and income tax are withheld, in this order — the order matters, because the tax is calculated on what remains after contributions, not on the gross amount.

Out of 5,000 gross lei, in the August 2026 example the employee receives 2,982. The difference of 2,018 does not stay with the company: 1,750 are the employee's contributions, remitted to the state, and 268 is income tax. The company gains nothing from the withholding — it is merely the intermediary obligated to remit it.

What the employee actually costs

This is business owners' most common calculation mistake. The cost is not the gross salary, because on top of it the company also owes a contribution of its own.

The gross is not the cost, and the net is not the gross. Three different figures, which a business owner confuses until they see them laid out side by side.

The two entries of the month

In accounting terms, payroll produces two distinct operations, at different times: recording the amounts owed, at the end of the month, and their actual payment, usually in the following month.

The distinction echoes the idea from Chapter 15: the expense is recorded when the obligation arises, not when the money leaves. December's salary is December's expense, even if it is paid in January — otherwise the year's result would be off by a full month of wages.

Why payroll isn't calculated by hand

The example above is the simplest case possible: a full salary, a full month, no absences. Reality brings paid leave calculated on the average of the last few months, sick leave paid partly by the company and partly by the fund, part-time schedules with their own rules, bonuses, deductions, advances, meal vouchers.

Each of these has its own formula and interacts with the others. That's why payroll is the area of accounting where manual calculation is riskiest and automation most valuable — with one condition: the rates and thresholds must come from a dated source, not be hardcoded, since they change often.

What's next

A correctly calculated payroll is a necessary condition. It is not, however, sufficient: some of the employee's entitlements are not granted automatically, but only if someone requests them. That is the subject of the next chapter, and it is the most expensive chapter of this part.

Paid leave: same money, different label

Vacation days are not paid with the salary but with an allowance calculated on the daily average of gross earnings over the last three months. And from this follows something surprising: at a constant salary, a month with vacation looks identical to one without.

This equality, however, only holds at a constant salary. If in the last three months the employee had bonuses or overtime, the average is higher than the base salary — and the month with leave costs the company more than a regular one. This is why the allowance is calculated on the average and not on the current salary: the law protects the employee's usual income, not that of the month in which they leave.

Sick leave in 2026: a reduced day and two funding sources

For certificates issued between 1 February 2026 and 31 December 2027, OUG no. 91/2025 introduced a temporary rule: the allowance is calculated and paid with a one-day reduction. For temporary incapacity for work under the normal regime, the employer bears the allowance from the second day up to and including the sixth day of incapacity, and FNUASS from the day following those borne by the employer. The working days within these intervals are paid; calendar days determine the position within the leave period. There are exceptions and allowances borne entirely by the fund, which must be checked by code.

For the five-day certificate starting on Monday, five working days are missing from the salary, but only four are compensated under this temporary regime. For the certificate starting on Thursday, the weekend occupies two of the calendar days within the employer's interval, leaving only two working days compensated. For the ten-day certificate, working days starting from the seventh calendar day fall under FNUASS's responsibility.

For certificates issued between 1 February 2026 and 31 December 2027, do not automatically apply the historical rule 'the employer bears the first five days.' Identify the reduced day, count the intervals in calendar days, and pay only the working days corresponding to each source.

Separately from the funding source, the allowance percentage, calculation base, and CAS, CASS, CAM treatment depend on the code and duration. For an uninterrupted leave, the reduction is applied only once per episode, regardless of the number of certificates, per Law no. 64/2026 and the norms updated by Order no. 506/2026. From 1 July 2026, an extension into the following month that moves the episode into a higher percentage bracket may trigger a recalculation of the difference for the previous month; the difference is included and declared in the month of recalculation, per OUG no. 89/2025.

The error is not visible through the payroll balancing correctly: the software can perfectly calculate an outdated rule. An effective check compares the certificate's date with the rule's version, then reconciles the days and amounts borne by the employer and by FNUASS with D112 and with the receivable to be recovered.

Part V · What you have, what you produce, who works · phases 4–6

Chapter 22

Rights that are lost if you don't claim them

This chapter was written after an incident, and the incident deserves to be told, because it is more instructive than the rule it illustrates.

In an accounting program, the employee record had a field called "Dependents". The field appeared filled in with zero — the digit zero sat there, gray, in the box. In reality it was empty: the zero was only a visual placeholder, not a value. And the program only granted the personal deduction if that field had been filled in.

The result: employees whose records had been filled in normally, without anyone touching that field, received no deduction at all. In August 2026, at a gross salary of 5,000 lei with no dependents, that means 57 lei extra tax each month — 684 lei per year per person, paid to the state without being owed and deducted from the net pay of people who had no way of finding out.

A right granted "if requested" is a right that gets lost. Not out of bad faith, but because no one checks what wasn't flagged to them.

What the personal deduction is

The personal deduction is a portion of salary income that is not taxed. It's not an amount received on top, but a reduction of the tax base — its cash effect is 10% of the deduction's value, i.e., exactly the tax that is no longer paid.

It is granted at the basic job, meaning the main place of employment, and its size depends on two things: how many dependents the employee has and how large their salary is.

At a salary of 5,000 lei, with the minimum wage of 4,325 lei applicable from July 2026, the basic deduction is 570 lei with no dependents — and the tax drops from 325 to 268 lei. With two dependents, the deduction rises to 1,000 lei, and the tax drops to 225 lei. The calculation is automatically redone when the minimum wage changes; a figure copied from the previous semester immediately becomes wrong.

Zero doesn't mean "none"

The core of the incident lies in a confusion that appears often, not just in software: between "I have zero dependents" and "I haven't filled in the field."

These are different things. An employee with zero dependents is entitled to the basic deduction — the smallest of the options, but not zero. An unfilled field says nothing about their situation; it only says that someone wrote nothing there.

A system that treats them the same silently chooses the option that disadvantages the person. And if no one looks at the "deduction" line on the payslip — and usually no one does, because the net pay seems plausible — the error can last for years.

How to check, in two minutes

The check requires no specialized knowledge. Open the payslip for the month and look at the deduction column.

The last column of the table also contains the only case where zero is the correct answer: a second job. The deduction is granted only once, at the basic job — and if an employee receives it at two jobs at once, the difference will be reclaimed from them.

The general rule, beyond this case

The personal deduction is the costliest example, but the pattern is broader. Anywhere a right is granted "upon request" or "if filled in", it is lost for those who don't know they must ask — that is, exactly for those who would need it most.

That's why, for any system that calculates entitlements, the right question isn't "what happens if I check the box", but "what happens if I check nothing at all". A poorly chosen default costs real money, silently, on every run.

The curve, in figures

The chapter said the deduction decreases proportionally with salary. It's worth seeing the whole curve, because it produces an effect that neither the employee nor the business owner suspects.

The effect nobody sees

Because the deduction decreases along with salary, a raise granted within the phase-out band reaches the employee in a smaller proportion than one granted above it. It's not an impression — it can be measured.

The difference is small on a single raise and becomes visible in a negotiation: an employee earning 5,000 lei gross who asks for "another 200 in hand" needs a bigger raise than one earning 7,000, even though the request sounds identical. Anyone calculating the required gross with the usual formula — net divided by 58.5% — gets it wrong exactly in the band where most salaries fall.

Between the minimum wage and the minimum wage plus 2,000, each additional leu of raise brings less into hand than above the threshold. It's the only zone in the pay scale where the effective rate is higher than the legal one.

When the situation changes during the year

The deduction is not set once and for all. It depends on the number of dependents, and that changes — a child is born, a parent becomes a dependent, a child turns 26 or finishes their studies.

The rule is that the change takes effect starting THE MONTH FOLLOWING the one in which it was communicated, based on the employee's declaration and supporting documents. Two practical things follow from this: the change must be communicated, not assumed; and a delay in communication is not retroactively recovered indefinitely.

The consequence for the employer is a filing obligation, not a calculation one: the own-responsibility declaration and the documents stay in the personnel file, since they are the basis for any deduction above the basic level. Without them, in an audit, the deduction granted becomes tax owed.

Part V · What you have, what you produce, who works · phases 4–6

Chapter 23

Fixed Assets and Depreciation

A pack of paper bought today is consumed this month. A piece of equipment bought today is consumed over five years. Accounting must distinguish between them, otherwise the result for the month in which you bought the equipment would show a loss that doesn't actually exist.

The mechanism that makes this distinction is called depreciation, and it is the best example in the whole book of the principle stated in chapter 1: an expense is recognized when the asset is consumed, not when the money leaves.

What Is a Fixed Asset

For accounting purposes, an item is recognised as a fixed asset when the company controls it, expects economic benefits from its use and intends to use it for more than one year; the capitalisation threshold is set by accounting policy with regard to materiality. The tax concept is narrower: from 2026, a depreciable fixed asset must have a tax value of at least 5.000 lei under Article 28(2)(b) of the Tax Code, a limit that may be updated annually for inflation by government decision. Document the two tests separately.

A 2.000 lei telephone used for three years may be expensed immediately under the company’s approved capitalisation threshold, while remaining a controlled item tracked in the inventory records if policy requires. A 200.000 lei stock of merchandise is not a fixed asset, regardless of amount: it is held for sale rather than use. Value alone does not determine classification.

Depreciation: Spreading the Cost Over the Useful Life

Depreciation spreads the value of an asset over the months it will be used. A piece of equipment worth 60,000 lei with a useful life of five years means sixty months of 1,000 lei each — and those 1,000 lei become the expense for each month.

The entry is the same every month: debit 6811 Operating Expenses regarding Depreciation and credit the corresponding depreciation account from group 28. No cash account appears — and it's important to understand why.

Depreciation is an expense that is not paid. The money left only once, at purchase; depreciation merely recognizes, gradually, that the asset is being consumed.

The practical consequence is that a company with high depreciation can have low profit and cash in the bank — and vice versa. This is the most common case where the business owner asks where the money is if profit is low, and the answer is: in the equipment bought last year.

When It Starts

Depreciation begins in the month following the asset's commissioning, not the month of purchase and not from the date of purchase. Equipment received on March 20th is first depreciated in April, with the full month's amount.

Commissioning is a distinct moment from purchase, and the distinction has its own account: 231 Assets Under Construction. Equipment bought in March and installed in June sits in 231 until June and is not depreciated at all in the meantime — because it has not been used.

The Useful Life Is Not Chosen by Preference

The normal useful life comes from an official catalog, approved by government decision, which covers practically all categories of assets. The catalog does not give a single figure, but a range — for example, eight to twelve years — and the choice within the range belongs to the company.

It's worth understanding what's at stake in that choice. A shorter useful life means higher depreciation, hence higher expense, hence lower profit and lower tax — but also an asset that disappears from the balance sheet more quickly. It's a decision with an effect spanning years, made only once, at commissioning, and one that no one ever revisits.

That's why a useful life typed in from memory is one of the most costly silent mistakes in accounting: it produces no visible error, it unbalances nothing, and it affects the result every month, for years on end.

Three Methods, Same Total

The law allows three calculation methods. All three depreciate exactly the same value — they differ only in pace.

Choosing the method is, therefore, a decision about where you want the profit to fall, not about how much you depreciate. The straight-line method is the default and most commonly used; the other two are chosen deliberately, usually for tax reasons.

Special 2026 rule: superaccelerated depreciation

For new assets in subgroup 2.1 — technological equipment, machinery, tools and production installations — and subgroup 2.4 — animals and plantations — acquired or produced and commissioned in 2026, Article 28(8^1) of the Tax Code permits superaccelerated depreciation: up to 65% of the tax value in the first year, with the remaining value allocated over the remaining normal useful life. This is a temporary tax facility; the fixed asset register must separately identify eligible assets and their commissioning dates.

The 5.000 lei threshold and depreciation of up to 65% are TAX RULES applicable from 2026. The ACCOUNTING recognition policy remains a documented decision of the entity; the two records may legitimately carry different values.

The Two Depreciations That Don't Match

Here we encounter one of the few situations where accounting keeps two parallel calculations for the same asset. Book depreciation reflects actual consumption and enters the accounting result. Tax depreciation is calculated according to the rules of the Tax Code and enters the profit tax calculation.

Usually they match. When they don't — because the company chose one method for accounting and another for tax purposes, or because the law caps the deduction — the difference is adjusted in the profit tax calculation, without affecting the accounting entries.

The most common case is the car: for passenger vehicles with no more than nine seats, tax-deductible depreciation is capped at 1,500 lei per month. A car worth 200,000 lei is depreciated in full for accounting purposes, but for tax purposes only up to this cap — the rest remains a non-deductible expense, every month, for the entire useful life.

The End: Disposal or Sale

An asset leaves the company's assets in two ways, and the two are not recorded the same way. Scrapping means taking out of use an asset that is no longer of use to anyone. Disposal means selling it — with a price, an invoice, and VAT.

In both cases, the operation has two halves, and the second is the forgotten one: removing the asset from the records. Accumulated depreciation is canceled and the remaining undepreciated value is expensed. Without this part, the asset remains on the balance sheet even though it no longer exists, and the result of the disposal comes out wrong.

What Comes Next

An asset can be in a company's use without having been purchased. The most widespread form is leasing — and the monthly leasing installment is, in accounting terms, the best example of a payment that is not an expense. It is the last chapter of this part.

The Car: Three Limitations That Get Confused

The passenger vehicle is the asset with the most complicated tax regime in a small company's entire accounting, and the reason is that the law limits it three times, in three different places, with three different bases. Anyone who confuses them makes mistakes in both directions.

The first limitation has an effect that many miss: the undeducted VAT is not lost, it enters the COST of the asset. The car's entry value is not 200,000, but 221,000 lei — and depreciation is calculated on this value.

The 1,500 lei cap is MONTHLY and is not carried forward. A year with eleven months of depreciation has eleven caps, not twelve — and the calculation "12 × 1,500" gives a deduction larger than the legal one.

All three limitations are lifted if the vehicle is used EXCLUSIVELY for business purposes, and the burden of proof lies with the company: a trip log, or other records showing the routes traveled. Without it, the limitation applies — not as a penalty, but because the law presumes mixed use. It's one of the few situations where a daily record is worth, in money, more than any optimization.

Part V · What you have, what you produce, who works · phases 4–6

Chapter 24

Leasing

Leasing is the arrangement through which a company uses an asset it has not purchased. In accounting terms, this is where it becomes clearest that a monthly payment is not, in itself, an expense — and treating it as one is among the most common mistakes in small-business accounting.

Everything depends on a single thing: what type of leasing it is. And the answer isn't found in the name of the contract, but in what happens at the end of it.

Financial or operating

The third row is the one that clarifies everything. Under financial leasing, you bear practically all the risks and benefits of use: you pay the installment regardless of whether the asset is still useful to you, you insure it, you maintain it, and at the end you buy it for a nominal value. Legally it isn't yours; economically, it is.

Under operating leasing, by contrast, you're in the position of a tenant. There, the installment really is an expense, recorded in account 612 Expenses with royalties, leases, and rents, and the asset doesn't appear anywhere on your balance sheet.

Entering into financial leasing

When the contract is signed, two things are recorded at once, at the financed value: the asset, in its fixed-asset account, and the liability to the financier, in account 167.

This is an entry that seems surprising at first glance — both an asset and a liability appear on the balance sheet, even though nothing has been paid and you're not the owner. That's exactly the point: the balance sheet shows that you use an asset of that value and that you owe that amount. Both are true.

From this moment on, the asset is depreciated in your books, following the rules from the previous chapter. Depreciation is your monthly expense for using it — and it's the only expense related to the asset itself.

The installment: three components, three destinations

This is the core of the chapter. The monthly installment looks like a single amount, but it contains three completely different things, and each goes to a different place.

Out of the 3,011.09 lei paid, the month's expense is 750 lei. The rest is debt repayment and recoverable tax. To this is added the depreciation of the asset, which is a separate expense and is not paid out.

Under financial leasing, the month's expense is the interest plus the depreciation — not the installment. The installment is, for the most part, repayment of a debt.

The mistake that gets made and its consequences

The widespread reflex is to record the entire installment as an expense, most often in 612, as if it were rent. It seems harmless — the money really did go out, the amount is real — and it produces four effects at once, all of them wrong.

First, the year's expense is inflated by the principal portion, so profit is understated. Then, if the asset is nevertheless depreciated, the same value enters expenses twice: once through the installment, once through depreciation. Third, the balance sheet shows neither the asset nor the liability — two omissions that make the company look smaller than it is. And finally, the interest disappears from the category of financial expenses, which distorts any analysis of the cost of financing.

This is, once again, the pattern encountered in Chapter 11: not a calculation error, but a misclassification. The trial balance closes perfectly in all four cases.

The schedule is the source of the figures

The split of the installment between principal and interest is not estimated and not divided into equal parts: it comes from the repayment schedule attached to the contract. With a constant-annuity schedule, the first installments are almost entirely interest, and the last ones almost entirely principal — even though the amount paid is identical.

In the example above, over the life of the contract, 100,000 lei of principal and 19,448 lei of interest are paid. A company that records the installment as "half-half" is wrong every month, in both directions, and only reaches an approximately correct total by chance at the end.

This is why the schedule is worth entering into the accounting software only once, when the contract is signed, with each installment then taken directly from it. It's an operation that takes two minutes and saves forty-eight monthly decisions made from memory.

What comes next

This chapter closes Part Five. The company now has records of everything it owns, produces, and pays its people: inventory counted and valued, production with cost, stock reconciled against reality, wages calculated correctly, assets depreciated over their useful life.

Next comes the part where all of this turns back toward the state: value added tax, returns, and their deadlines. This is phase seven, and it's the one where mistakes cost the fastest.

Part VI · What doesn't come from a document · phase 7

Chapter 25

Adjustments: the accounting of time

So far, every entry has had a document behind it: an invoice, a bank statement, a payroll, an inventory count report. This part is about the entries that have no such thing.

Not because the document is missing, but because the operation did not take place between the company and someone else. It took place between the company and time. And time does not issue invoices.

Why they are needed

Accounting divides a continuous activity into periods with a beginning and an end. The division is artificial — customers do not stop buying on December 31 — but necessary: without it, you could never say how much the company earned.

The problem is that the cut always falls through the middle of something. Insurance paid in October also covers months of the following year. A client who stops responding does not turn into a loss on a specific date. A euro collected in March is not worth the same in December.

Adjustments (regularizări) are the entries through which the accounting period becomes true. Without them, the trial balance closes perfectly and the result is false.

The four families

There is also a fifth category, less discussed: expenses for which the service was provided but the invoice has not arrived. These are not adjustments in the strict sense, but early recognitions, and are recorded in accounts 408 Suppliers — invoices not yet received and 418 Customers — invoices to be issued.

The provision: the probable obligation

A provision is set up when the company has a probable obligation, of estimable size, arising from an event that has already occurred — but whose exact amount or due date is not yet known.

The classic example is litigation. A former employee has sued the company, the lawyer assesses the chances of loss as high, and the probable amount is 30,000 lei. There is no invoice, no ruling, no document that requires payment. There is, however, a real risk, born from a past fact.

The entry debits 6812 Operating expenses regarding provisions and credits 151 Provisions. When the litigation ends, the provision is reversed — 151 is debited against 7812 — regardless of whether the company won or lost; if it lost, the actual payment is recorded separately, at its real value.

It is worth noting that a provision does not set money aside. There is no blocked bank account, no liquid reserve. The provision only recognizes, in the result, an obligation that already exists — just as depreciation recognizes a consumption that has already occurred.

What makes them dangerous

Adjustments have three properties that make them the preferred spot for errors, and it's worth naming them in advance, since they recur in all the chapters that follow.

First, nothing triggers them. No paper arrives to remind you of them. If nobody does them, they simply don't get done — and nothing signals the absence.

Then, their absence unbalances nothing. The trial balance remains perfectly closed without any adjustment. It's a variant of the trap from chapter 2: the balance check says nothing about completeness.

Finally, all of them require judgment. What is the probability of losing the litigation? Will this client still pay? How many months of the insurance belong to next year? These are questions a person answers, not a formula — and the answer must be documented, because otherwise no one can reconstruct it two years later.

What follows

The four chapters that follow take up, in turn, the adjustments that appear at almost any company. The first is the most mechanical of them and, precisely for that reason, the one where mistakes slip in most easily: bringing foreign currency to the closing exchange rate.

Part VI · What doesn't come from a document · phase 7

Chapter 26

Foreign Currency Revaluation

A receivable of 10,000 euros is not a fixed amount. It's a fixed amount in euros and a variable amount in lei — and Romanian accounting keeps records in lei.

The difference between the two is called exchange rate difference, and it is income or an expense just as real as any other. The company did nothing, sold nothing, paid nothing — and yet it gained or lost money.

How the Mechanism Works

An invoice issued in March, for 10,000 euros at an exchange rate of 4.97 lei, enters the accounts as a receivable of 49,700 lei. If by December 31 the client still hasn't paid, and the closing rate is 5.08, the same receivable is worth 50,800 lei.

The 1,100 lei is financial income, recorded by debiting the receivable account and crediting account 765 Income from Foreign Exchange Differences. The receivable on the balance sheet thus reaches its correct value at the closing date, and the gain appears in the result.

The Sign Rule, and Why It Gets Wrong

Here lies the trap of this chapter, and it has cost real companies and real software. The same rising exchange rate produces opposite effects on receivables and on liabilities.

The logic is simple once stated: if you are owed 10,000 euros and the euro becomes more expensive, you will receive more lei — you have gained. If you owe 10,000 euros and the euro becomes more expensive, you will pay more lei — you have lost.

The mistake doesn't come from misunderstanding the rule, but from applying it mechanically, with a single formula, to all accounts at once. A program that calculates the difference as "new value minus old value" and records it as income whenever it's positive will record losses on liabilities as gains — with the sign reversed, every year, silently.

The sign of the difference is not read from the figure, but from the nature of the item. The same positive difference is a gain on a receivable and a loss on a liability.

What Gets Revalued and What Doesn't

The second source of errors is scope. Not everything purchased in foreign currency gets revalued — on the contrary, most of it doesn't.

The distinction deserves to be understood, because it is more than a convention. Goods purchased for 1,000 euros remain the same goods regardless of the exchange rate; their cost is what was paid at that time, and the gain will show up at sale. A receivable of 1,000 euros, on the contrary, is exactly the right to receive 1,000 euros — so its value in lei changes along with the exchange rate.

Advances are the surprising case. An advance paid to a foreign supplier looks like a receivable, but it's not the right to receive money, it's the right to receive goods — so it stays at the exchange rate from the day of payment. A program that revalues them produces differences that don't exist.

Practice

Revaluation is an operation that shouldn't consume thinking time. Foreign currency balances are known, the closing rate is public, the sign rule is fixed. A well-made program proposes the list of accounts to be revalued, with the foreign currency balance and the rate, and the accountant confirms.

What cannot be automated is checking the scope: if an account entered the list because it was touched at some point by a foreign currency transaction, but its balance has long been in lei, revaluing it produces an invented difference. A minute spent reading the list before confirming is worth more than any check done afterward.

What Happens at Settlement

The revaluation at closing is not the final act. When the receivable is collected or the liability is paid, a second difference appears — and this is where the mistake is made that doubles everything recorded before.

The difference at settlement is calculated relative to the LAST revalued value, not to the rate on the day of the invoice. That rate has already been superseded once, at closing, and the difference from then is already in the result.

The typical mistake is calculating the difference at collection relative to the invoice rate: 51,200 − 49,700 would give 1,500 lei recognized a second time, in the second year, on top of the 1,100 already recorded in the first. The result of the second year would be inflated by 1,100 lei, while the first would remain correct — so nothing would offset and nothing would signal an error.

After a revaluation, the initial exchange rate is no longer a reference point. The book value of the receivable is the revalued one, and any subsequent difference is measured from there.

How Often Revaluation Occurs

The rule states "at every period-end closing for which financial statements are prepared," and the wording is broader than it seems: not just on December 31.

A company reporting quarterly revalues quarterly. One that prepares interim statements for a bank or for a shareholder revalues then too. Omitting interim closings doesn't change the annual total, but it shifts the result between periods — and if those periods were reported to someone, the shift has consequences.

And for companies paying microenterprise income tax, there is a specific rule, easy to miss: favorable exchange rate differences enter the taxable base not line by line, but through the NET favorable difference, cumulated, at the end of the quarter. Calculating per transaction overestimates the base.

Part VI · What doesn't come from a document · phase 7

Chapter 27

Adjustments for Doubtful Receivables

An invoice issued and uncollected sits on the balance sheet as an asset, at its full value, no matter how much time has passed since its due date. So the balance sheet states that the company is owed that amount — and at some point, that statement stops being true.

This chapter is about that moment and about how it is recognized in accounting, without erasing anything. Because a doubtful receivable does not disappear: it remains owed. Only the likelihood of collecting it has diminished.

Three States, Not Two

The usual reflex divides receivables into collected and uncollected. Accounting needs an intermediate state, because between "the client is late" and "the money is lost," months or years usually pass.

The first transition is a simple reclassification: 4118 is debited, 4111 is credited, for the full amount of the receivable, VAT included. No income statement account is touched — the receivable is the same, only moved into a drawer that says it's problematic.

The Adjustment: The Probable Loss, Not the Certain One

Reclassification changes nothing in the result, but the balance sheet still shows a full receivable. The step that corrects this is called an impairment adjustment.

The adjustment does not erase the receivable. It places alongside it, on the balance sheet, a negative value stating how much of it will probably not be collected. The receivable remains 10,000 lei; the adjustment of 10,000 lei says that, realistically, its value is zero.

Account 6814 Expenses regarding adjustments for the impairment of current assets is debited, and 491 Adjustments for the impairment of receivables is credited. The expense appears in the result of the period in which the loss became probable — not in the period in which it will be definitively established.

The adjustment recognizes that the value of a receivable has decreased. It does not recognize that it has disappeared — that's a different step, and often it never comes.

What Happens If the Client Pays After All

It happens more often than one might think, and accounting must be able to reverse it. This is where you see why the adjustment is a better construction than writing off the receivable.

The adjustment is reversed: 491 is debited, and 7814 Income from adjustments is credited. The expense recorded last year is offset by income this year, and the receivable is collected normally. Nothing was destroyed and nothing needs to be reconstituted.

If, on the contrary, the receivable proves definitively lost, it is written off through 654 Losses from receivables and various debtors, and the adjustment previously set up is reversed at the same time. And if, years later, a debtor considered lost pays, the amount is recorded as income in 754 Income from reactivated receivables.

The Numbers, in a Case

The last line of the entry is the one that clarifies the purpose of the entire mechanism. The loss is not recognized twice: once through the adjustment, when it became probable, and again through the write-off, when it became certain. The second time, only the label changes.

The Complete Monograph

Placed one below the other, the four operations cover the entire path. They deserve to be read at the level of lines, not just accounts — the third one has two, and it's precisely the second one that gets forgotten.

The net effect of the two lines at write-off is zero on the result — and this is the check that confirms the entry is complete. If the write-off changes the profit, either the adjustment had not been set up, or the second line is missing.

The expense is recognized only once, when the adjustment is set up. The write-off does not add a loss — it only changes its label from "probable" to "certain."

The VAT Remains Collected

This is the part that comes as an unpleasant surprise. The company issued the invoice, collected the VAT, and paid it to the state. The client never paid. The VAT, in principle, remains owed.

The law allows a reduction of the tax base for uncollected receivables, but only in expressly enumerated situations — essentially, in insolvency, bankruptcy, or debtor reorganization procedures — and only within a limited timeframe. It is not a correction made simply because a client fails to pay.

The practical consequence is worth stating bluntly to a business owner: for an uncollected invoice of 12,100 lei, the company loses not 10,000, but 12,100, because the 2,100 lei of VAT has already been paid to the state from the company's own money.

Judgment, Not the Calendar

The 270 days in the fiscal law are a deductibility threshold, not a definition of a doubtful receivable. In accounting terms, a receivable becomes doubtful when there are reasons to believe it will not be collected — and a client who enters insolvency thirty days past the due date is doubtful immediately, not nine months later.

That is why the decision remains that of the accountant and the administrator together, and it is put in writing. In an audit, the question is not whether the adjustment was correct, but what it was based on — and an answer given from memory, two years later, is no substitute for documentation.

Why the Deductible Record Must Be Kept from the Start

The adjustment has an asymmetric tax regime, and the asymmetry is paid for two years later, if nothing was recorded at setup.

At setup, only 30% is deductible. At reversal — when the client pays or the receivable is written off — the income is taxable in the SAME proportion: 30% taxable, 70% non-taxable. The two match only if it is known how much was deductible at setup.

Without the analytical record for the two portions, one of two things happens at reversal: either all the income is declared as taxable, and the company pays 16% on 8,470 lei that was not owed, or nothing is declared, and the difference is discovered during an audit. Both cost more than one extra column in a table.

The Adjustment Is NOT a Definitive Loss

This is worth stating explicitly to a business owner, because the term "impairment adjustment" sounds like giving up: the receivable remains owed, and the right to pursue it does not expire. The adjustment is a statement about the PROBABILITY of collection, not about the right to collect.

The practical consequence: after the adjustment is set up, pursuit of the receivable continues — demand letters, court action, enforcement. And if these efforts succeed, the adjustment is reversed and the money comes in normally. A company that treats the adjustment as a loss stops pursuing exactly the receivables it recognized as at risk.

The adjustment changes the figure on the balance sheet, not the relationship with the client. The receivable remains, pursuit continues, and if the money comes in, everything recorded is reversed.

Part VI · What doesn't come from a document · phase 7

Chapter 28

Time allocation

An insurance policy of 12,000 lei paid on October 1st, for twelve months, is not October's expense. Nor is it this year's expense. Of it, three months — 3,000 lei — belong to this year, and the rest belongs to next year.

The mechanism through which this split is made is called time allocation, and it is the most direct application of the principle with which this book began: the expense belongs to the period in which the service is consumed, not the one in which the invoice is paid.

The two accounts

It's worth noting that 471 is an asset, even though it looks like an expense. The company has paid, but in exchange it has a right: insurance coverage for the coming months. That right has value, so it sits on the balance sheet as an asset until it is consumed.

How it is recorded

There are two movements, at different moments. When the invoice is received, the entire amount goes into 471 — not into the expense account. Then, each month, the portion belonging to that month is moved from 471 into the appropriate expense account.

Without this split, the current year's result would be lower by 9,000 lei, and next year's higher by the same amount. Taken together across two fiscal years, nothing changes — but neither one, taken separately, would be accurate.

The trap: the monthly transfer doesn't happen by itself

This is the most frequent flaw in this area, and it's important to state it explicitly, because it's counterintuitive: the entry into 471 is easy to remember — it comes with the invoice. The monthly transfer comes with nothing.

No document arrives in November saying "move 1,000 lei from 471 to 613". If no one does it, the amount stays in 471, and the account swells year after year with insurance, subscriptions, and rent that were consumed long ago.

The balance of account 471 is a list of promises. If it isn't shortened every month, it becomes a list of forgotten expenses.

The check is simple and worth doing annually: every amount remaining in 471 must correspond to an identifiable future period. If you can't say which month it covers, it's no longer a prepaid expense — it's an expense you forgot to recognize.

Don't confuse it with advances

This is a frequent confusion, since both concern money that moves before the performance takes place. The distinction lies in what exactly has already been established.

The symmetrical case

Account 472 works identically, in mirror image. A company that collects an annual subscription of 12,000 lei in January hasn't earned 12,000 lei in January: it has received money and taken on the obligation to perform for a year.

The amount goes into 472, and each month 1,000 lei is transferred to revenue. Forgetting the transfer has the opposite effect here — understated revenue and a fictitious liability on the balance sheet — but with the same cause: no document arrives requesting it.

How finely it's worth splitting

The rule says every expense is allocated across the periods it covers. Applied literally, it would require splitting a 200 lei subscription over twelve months — that is, eleven entries of 16.67 lei each, for an amount that changes nothing.

This is what the materiality threshold exists for: an amount below which the effort exceeds the information gained. The threshold is not set numerically by law — it is established by the company and written into the accounting policies manual.

The threshold must be WRITTEN, not kept in mind. A threshold decided case by case isn't a threshold: it's an inconsistency, and it violates the principle of consistency of methods.

A practical criterion, often used: allocate what exceeds the threshold AND crosses from one fiscal year into another. An expense that is fully consumed within the same year affects the same result however it's allocated — so splitting it by month is useful only if the company reads monthly results.

Three mistakes made with account 471

The fourth row is the most frequent. A stack of paper bought for a year is not a prepaid expense: it's inventory, and it is consumed as it is used, not as time passes. The difference seems subtle and becomes visible at stocktaking, where inventory is counted and prepaid expenses are not.

The closing check, in two minutes

For the balances in 471 and 472, a schedule is kept: for each amount, what period it covers and how much remains. The schedule's total must exactly match the accounting balance.

The check has a useful side effect: it forces you to name the period. An amount for which you cannot write "covers months X–Y" has no business being in 471, and the discovery happens here, not two years later, when no one remembers where it came from.

Part VI · What doesn't come from a document · phase 7

Chapter 29

You reverse it, you don't delete it

All the chapters so far have assumed that the entry is correct. This chapter is about what happens when it isn't — and it is, probably, the most important principle of discipline in the whole book.

The rule fits in one sentence: an accounting entry is never deleted and never modified. It is corrected through another entry, which remains linked to the incorrect one.

Why it isn't deleted

The reason isn't bureaucratic. Accounting in which entries can be modified after they've been made no longer proves anything — because any figure in it could have been written retroactively, and no one could tell which.

The evidentiary value of accounting lies precisely in the fact that the past is closed. A ledger that can be revised isn't a ledger at all: it's a permanent draft. That's why the law requires that errors be corrected through new entries, and the trace of the error must remain.

The trace of the mistake is part of the evidence. Accounting with no visible errors isn't clean accounting — it's accounting in which nothing can be verified anymore.

The practical consequence is that a serious accounting program refuses to delete an entry that has already been posted. Not as a precaution, but as a property: if deletion were possible, all the other guarantees would be decorative.

Draft and posted entry

The rule, however, doesn't apply from the very first moment. Between the moment an entry is composed and the moment it actually enters the accounting records, there is an intermediate state in which it can be freely corrected or deleted.

This boundary is useful for another reason too: it separates the place where work is done from the place where responsibility is taken. An operator can freely correct a draft; posting is the act by which someone takes ownership of the figure.

What a reversal (storno) looks like

The reversal entry uses the same accounts and the same amounts as the incorrect entry, with the sign changed. The effect is that the two cancel each other out, and the correct entry is made separately.

It is not dated in the period of the incorrect entry, but in an open period — usually the current one. The reason is the same as at the beginning of the chapter: a closed month has already been reported, and rewriting it would contradict declarations already filed.

In red or in black: two conventions, only one correct

There are two ways to accounting-wise cancel an entry, and the difference between them seems purely aesthetic. It isn't.

Reversing "in black" — swapping debit for credit — gives the right balance, but inflates the turnover: the revenue account reports 10,000 lei of activity on the credit side and another 10,000 on the debit side, that is, activity that never existed.

Turnover and profit come out correct, because they're read as a difference. But the turnover columns in the trial balance and the total of the journal register lie — and it's those that the accountant reads, not the calculator. In a company with many corrections, the trial balance ends up showing double the actual volume of activity.

That's why the correct convention is the "in red" reversal: the same accounts, in the same direction, with the amount negated. The name comes from the era of handwritten ledgers, when amounts to be canceled were written in red ink and subtracted, not added.

What a reversal doesn't solve

The reversal corrects the accounting records. It doesn't correct documents that have already left the company, and this is where the most costly confusion arises.

If the incorrect invoice was sent to the customer and reported in the VAT return, canceling it internally changes nothing for the customer or for ANAF. A correction invoice — a credit note — is needed, issued, transmitted, and reported in its own right.

The rule, in short: the reversal fixes what was written; the correction document fixes what was communicated. An error that has left the company needs both.

Corrections that aren't made through a generic reversal

There is a category of entries where the accounting reversal alone does more harm than good: those that also moved something besides accounts.

A sale of goods also decreased the quantity in the warehouse record. A reversal that touches only the accounts leaves the accounting correct and the inventory records wrong — and the two will diverge silently, until the inventory count puts them face to face, months later.

In such cases, the correction must be made at the level of the document that produced both effects, not at the level of the resulting accounting entry. It's the same idea as in Chapter 11: don't improvise with the general tool when there's a suitable one.

What's next

This chapter concludes Part Six, and with it, all the entries. The company's accounting now contains everything that happened, including what happened wrongly and how it was corrected.

Next comes the part where the same operations are viewed from a different angle: the registers. The same figures, once in chronological order and once grouped by account — and the tool that says whether everything balances.

Six situations, six treatments

"You reverse it, you don't delete it" is the rule. Its application, however, differs depending on how far the error has traveled — and the distance is measured by three thresholds: posting, month closing, and declaration filing.

The observation that orders the table: the error doesn't become harder to correct in accounting terms, it becomes harder to correct COMPLETELY. The reversal always resolves one's own records; the rest of the third column are things that have left the company and can't be withdrawn through a single entry.

What isn't reversed

There are operations where the accounting reversal alone does more harm than good — and they all have in common the fact that they moved something besides accounts.

The reversal corrects ACCOUNTS. When the operation also moved a quantity, a register, or a calculation, the correction must be made where the movement occurred — otherwise the two records diverge silently.

Part VII · The registers and verification · phase 8

Chapter 30

Chronological: the general journal

All the entries made so far exist somewhere. This part is about exactly where — and about why the same transactions are written twice, in two different registers, without this being unnecessary duplication.

The first of these is the general journal: everything that happened, in the order it happened. It's the register the law requires first, and it's the simplest of all.

What it looks like

One row for each line of each accounting entry, in chronological order, with the sequence number, date, document, explanation, debit account, credit account, and amount.

The most useful observation about this table is what it doesn't contain: no balance, no total per account, no grouping. The general journal doesn't answer the question "how much do I have in the bank." It answers the question "what happened on June 15."

What the order is for

Chronological order isn't a presentation convention — it's the whole point of the register. A continuously numbered journal, with no gaps, is proof that nothing was inserted afterward between two already-recorded transactions.

From this follow two rules that seem pedantic but aren't: numbering is continuous throughout the financial year, and a number once assigned is never reused. A gap in the numbering is a question someone will have to answer.

The general journal isn't a list of transactions. It's proof that the list is complete and that it hasn't been tampered with.

This is also why corrections are made through reversal (storno), as seen in chapter 29: a reversal entry receives the next free number and appears under its own date, not the date of the error. The journal thus preserves both the mistake and the correction, in the order they occurred.

The total that ties everything together

The journal has a single summary figure: the total of the amounts recorded in the period. In the example above, for June, it is 76,659.17 lei.

This figure seems pointless — it represents neither revenue, nor expenses, nor cash moved. Its value lies elsewhere: it must equal the total turnover in the trial balance, both on the debit and on the credit side. It's the first of the correlations that tie the registers together, and chapter 32 brings them all together.

Auxiliary journals

For a company with a few hundred transactions a month, a single journal is enough. For one with tens of thousands, it becomes unreadable — and that's when auxiliary journals are used: one for purchases, one for sales, one for bank, one for cash, one for miscellaneous transactions.

Each gathers transactions of the same kind, and their totals are carried into a centralizing journal, which remains the general journal proper. The structure changes nothing about the substance: the same transactions, the same order, just grouped so they can be read.

It's worth knowing that the sales and purchases journals also have a second, tax-related use: the VAT return is compiled from them. That's why they appear even at small companies, where volume wouldn't otherwise require them. Chapter 35 returns to them.

What comes next

The general journal tells everything that happened, but says nothing about state. To find out how much you have in account 401 at the end of the month, you need the same transactions viewed from another angle: grouped by account. That's the general ledger, and it's the next chapter.

The sequence number: how and when it's assigned

Numbering seems like an administrative detail until someone asks why a number is missing. Then it becomes the only question that matters, and the answer depends on a choice made long before: at what moment the number is consumed.

The same logic determines the order: the number is assigned in the chronological order of the posted entries, not in the order they were typed. A document from June 3 entered on June 20 receives the number it's entitled to by date, not the last free one.

VAT journals: the same transactions, a different criterion

The sales journal and the purchases journal are registers for VAT purposes, distinct from accounting auxiliary journals — and confusing the two produces, year after year, mismatches between the accounting records and the VAT return.

The correlation to check monthly is simple and catches almost everything: the VAT total in the sales journal must equal the credit turnover of account 4427, and the one in the purchases journal must equal the debit turnover of account 4426. In the book's example: 2,940 and 2,100 lei, in both places.

If the VAT journal and the trial balance don't agree, the return built on either of them will be wrong. The check is done before filing, not after.

Part VII · The registers and verification · phase 8

Chapter 31

Systematic view: the general ledger

The same transactions, the same amounts, the same month — but arranged differently. The general ledger takes everything from the journal and groups it by account, not by day.

The result is a ledger card for each account, and that card answers the question the journal could not: what is the situation of this account, where did it start, and how did it get here.

An account's ledger card

The structure is the same for all accounts, regardless of their nature: opening balance, the period's movements in chronological order, the turnover on each side, and the closing balance.

This card tells, on a single screen, the complete story of the merchandise in June: the business started with 20,000 lei of merchandise, purchased another 10,000, sold merchandise that had cost 8,000, and was left with 22,000 in the warehouse.

It's worth noting that the same event — the sale on June 15 — appears in the journal as three lines, of which only one is visible in the ledger card for account 371. This is not a loss of information: the other two appear in the ledger cards for accounts 4111, 707, and 4427. Each account sees, from the transaction, only the part that concerns it.

The journal and the general ledger contain exactly the same figures. They differ only in the question they answer: "what happened then" versus "what is the situation here".

The path back

The most important property of the general ledger is not that it shows the balance, but that it shows what it's made of. Any figure in a report can be unpacked, step by step, down to the document that produced it.

The path always has the same four steps: the balance sheet or the report you started from points to an account in the trial balance; the account in the trial balance points to its ledger card in the general ledger; a movement in the card points to the journal entry; the entry points to the supporting document.

Accounting software makes this journey in four clicks. A paper file makes it in four quarter-hours. But the possibility of making it is the same, and it is mandatory: a figure that cannot be traced back to a document has no business being in a financial statement.

Synthetic and analytic

Account 401 Suppliers has a balance. But the real question for any business owner is not "how much do I owe", but "whom do I owe". The answer lies in the account's analytic breakdown: the same amount, split by each partner.

The relationship between the two is a fixed rule: the sum of the analytic balances must equal the balance of the synthetic account. Not approximately — exactly. If they are not equal, one of them is wrong, and the wrong one is almost always the analytic breakdown, because that's where partners recorded under two different names get lost.

This check is worth doing monthly and is among the few that can be fully automated. Software that displays the analytic breakdown without confirming agreement with the synthetic account offers a nice-looking but unreliable list.

What the general ledger is not

Two confusions deserve clarification, since they often arise among those learning on their own.

The general ledger is not the trial balance. The general ledger contains the movements — all of them, one by one. The trial balance contains only the totals of each account, on a single line. One is the material, the other is the summary.

Nor is it a separate register, written a second time. In paper-based bookkeeping it actually was: the same transaction was copied manually into two registers, and the copying was the source of errors. In software, both registers are two views of the same collection of entries — which is why they can never differ.

What's next

With the journal and the general ledger, the bookkeeping is complete: everything is recorded, in chronological order and grouped by account. The question remains whether everything adds up — and for that there is a single tool, the oldest in accounting. That is the next chapter.

Which accounts require analytic development

Analytic detail is not kept everywhere — that would be work without information. It is kept where the synthetic balance does not answer the question the account exists for.

For inventories, the breakdown is done by custody/management unit and by product, and for expenses with limited deductibility — protocol, sponsorship, vehicles — by category, because otherwise the year-end tax calculation cannot be done without reconstruction.

What happens when the analytic breakdown doesn't agree

The equality between the sum of the analytic balances and the synthetic balance is the cheapest check in accounting and one of the most productive. When it doesn't work out, the causes are few and identifiable.

An analytic breakdown that doesn't agree with the synthetic account is not a reporting error: it's a data error. And the wrong figure is almost always the one in the breakdown, because the balance is checked against totals.

Part VII · The registers and verification · phase 8

Chapter 32

The trial balance

The trial balance is the oldest control tool in accounting and, five hundred years later, still the most used. It is prepared monthly, before any report or declaration, and answers a single question: does everything balance?

Its form is a list: one row for every account that was touched, eight columns of figures, a totals row. Its power lies in the last row.

The eight columns

Every account appears with the balance it started from, what moved during the month, the sum of the two, and the balance it ended at — each, on debit and on credit.

Row 4111 Customers deserves a closer look: it has movement on both sides and no balance. The customer was invoiced 16,940 lei and paid exactly that amount in the same month. The activity existed, the balance is zero — and a trial balance showing only balances would hide it completely.

The four equalities

The total at the bottom of the table contains four pairs of figures, and each pair must consist of identical numbers. These are the four series of equalities, and each one proves something different.

The correlation with the journal entry register is the most useful of all, because it links two registers prepared along different paths. The journal collects lines in chronological order; the trial balance collects them by account. If the totals don't match, a transaction entered one and not the other.

What happens when it doesn't balance

In accounting kept on paper, a trial balance that wouldn't close meant hours of searching. In a software program, the situation shouldn't arise at all: entries are validated at recording time, and an unbalanced entry cannot be saved.

When it does occur, the cause is almost always one of three: opening balances carried over incorrectly at the start of the year, a transaction recorded directly in the database, bypassing the application, or a propagated rounding. The difference itself is the first clue — if it's divisible by nine, it's almost certainly a digit transposition.

The trial balance is done monthly, not annually. An error found in the month it occurred costs a minute; the same error found the following March costs a day.

What can be read from a trial balance, beyond the equalities

An experienced accountant doesn't look at the trial balance as a balance test, but as an X-ray. Several things can be seen immediately, without any calculation.

What comes next

A trial balance closed on all four equalities is good news. But it isn't the news most people think it is — and the next chapter is the most counterintuitive in the whole book.

Three forms, the same trial balance

The trial balance can be prepared with three, four, or five series of equalities, and the choice doesn't change the information, only how much of it is visible.

What doesn't change in any form is the number of accounts: the trial balance includes ALL accounts touched during the period, including those with zero balance, if they had movement. An account that received and gave the same amount existed that month, and a trial balance that omits it hides real activity — the case of account 4111 in the example above.

The analytical trial balance, the check that doesn't follow from the synthetic one

The trial balance above is synthetic: one row per account. Alongside it, for accounts that require it, an analytical one is prepared — one row per partner, per stock location, per employee.

The link between them is a strict equality: the sum of the analytical balances of an account must be exactly its synthetic balance. And this check does NOT follow from the four equalities of the synthetic trial balance — it closes perfectly even if the analytical detail is wrong, because the balance check operates on totals, not on the breakdown.

The four equalities say nothing about the analytical detail. It's a separate check, and it's the one that uncovers partners entered twice under slightly different names.

What's done with it, after it's closed

A verified trial balance is not an end in itself, but the starting point for all the other work of the month. The order is fixed and is worth remembering as a list.

Part VII · The registers and verification · phase 8

Chapter 33

Balanced does not mean correct

This chapter contradicts the most widespread belief about bookkeeping: that a trial balance which closes means the bookkeeping is correct.

It doesn't. The trial balance proves only one thing — that every amount was written twice, once on debit and once on credit. About the correctness of those amounts, about the accounts chosen, about transactions missing entirely, it says absolutely nothing.

The trial balance verifies form, not substance. It's a necessary condition, and one of the weakest conditions in bookkeeping.

The six errors that get through

They're worth listing, because they aren't exotic cases: they are exactly the mistakes made in practice, and none of them disturbs any equality.

The second row deserves elaboration, since it's the most frequent. Goods recorded to a consumable-materials account don't unbalance anything, but they produce four consequences: the merchandise inventory is understated, the materials inventory is overstated, the cost of goods sold can no longer be correctly released at the time of sale, and at physical inventory the figures won't match anything in the warehouse.

And the third row is the most frightening, precisely because it's the simplest: an invoice for 10,000 lei entered as 1,000 passes any balance check. Nothing in the bookkeeping flags it. The only thing that uncovers it is someone comparing it against the document.

The case that gave the chapter its name

The symmetric error deserves to be told through a concrete case, because it's hard to believe until you see it.

A company has a liability of 100,000 euros to a foreign supplier. At closing, the exchange rate rises. The rule, from chapter 26, says that a liability that increases produces a loss. A program applying a single formula to all accounts instead records it as a gain.

The resulting entry is perfectly balanced: debit one account, credit another, the same amount. The trial balance closes. All four equalities are respected. And the year's result is wrong by double the difference — once because the loss is missing, and again because a nonexistent revenue appears.

This isn't a made-up example. It's a real error, found in an accounting program in actual use, which had produced years of wrong results without any automatic check ever flagging it.

What is verified, then

If the trial balance isn't enough, the natural question is what is. The answer isn't a single tool, but a set of confrontations with something outside the records — and the book has already introduced all of them.

The last row is the only one that catches all six errors from the previous table, and it's the only one that no program can do in someone's place. It's also why bookkeeping remains a profession, not a calculation function.

What comes next

One more report remains in this part, and it's the only one that verifies nothing: it's built to be used in running the business. It shows who owes you, whom you owe, and since when — and aging is the information missing from all the other reports.

What it does catch, though

The chapter has insisted on what the trial balance doesn't catch, and rightly so. But it's worth also saying what it does catch, because otherwise it seems useless — and it isn't.

The divide-by-2 and divide-by-9 rules aren't bookkeeping superstitions: they follow from arithmetic. An amount moved from one side to the other changes the difference by double its value, and transposing two adjacent digits always produces a multiple of nine. They don't say where the error is, but they narrow the search from the whole trial balance down to a single type of amount.

What is done instead of the missing check

If balance can't confirm correctness, the practical question is what does. The profession's answer is sampling: not everything gets checked, what's checked is chosen.

There are three criteria for the selection, and they explain why a well-made file doesn't contain random checks. Large transactions are chosen, because there an error carries weight; unusual ones, because there the classification is debatable; and those at the edge of the period, because that's where cut-off errors occur.

Balance verifies itself. Correctness is verified through selection — and good selection is the kind that goes where an error would cost the most, not where it's easiest to look.

Part VII · The registers and verification · phase 8

Chapter 34

The Aging Schedule

All the reports so far have been made to verify or to report. The aging schedule is the only one made to run the business — and, perhaps for that reason, the only one a business owner reads on their own initiative.

It answers three questions the trial balance cannot reach: who owes me, whom I owe, and, above all, since when.

Why the Balance Isn't Enough

Account 4111 Clients (Trade Receivables) might show a balance of 200,000 lei. The figure is true and almost useless, because it doesn't say whether the 200,000 are invoices issued last week or receivables two years old.

The first situation is normal. The second is a company in danger, with the same balance and the same balance sheet. The difference between them shows up nowhere in the general ledger — only in the age of the balances.

A receivable isn't measured only in lei, but in lei and in days. The second dimension is missing from all mandatory reports.

What It Looks Like

The aging schedule places each partner on a row and splits their balance into age brackets, counted from the invoice date or from the due date.

Read this way, the table says something the balance didn't: 31,000 lei are older than ninety days and belong to a single client. That is no longer a receivable to be tracked, but a decision to be made — chapter 27 shows what happens to it in the accounting.

How Aging Is Determined

This is the technical part of the chapter, and it deserves to be understood, because on it depends whether the table tells the truth.

A client usually has several invoices and several payments, and payments rarely arrive with the note "this settles invoice number such-and-such." It must therefore be decided which invoice each collection settles, and the entire distribution across intervals depends on this decision.

The natural convention is settlement in order of age: the first payment settles the oldest invoice, the remainder carries forward. This is also the one implicit in law absent other instructions, and the prudent one — if you err, you err by showing the balance older than it is, not newer.

The opposite convention — payment settles the most recent invoice — would make old debts appear forever fresh, and the table would become the exact opposite of an alert tool.

The Trap of Transfers Between Accounts

There is a typical error in building the aging schedule, worth describing because it is invisible and doubles the figures.

A receivable can change accounts without changing size. A delivery note sits in 418 Clients — Invoices to be Issued; when the invoice is issued, the same receivable moves to 4111. Symmetrically, an expense without an invoice sits in 408 and moves to 401 when the invoice arrives.

If the mechanism that builds the aging schedule looks at each line separately, it will see at 4111 a new receivable of 10,000 lei and will not see that elsewhere a receivable of 10,000 has just disappeared. The result: the same client appears with 20,000 lei, half of which doesn't exist.

The sign that this has happened is easy to recognize: the aging schedule shows a different total than the account balance. When two reports about the same thing don't match, the one built by aggregation is always the wrong one — not the balance.

What to Do With It

An aging schedule read monthly changes three things in running a business, and none of them is accounting.

First, the order of phone calls. You don't call every overdue client, but call first those in the oldest column with the largest amount — a list of five names, not fifty.

Then, the decision whether to keep delivering. A client with 45,000 lei overdue for more than sixty days should not receive a new order without a discussion, however tempting the order might be. Accounting doesn't make the decision, but it provides the figure without which the decision would be made by instinct.

Finally, the mirror. The same report exists for suppliers, and comparing the two tables shows whether the company finances its activity from supplier credit or the other way around. It is the simplest cash-flow analysis one can perform, and it requires no calculation.

What Comes Next

With this chapter, part seven comes to a close. The company's accounting is complete, verified from every angle that can be verified, and read as a management tool.

Next comes the part where it turns toward the state. Not as a separate obligation, but as a result: declarations aren't filled in, they are composed from the month already recorded. And the first of them is the most frequent and the most misunderstood — value added tax.

The Supplier Aging Schedule, Read in Reverse

The same report exists for payables, and there it is read exactly the other way around: what is a warning signal for clients is, up to a point, a resource for suppliers.

An old debt to a supplier isn't necessarily a bad sign — it could be an invoice under dispute, or one for which a long term was agreed. It becomes a problem when no one knows anymore which of the two it is, and that happens within a few months.

The Two Aging Schedules, Read Together

The real information emerges from the comparison. If you collect on average in 60 days and pay in 30, you finance the activity from your own money; if it's the reverse, you finance it from supplier credit.

From the average age of receivables, a simple indicator is calculated: the average collection period, i.e., the client balance divided by daily sales. At 156,000 lei receivables and sales of 30,000 lei per month, it is approximately 156 days — a number that says more than any discussion about "clients who pay late."

Receivables and payables are read together. Separately, each shows a list of names; together, they show whether the company is financing its growth or paying for it.

Part VIII · The state · phases 9 and 13

Chapter 35

VAT: collected, deductible, payable

Value added tax is the most misunderstood obligation of a company, and the confusion starts from something simple: the VAT money passes through the company's account, but it is never the company's own.

The company collects it from the customer along with the invoice, holds it for a few weeks, and transfers it to the state. Its role is that of a collector, not a payer — even though the return says "payable."

The two accounts

The entire mechanics of VAT rest on two accounts and the difference between them.

The "value added" in the tax's name is exactly this difference. The company pays tax only on what it has added to the value of the good — the rest had already been paid by its suppliers, and the company deducts it.

The return, in real figures

Here is June for a small company, as the VAT return comes out of its own accounting records, without any intervention.

The company collected 2,940 lei of VAT from customers and paid 2,100 to suppliers. It owes the state the difference of 840. It has neither lost nor gained anything: the 840 lei were, from the very start, the state's.

The return is not filled in, it is composed. Every figure in it comes from accounting entries already recorded — if you have to type something in, an entry is missing somewhere.

When the deductible is larger

This happens often and is not an anomaly: a month with large investments, inventory built up ahead of the season, an exempt export with the right of deduction. The difference becomes VAT to be recovered.

The company then has two options: to request a refund, with a tax inspection in most cases, or to carry forward the amount to the following month, where it will be deducted from the VAT payable.

Carrying forward is the natural and most commonly used mechanism. It is also why the return has, at its head, a line for the previous period's balance: VAT is not closed month by month, but flows from one month to the next until it is settled.

When VAT is not deductible

The right of deduction is not automatic. Three conditions must be met at once, and the absence of any one of them cancels it.

To these are added the express limitations: expenses without economic purpose, entertainment expenses above the ceiling, and, the most frequent case, passenger cars — for which the deduction is limited to half, if the vehicle is not used exclusively for economic purposes.

Monthly or quarterly

The fiscal period depends on turnover and on intra-Community operations. The practical rule: small companies file quarterly, large ones monthly, and an intra-Community acquisition of goods immediately changes the period to monthly, starting from the very month in which it occurred.

The surprising consequence is that the change is neither optional nor postponable: a single purchase from a supplier in another Union country brings with it the obligation to switch to monthly filing and to report the change. It is one of the few situations where a single operation changes the company's reporting regime.

The link with the registers

The return is composed from the sales journal and the purchases journal, encountered in chapter 30. These are registers for VAT purposes, distinct from the accounting journals, and their totals must correspond to the turnover of accounts 4427 and 4426 in the trial balance.

This correlation is the last check before filing and is always worth doing: if the journals and the trial balance do not say the same thing, one of them is wrong, and the return built on either one will be just as wrong.

The chargeable event and chargeability

This is the most frequent source of discrepancies between the accounting records and the return, and it comes from a distinction that the document does not show: the moment the operation OCCURS is not necessarily the moment the tax BECOMES DUE.

The first is called the chargeable event, the second chargeability. In an ordinary supply they coincide, and the distinction seems unnecessary. In four situations they do not coincide — and that is exactly where the month is gotten wrong.

The third row deserves to be read twice, since it is the most counterintuitive: for an intra-Community acquisition without an invoice received by the 15th day of the following month, the tax becomes chargeable regardless, on that date. The document is not waited for — a self-invoice is issued.

The return is composed according to CHARGEABILITY, not the document date. It is the only rule in this chapter that a program cannot guess from the invoice: it needs to know what kind of operation it is.

The advance, in figures

The advance case deserves to be unpacked, since it produces three operations for a single sale and because the second one is the one that gets forgotten.

The regularization of the advance is done in the final invoice itself, not as a separate operation — and this is why the document type for the final invoice with an advance is distinct from the ordinary one. It is the same pattern as the inventory write-off in chapter 11: a line that is not visible from the document, but without which the result is wrong.

Part VIII · The state · phases 9 and 13

Chapter 36

Special regimes

The previous chapter described the general rule: the tax is owed at delivery, by the supplier, on the full value. There are, however, situations in which each of these three elements changes — the moment, the person liable, or the taxable base.

These are not exotic exceptions. An ordinary company usually encounters two or three of them, and the wrong application of any one produces either tax paid for nothing, or a liability discovered during an audit.

The map

VAT on collection: the tax follows the money

Under the normal regime, VAT becomes chargeable when the invoice is issued. The company owes it to the state even if it has not collected it — and an uncollected invoice becomes, as seen in chapter 27, a loss greater than the uncollected amount.

The VAT on collection regime shifts the moment: the tax becomes chargeable when the customer pays. For a company with customers who are late payers, the cash-flow advantage is considerable.

The price is symmetrical and is often forgotten: the company can no longer deduct the VAT from suppliers either, until it pays them. The regime therefore applies on both sides, and a company that pays promptly and collects slowly gains, while one that does the opposite loses.

Accountant, this is where account 4428 Non-chargeable VAT finds its purpose: the tax sits there, on sales as on purchases, until the moment of payment, when it moves into 4427, respectively into 4426.

The scheme is optional and has its own threshold, distinct from the small-enterprise exemption threshold: 5.000.000 lei for 1 March–31 December 2026, increasing to 5.500.000 lei from 1 January 2027. Entry, exit and threshold crossings are notified, and actual application is verified in ANAF’s public register; an internal setting that disagrees with the register does not change chargeability for counterparties.

Reverse charge: the obligation shifts to the buyer

This is the mechanism that surprises most on first encounter: the supplier issues the invoice without VAT, and the buyer records the tax itself, at once as a liability and as a deduction right.

The result is that, in most cases, the effect on cash flow is zero: the same amount appears in 4427 and in 4426, and the VAT return picks it up in both columns. This is also why the operation seems, at first glance, pointless.

The point is different: the reverse charge eliminates the possibility that the supplier collects the VAT and disappears without remitting it. It applies to intra-Community acquisitions, to imports of services, and, within the country, to categories exposed to fraud — cereals, timber, waste, certain construction works.

The practical trap is that the omission is not visible: if the company does not record the self-charge at all, the VAT return is smaller on both columns and appears correct. It is only discovered by cross-checking against the partner's declarations, at the tax administration level.

Under reverse charge, disregarding the obligation does not produce a payment difference. It produces a false declaration — which is more serious.

The margin scheme: tax only on the markup

A dealer who buys a used car from a private individual receives no VAT invoice, so has nothing to deduct. If they were to charge tax on the entire selling price, they would pay tax on a value they did not add themselves.

This is what the special margin scheme is for: the tax is calculated only on the difference between the selling price and the purchase price. It applies to second-hand goods, works of art, antiques and, with its own rules, in tourism.

Two consequences are worth remembering. First: the invoice issued does not show VAT — the buyer has nothing to deduct, and the mention of the special regime is mandatory. Second: the company must keep a separate register of goods under this regime, because the margin is calculated for each item individually, not on the total.

Pro-rata: partial deduction

A company that has both taxable operations and exempt operations without deduction rights — a medical clinic with a pharmacy, a school with consulting activity — cannot fully deduct the VAT on shared expenses.

Pro-rata is the proportion in which it deducts them: the share of operations with deduction rights in the total operations. During the year a provisional pro-rata is used, and at year-end the final one is calculated and the difference is adjusted.

A detail that changes the figure: the final pro-rata is rounded UP, to the nearest unit. A proportion of 70.2% becomes 71%, not 70%. Rounding down, done out of arithmetic reflex, takes away part of the company's deduction every year.

Triangular operations

The last regime on the list is the rarest and most confusing: three companies from three states of the Union, goods moving directly from the first to the third, while the one in the middle — which buys and resells — is not allowed to register in either of the other two countries.

The simplification measure resolves exactly this: the acquisition made by the company in the middle becomes non-taxable in its country, and the payment obligation shifts to the final buyer. The price is a strict declarative formality — the operation must be reported with a distinct code in the recapitulative statement, and the invoice must bear the corresponding mention.

It is worth knowing that it applies and that it is simple once understood, because the alternative — VAT registration in a second state — costs ten times more than the operation itself.

Pro-rata, over a whole year

Pro-rata is the regime most often applied incorrectly, because it has three distinct moments and because its rounding is not done like any other. Here is a full year, for a company with taxable and exempt operations.

The rounding deserves emphasis, because it is the only one in accounting that always goes in the company's favor and because it is easy to get wrong out of reflex. At 85.25%, arithmetic rounding would give 85%, i.e. 7,140 lei deducted — 84 lei less than is due. The amount is small over one year; the rule, applied wrongly and systematically, is not.

The final pro-rata is rounded UP, to the nearest unit. Not to the nearest whole number — up, always. It is written into the law, and it is the only rounding that is not negotiable.

What is included in the calculation and what is not

The denominator of the pro-rata is not "all revenues." There are operations that do not enter the calculation at all, and including them distorts the proportion — usually to the company's disadvantage.

And one last rule, often forgotten from year to year: pro-rata applies only to acquisitions with MIXED use. What is purchased exclusively for operations with deduction rights is fully deductible, and what is purchased exclusively for operations without such rights is not deductible at all. Applying pro-rata to all inputs is an error in both directions at once.

Part VIII · The state · phases 9 and 13

Chapter 37

Declarations and their deadlines

The number of declarations a Romanian company owes is intimidating at first glance and becomes manageable as soon as you understand the principle: each declaration corresponds to an obligation, and obligations arise from what the company does, not from what someone chose to require.

A company with no employees does not file D112. One with no intra-Community transactions does not file D390. The list is not the same for everyone — it is derived from the tax regime and from actual activity.

Who owes what

The organising principle is that returns are not independent. D394 covers the same reporting period as D300 and must reconcile to it, but has a distinct deadline: the 30th day, inclusive, of the month following the reporting period, or 28/29 February for January. D406 contains detailed data that must agree with the accounting and tax records. Annual financial statements reconcile to the final trial balance and the returns filed for the relevant periods.

The declarations are different views of the same data. If they don't agree with each other, the declarations are not the problem — the accounting behind them is.

The company's regime decides the list

Two companies with the same turnover can have completely different lists. Three choices, made once, determine almost everything.

The month's calendar

For an ordinary company, a VAT payer with employees, the month has a fixed rhythm. It's not complicated, but it is strict — and easier to keep as a list than as a memory.

First, the previous month's documents are closed and the bank statement is reconciled. Then adjustments are made. Then the trial balance is prepared. Only after the trial balance are the declarations generated — because they are composed from it. Finally, they are filed, the obligations are paid, and the evidence is archived.

The order is not a preference. A declaration generated before the trial balance will need correction afterward, and correcting an already-filed declaration costs more than a day's wait. Part nine of the book covers this order in detail.

The annual close

Besides the tax returns, the fiscal year ends with obligations that are not owed to ANAF and that are most often forgotten: the inventory register and inventory documents, financial statements filed on time, and the general meeting's resolution approving the statements and distributing profit.

The last point deserves emphasis for a business owner: dividends cannot be distributed without a shareholders' resolution, and the resolution cannot be adopted without approved financial statements. This order is legal, not accounting, and cannot be shortened.

What comes next

A correctly prepared declaration is half of the obligation. The other half — and, in an audit, the part that matters — is proof that it was filed and received. That is the next chapter, and it is the shortest in the book.

The fiscal vector: what the state thinks you owe

There is a list of your declarative obligations kept not by you, but by the tax administration. It's called the fiscal vector, and it's the configuration according to which the system expects declarations from your company.

The consequence matters more than it seems: if the vector says you are a monthly VAT payer, the system expects a return every month — even if the company no longer has any activity. Not receiving it, it sends a notice. Conversely, if the vector doesn't contain an obligation you actually have, the absence of the declaration goes unflagged, but the obligation remains.

The fiscal vector does not update itself. It is a declaration of mentions filed by the company — and between what you do and what the state thinks you do, there can be a gap that generates notices for years.

How the list is derived, in practice

A well-built program does not keep a hand-written list of the company's declarations: it derives it from the company's profile. Three or four attributes decide almost everything, and the rest follows.

What is paid, beyond what is declared

Filing and payment are two distinct obligations, with distinct sanctions. They can be fulfilled one without the other, in both directions — and the most unpleasant case is a declaration filed with the amount unpaid, because that's where late-payment charges accrue on a debt acknowledged in writing.

Declarations arising from a situation, not from a regime

The table at the beginning of the chapter is the usual list: what an ordinarily active company owes, month by month. There is, however, a second group, tied not to the tax regime but to circumstance. They are never filed — until the month in which they are.

The distinction matters for whoever keeps the records. The first list can be written into a calendar at the start of the year and followed mechanically. The second cannot: it is derived from the month's transactions, and if no one looks at the transactions, the obligation goes unnoticed. It is the only place in the tax calendar where the absence of a declaration doesn't prove it wasn't owed.

The first four arise from a transaction that the accounting records already contain. A program that keeps the records can propose them on its own; a person keeping the records must know they exist, otherwise they won't look for them.

When the declaration is wrong

No filed declaration can be deleted. It can only be replaced or corrected, and how this is done depends on what kind of declaration it was — a distinction usually learned by getting it wrong once.

The procedure depends on the form. D394 and D112 D300 is the important exception: no amending statement is submitted for the same period. Operations omitted and other corrections provided for in the instructions fall into the ranks of regularisations of a subsequent statement and material errors strictly defined follow the application and decision of the tax body.

Declarations that establish payment obligations — D100 — are not rectified the same way. For them there is the D710 correction declaration, which states what amount was declared incorrectly and what the correct one is. The reason for the difference is that D100 doesn't describe transactions but creates a debt in the payer's record: the correction must reach that debt, not retell the month.

Rectifier, D710, regularization D300 and application for material error correction are different procedures. The form and the cause of the error decide the way; the generic term "amending" is not sufficient.

The practical rule for both: the correction is made after the bookkeeping has been clarified, not instead of it. An amending return filed on top of a trial balance that is still unresolved will itself need correcting, and a string of amendments for the same month is the first thing an inspector will notice.

Obligations that accompany the goods, not the bookkeeping

All the returns discussed so far describe what was recorded. But there are two obligations that don't concern the recording itself, but the goods themselves: one is fulfilled before a truck departs, the other statistically reports the movement of goods across the border. Neither is a tax return in the usual sense — the first doesn't establish any amount, the second isn't even filed with ANAF — and that's exactly why they get overlooked.

e-Transport is the system through which the state tracks the transport of goods. Before the goods are set in motion, the transport must be declared and receives a unique code, which must accompany the goods along the way. The key word is "before": a code obtained after the truck has left fixes nothing, and roadside checks happen on the road, not in the bookkeeping.

In practice, the declaration is made from the delivery note — the document that is drawn up anyway when goods leave — plus data that isn't found in the bookkeeping: the vehicle, the route, who organizes the transport. Software that already has the delivery note can prepare the rest; what it can't do for you is know the license plate number.

Intrastat is something else, though it concerns the same goods. It's a statistical return, filed with the National Institute of Statistics, not with ANAF, and it establishes no tax. It becomes mandatory when the company exceeds an annual threshold, calculated separately for each direction: one for what comes in from the Union, another for what goes out. You can be obligated on one direction and free on the other.

Intrastat counts goods, D390 declares transactions. They're calculated from the same invoices, but they can't substitute for one another: one goes to statistics, the other to the tax authority.

It's also worth highlighting the difference from intra-Community services: they're included in D390, but not in Intrastat, because nothing physically crosses the border. This is the most common confusion between the two — and it's immediately obvious, because the totals never match.

Part VIII · The state · phases 9 and 13

Chapter 38

Filing and Proof

This chapter is short because it has a single idea, and that idea matters more than its length: the document that counts is not the declaration, but the confirmation that it was received.

A perfectly prepared declaration, saved on the accountant's computer, has no legal value whatsoever. A declaration with a mistake, submitted and confirmed, exists. The first is penalized as not having been filed; the second is corrected.

The Receipt

When filing electronically, the system returns a confirmation file — the receipt. It contains the submission identifier, the exact timestamp, and the status: accepted or rejected.

The receipt is the only proof that the obligation was fulfilled. Not the screen display, not the email from the software, not the fact that the button was pressed. During an inspection, the question is "show the receipt," and the answer "I did file it, I'm sure" carries no weight.

Without a receipt, a declaration is an intention. With a receipt, it is a fact. The distinction is the same as between a draft and a published article, from chapter 29.

From this, a simple practical rule follows: the receipt is downloaded and archived in the same folder as the declaration, immediately, not "when it's needed." Two years later, retrieving it from the tax system is possible, but not always convenient — and an inspection does not wait.

The Three States

An electronic submission does not have two possible outcomes, but three, and confusing the first two is the source of the most unpleasant surprises.

The third row deserves emphasis. A submission that was not completed — interrupted connection, expired session, certificate not uploaded — produces no error message to the taxpayer in many situations. The company only finds out when a payment notice arrives, months later, with late-payment increases having accrued in the meantime.

The rule, for any system that files automatically: "could not verify" does not mean "it's fine." A program that reports a submission as successful simply because it did not receive an error makes exactly the mistake an experienced accountant avoids.

Who Signs

Electronic filing is done using a qualified digital certificate, previously registered with the tax administration for that particular company. The certificate belongs to an individual — the administrator, the accountant, or the authorized representative — not to the company.

The practical consequence is usually discovered at the worst possible moment: if the person holding the certificate leaves and the power of attorney has not been updated, the company can no longer file anything until the formalities are redone. It is worth checking before the deadline, not on the day of it.

The Virtual Private Space (SPV)

Communication with the tax administration is no longer optionally electronic: companies are required to enroll in the Virtual Private Space (Spațiul Privat Virtual, SPV), and administrative acts are communicated through it.

The consequence is more important than it seems. An act communicated through this channel is considered communicated, and appeal deadlines run from that moment — regardless of whether anyone opened the message. An unread electronic mailbox produces the same effects as a refused registered letter.

That is why periodic checking is not an administrative formality, but a protective measure. It naturally fits into the monthly rhythm from chapter 37, alongside the filing itself.

What Comes Next

With this chapter, Part Eight closes. The company has calculated its obligations, has declared them, and has proof that it declared them.

Next comes the part about the moment when the month or year becomes final: closing. There, the order of steps matters more than anywhere else, and skipping one only becomes apparent months later.

What Is Kept, and For How Long

The receipt is not useful on the day of filing — at that point everyone already knows it was filed. It is useful years later, and that determines its retention regime: it is archived together with the declaration, for the entire period during which the authority may still establish obligations for that period.

The order in the table is also the order in which an inspector will ask for them: first what you declared, then proof that you declared it, then where the figure comes from. A file containing only the first two answers the compliance question and leaves the substantive one open.

Corrective Filing: What Can Be Done, and Until When

A wrongly filed declaration is corrected through a corrective declaration. The mechanism seems simple and has three subtleties that, together, produce most of the problems.

The second row changes how errors discovered late should be viewed: the legal correction deadline is long, but the practical one closes on the day the inspection notice arrives. That's why a known error should not be postponed — it is corrected while the correction is still within the company's control, not the inspection's.

The receipt proves that you filed. It does not prove that you filed CORRECTLY — nor that you paid. These are three distinct obligations, with three distinct proofs.

Two Paths, Not One

It's easy to believe the Virtual Private Space is a single-window service. In practice there are two different paths, with different responses, and confusing them means looking for proof where it isn't.

Tax declarations are validated and filed as a file, and the proof is the receipt described above. Electronic invoices go a different route: they are uploaded one by one, receive a registration number, and the response comes later — either the sealed invoice or the errors. There is no filing receipt for an invoice, and no upload number for a tax return.

The second thing that gets forgotten: messages do not stay there forever. The space is a communication channel, not an archive — documents must be downloaded and kept with you, along with the rest of the month's file. A company that relies on the fact that "they're in the SPV anyway" discovers their absence exactly when it needs them.

The Virtual Private Space is the mailbox, not the archive. What arrives there must be downloaded and put in the file; otherwise the proof lives on a server that isn't yours.

Part IX · Closing · phases 10–11

Chapter 39

Closing the month, step by step

Closing the month is not an operation, but a sequence. Everything that makes it up has already been described in the previous chapters — the documents, the bank statement, the adjustments, the VAT, the declarations. What is added here is their order, and it matters more than it seems.

The reason is simple: each step relies on the one before it. A step taken too early will have to be redone, and redoing it always costs more than waiting.

The six steps

It's worth noting that the first three steps produce records, the fourth produces documents for the state, and the last two produce nothing: they are acts of accountability. Closing is not an accounting operation, but a decision about accounting already done.

Why the order is not negotiable

The most frequent temptation is generating declarations before the documents are complete — usually because the deadline is approaching. The consequence is predictable: an invoice arriving in the meantime changes the VAT return, and the declaration already filed must be corrected.

The second temptation is reconciling the bank statement after the VAT adjustment. For a company using VAT on collection, this order is simply wrong: chargeability depends on payments, so an adjustment made before reconciliation works with incomplete data.

Every skipped step is paid for later, with interest. A step done on time costs minutes; the same step redone after filing costs an amended return.

The four states of a step

In a well-built flow, a step doesn't just have "done" and "not done." It has four states, and the distinction between them is what turns a list into a tool.

The distinction between "blocked" and "open" is what changes the way of working. A blocked step is not an outstanding task, but one that hasn't started yet — and a system that displays them the same way produces a list of ten things to do where, in fact, there is only one.

Who is responsible for each step

In a small business, the same person does all the steps. In one with a team, they get divided — and then a question arises that isn't accounting-related: who exactly is responsible for each, and by when.

It's worth recording, since it's the only information in the whole flow that cannot be deduced from the data. What has been recorded can be seen in the accounting; who was supposed to handle it and by what date is nowhere visible if no one writes it down.

It's a general observation about any workflow: only what cannot be deduced needs to be persisted. The rest is read from the data, each time, fresh — and the next chapter shows why.

What comes next

A question remains that the flow above raises without answering: how does anyone know a step is done? The usual answer — because I checked it off — is the wrong one, and the next chapter explains why.

What it looks like on a screen

The flow is not a checklist, but a state read from the data. Here is June for the example company, as the application sees it on July 14 — with the blockages spelled out, not just a traffic light.

It's worth noting what each blockage says: not "not ready," but exactly what's missing and, where relevant, by when. A flow that showed six red lights would be a dashboard; one that says "an invoice not submitted to SPV" is a work list.

The second row is just as instructive: the bank statement is "done," even though the month isn't closed. A step's state doesn't depend on the month's state — it's derived from its own data, independently. That's why a step can revert from "done" to "open" without anyone touching it, if a new document arrives.

Who is responsible and by when

In a small business, the same person does all six steps, and the question doesn't arise. In one with a team, it becomes the main source of delay: everyone thinks the next step belongs to someone else.

The person responsible and the internal deadline are the only information in the whole flow that cannot be deduced from the data — there is nowhere in the accounting that shows who was supposed to handle it. That's why they are recorded, while the rest is not: it's the same rule as in the next chapter, memorize the decisions and recalculate the facts.

Internal deadlines are usually anchored to the legal one: if the declaration is filed on the 25th, the bank reconciliation has an internal deadline on the 8th, and the trial balance on the 15th. Anchoring serves a practical purpose — when the legal deadline shifts, everything else shifts too, without anyone redoing the calculation.

Part IX · Closing · phases 10–11

Chapter 40

Status is derived from data, not checked off

This chapter is about a design decision that looks technical but is, in fact, a matter of accounting truth. The question is how a system — or a person — knows that a closing step is finished.

There are two possible answers, and the difference between them is bigger than it seems.

The checkbox and derivation

A checkbox is a statement about a moment in the past: at that time, that thing was done. The problem is that the statement has no expiration date, while the reality underneath it keeps changing.

A manual checkbox remains true even after the data beneath it has become false. A closing workflow that lies is worse than having none at all.

What the lie looks like, concretely

Three common situations, all encountered at any company.

The bank statement was reconciled on the 5th of the month and checked as finished. On the 7th, an old invoice arrives, gets recorded, and one of the payments already reconciled turns out to be for it. The reconciliation is no longer complete — but the checkbox is still there, green.

The tax return was filed and checked off. An error is then discovered, an amended return is filed, but someone forgets to update the status. The system shows "filed" — which is true, but is no longer the relevant information.

The documents were declared complete. A week later, three more invoices from the same month are recorded. The checkbox doesn't turn back on, because nothing is tracking it.

How derivation works instead

A derived status is not remembered: it is calculated, every time it is viewed, from the data existing at that moment.

The practical advantage is that the status corrects itself. The three situations above — the late invoice, the amended return, the documents added later — automatically make the step open again, without anyone having to remember.

What is still persisted

The principle doesn't mean nothing gets written down. It means that only what cannot be deduced gets written down — and those things are few and of a different nature.

Who is responsible for a step. Until when. What was observed along the way. Who approved the month and when. Whether closing was forced past an unfinished step, and for what reason. All of these are human decisions: they exist nowhere in the data, so they must be recorded.

The distinction is worth keeping in mind beyond accounting. In a well-built system, decisions are remembered and facts are recalculated. Whoever remembers the facts will sooner or later end up with two versions of the same truth.

What comes next

Derivation solves the problem of the status that lies. But it raises another one: if everything is always recalculated, what exactly makes a closed month stay closed? The answer is the next chapter.

How to write a status that cannot lie

The difference between a checkbox and a derived status is clearest in how each one is written. A checkbox is a stored value: "step 2 = finished". A derived status is a question posed to the data, every time.

The price is a query on every read, and it does get paid: for a company with tens of thousands of records, the question "are there still unrecorded documents this month?" isn't free. The usual compromise is caching the result for a few seconds, not forever — a cache that expires on its own can't lag behind reality by more than its interval.

Where the rule applies, beyond closing

Once stated, the rule is recognizable in many places, and accounting is full of them.

The test that separates them: if you delete the value and can rebuild it exactly from the data, it's a fact and must be derived. If you can't rebuild it, it's a decision and must be kept — along with who made it and when.

Applied to accounting, the test also explains why the month's file contains written notes, not just numbers: numbers can be recalculated anytime, judgments cannot. An allowance for receivables can be rebuilt from data; the reason the receivable was considered doubtful cannot.

Part IX · Closing · phases 10–11

Chapter 41

Locking the period

A closed month is a reported month. Its trial balance underpinned filed returns, and those returns produced payments, receipts and, potentially, an inspection. From that point on, the month no longer belongs to the company: it's a public statement.

Period locking is the mechanism that makes the statement stable. After it, nothing can be recorded, modified, or deleted in that month.

Why discipline isn't enough

The natural objection of any experienced accountant is that nobody intentionally touches a closed month. That's true, and irrelevant: the problem isn't intent, it's accident.

An invoice with a wrongly typed date — 2026-05-30 instead of 2026-06-30 — slips into last month without anyone noticing. A correction made "to make it balance" in an old month seems harmless. An import of documents with old dates rewrites an entire period.

None of these is bad faith. All of them produce the same result: May's trial balance is no longer the one on which May's VAT return was filed, and the difference is discovered at the earliest during an inspection.

A locked period doesn't protect bookkeeping from bad people. It protects it from ordinary mistakes, which would otherwise be discovered as late as possible.

What happens to a necessary correction

Locking doesn't forbid corrections — that would be absurd. It moves them in time: an error in a closed month is corrected through a reversal entry dated in an open month — exactly the procedure from chapter 29.

The result is that the closed month remains as it was reported, and the correction appears where it can be seen and explained. Anyone reading both months sees both the error and the fix — which is more than they would have seen if the figure had simply been changed.

It's the same logic as reversal entries, applied at the scale of the period: nothing is erased, it's written alongside.

Forcing

There are situations where closing must be done over an unfinished step: a return that couldn't be filed due to technical reasons, a statement that hasn't arrived, a deadline that can no longer be postponed.

A system that completely forbids forcing will be bypassed — through direct database writes, parallel files, "I'll handle it another way." A system that allows it without a trace is as good as having no control at all. The solution is a third path: it's allowed, but at a cost.

The written reason is the requirement that matters most, and not for control purposes. It's for whoever reads it a year later — often the same person who wrote it — who will need to understand why that month looks the way it does.

When to lock

Good practice is to lock immediately after the month's returns are filed, not at year-end. The reason is that the window between filing and locking is exactly the interval in which accidental changes can slip in.

A company that locks monthly has, at any moment, only one open month. One that locks annually has twelve — and a date error lands somewhere among them without hitting anything.

What comes next

With locking, the month becomes final. The same logic applies, at a different scale and with its own operations, to the entire fiscal year — and there something happens that has never happened in any month: revenue and expense accounts disappear.

What "locked" means, technically

Locking isn't a label placed on a month, it's a guard that stands between any write operation and the database. The difference shows in what exactly it stops.

Verifying that the guard covers all paths isn't done by reading the code, but by trying: a locked month must refuse each of the five, with messages explaining why. It's the kind of property that's proven by testing, not by intent.

Correction of errors from prior periods

Locking moves the correction in time, but doesn't decide where it's recorded. That depends on a separate accounting criterion: how large the error is relative to the financial statements.

The tax consequence deserves separate mention: a correction that goes through retained earnings doesn't change the current year's accounting result, but may change the taxable base of the corrected year — thus triggering an amended return for that year. The two records are corrected via different paths, and forgetting the tax one leaves a discrepancy that gets discovered during an audit.

Period locking establishes WHEN the correction is recorded. The materiality of the error establishes WHERE. These are two separate questions, and both require an answer.

Part IX · Closing · phases 10–11

Chapter 42

Closing the Year

Twelve closed months do not make a closed year. At the end of the fiscal year an operation takes place that has no counterpart in any month: the revenue and expense accounts are extinguished, and their balance is gathered into a single account.

The operation is called closing the result accounts, and understanding it clarifies a question many business owners ask without ever getting a clear answer: where exactly does it say how much the company earned.

Why the result accounts disappear

The accounts in classes 6 and 7 measure an interval, not a state. Account 607 doesn't say "I have 8,000 lei worth of goods," but "during this period I gave out 8,000 lei worth of goods." These are flow accounts, and the flow belongs to the fiscal year that is ending.

On January 1, measurement must start from zero — otherwise expenses would accumulate indefinitely, and the third year's result would also contain the first two. Emptying them is not a formality: it is the condition for the following fiscal year to mean anything.

Balance sheet accounts carry over across years; result accounts, never. A balance remaining in a class 6 or 7 account on January 1 is always an unperformed closing.

How it's done

All revenue accounts are debited with their balance, and account 121 Profit or Loss is credited. All expense accounts are credited with their balance, and 121 is debited. In the end, the only thing that remains is the balance of account 121.

A concrete example: revenues of 500,000 lei, expenses of 420,000. After closing, account 121 has a credit balance of 80,000 lei. This is the figure that appears in the income statement as the gross result for the fiscal year.

The tax, which comes afterward

The order here is often confusing, because it appears circular: corporate income tax is an expense, so it should be closed into 121; but it is calculated based on the result, which is obtained precisely from 121.

The actual order has three stages. First, all revenue and expense accounts are closed, except for the tax account. This yields the gross result. Based on it, the tax is calculated and recorded, and then the tax account is also closed into 121.

What remains in 121 after the third stage is the net result — the profit that can actually be distributed. In the example above, with a tax of 13,280 lei, the net result is 66,720.

What is NOT closed

This is worth stating explicitly, because it is the mistake that empties reports: only the accounts in classes 6 and 7 are closed. All others — assets, liabilities, equity — keep their balance and carry over into the following year as opening balances.

Account 371 keeps the merchandise in the warehouse. Account 401 keeps the payables to suppliers. Account 5121 keeps the money. A closing that touched them would wipe out the company's assets, not its result.

The practical consequence, for anyone reading reports: after the annual closing, any report built on the turnover of the closed year's result accounts shows zero. Not because the activity has disappeared, but because it has been moved into 121. This is why financial statements are prepared on the trial balance BEFORE closing.

What comes next

We used a figure twice above without explaining it: the tax of 13,280 lei on a profit of 80,000. Anyone calculating 16% of 80,000 gets 12,800, not 13,280. The difference is not a mistake — it is the subject of the next chapter.

The three stages, in figures

The order of the three stages is best understood by tracking what account 121 contains after each one. Here is a year with 500,000 lei in revenues, 420,000 in expenses (excluding tax), and a tax of 12,800.

Anyone who closes all the expense accounts at once, including 691, obtains a gross result already reduced by the tax — and if the tax is then calculated on it, the resulting figure is smaller than the amount actually owed. The error is self-perpetuating: each recalculation deepens it.

What remains in 121 after December 31

Unlike accounts 6 and 7, account 121 is NOT settled at the closing of the fiscal year. It keeps the result until its distribution, which takes place in the following year, after the financial statements are approved.

So a balance in 121 on January 1 is normal and expected — unlike a balance in 607 or 707, which is always an unperformed closing. It is a small and often confused distinction, and the confusion leads to additional "closings" that empty the result before it has been distributed.

On January 1, the revenue and expense accounts must be ZERO. Account 121 is not — it carries the result until distribution, that is, until the shareholders decide what to do with it.

The microenterprise: same path, different base

For companies paying microenterprise income tax, the closing mechanism is unchanged, but the step between the stages looks different: the tax is not calculated on the result, but on revenues, and not at the end of the year, but quarterly.

The consequence is that, on December 31, the tax expense is already recorded — four quarters, already declared and paid. The third stage still exists, but it closes an expense that no longer depends on the result just calculated. The order remains the same; only the dependency between the steps disappears.

Part IX · Closing · phases 10–11

Chapter 43

From accounting result to tax result

A business owner who sees a profit of 80,000 lei and a tax rate of 16% expects 12,800 lei. They will pay 13,280. The difference doesn't come from a calculation error, but from the fact that the tax isn't applied to the accounting profit.

It is applied to the taxable result — a different figure, obtained from the first through adjustments required by law. This chapter shows where the adjustments come from and why they aren't optional.

Two results, two purposes

The accounting result measures the economic performance of the company. The taxable result measures the base that the state chooses to tax. They are different questions, so they have different answers.

The state refuses, for instance, to bear through lower tax a fine received by the company — otherwise it would be paying part of the sanction it itself imposed. The expense is real and reduces accounting profit; for tax purposes, it is not recognized.

A non-deductible expense is not a refused expense. It is a real expense, paid from the company's money, which the state does not deduct from the base.

The calculation, step by step

Note that the adjustments go both ways. Non-deductible items raise the base; non-taxable revenues and tax deductions lower it. It is not a penalty mechanism, but an alignment with a different definition of gain.

The legal reserve: the deduction that isn't an expense

The fourth row in the table deserves separate attention, since it is the one most often omitted — and omitting it costs money, not just accuracy.

The legal reserve is mandatorily set up through the allocation of profit, not through recording an expense. So it never appears in any class 6 account and cannot reach the tax base through the usual route.

The law nonetheless recognizes it as a deduction. A company that does not deduct it still must set up the reserve and pays 16% on it — in the example above, 640 lei per year, lost with nothing in return.

The two kinds of non-deductible items

This distinction is the most useful in the whole chapter, because it explains why the calculation cannot be done "by eye".

The confusion between the two is among the most costly in tax practice, because it produces errors of tens of percentage points, in either direction, without anything flagging them.

Carried-forward loss

A year with a loss is not permanently lost: the tax loss is carried forward and deducted from the profits of subsequent years. However, two limits apply, and both have changed recently.

The first is quantitative: the loss can be recovered up to a limit of 70% of the year's taxable result, not in full. The second is temporal: losses are carried forward for a limited number of years, and those not recovered within this interval are permanently lost.

The practical consequence is that losses must be tracked by year, not as a cumulative sum. A single sum has no age, so it cannot expire — and a company that keeps track this way will, silently, calculate too little tax, year after year. The correct order of consumption is from oldest to newest, precisely because the oldest expires first.

The tax records register

All these adjustments cannot be kept in memory and cannot be reconstructed a year later. The law therefore requires a dedicated register, in which each adjustment is recorded with its amount and its legal basis.

The register is not a formality: it is the only document that can explain, during an audit, why the declared tax differs from 16% of the balance sheet profit. Without it, the difference must be defended from memory — and memory is not evidence.

What's next

The tax is calculated, and the net result is established. What remains is the question the business owner has been waiting to have answered from the start: what can be done with it. This is the final chapter of this part.

A complete case, from trial balance to tax return

The rules covered so far are best seen laid out one after another, on the same company. Here is a full year, with all four kinds of adjustments at once — fixed percentage, cap on a base, deduction that isn't an expense, and credit deducted from the tax.

Step one: non-deductible items, in their order

The order is not indifferent, since some adjustments are calculated on a base that includes others. Fixed percentages per account are applied first, then caps that depend on a calculation base.

Step two: the deduction that isn't an expense

The legal reserve is deducted from the taxable result without ever having passed through an expense account. On a gross profit of 120,000 and capital of 200,000, the 5% rate gives 6,000 lei, and the cumulative cap of 20% of capital — 40,000 — leaves room for the entire amount.

Step three: the credit deducted from the tax

Sponsorship is not deducted from the base, like the others, but from the tax already calculated — which is why it could not be calculated together with them. It has two caps, and the smaller one applies.

Final tax is 25.432 minus 3.750, or 21.682 lei. The 10.000 lei sponsorship cost the company 6.250 lei in the current year; the balance was absorbed by tax that no longer had to be paid. Report the beneficiary and amounts in D107. Had the company not used the entire ceiling, the available amount could have been redirected through D177 under Article 42 of the Tax Code.

Three kinds of adjustment, three moments: fixed percentages and caps raise or lower the BASE; the legal reserve lowers it without being an expense; sponsorship is deducted from the TAX. Anyone who mixes them up gets a wrong result even if each individual figure is correct.

Research and development in 2026: enhanced deduction or tax credit

For eligible research and development activities, 2026 introduced a choice with a material effect: a corporate income tax payer may apply the incentive based on the enhanced deduction for eligible expenditure or elect the 10% tax credit governed by Article 20^1 of the Tax Code. The two mechanisms cannot be applied to the same eligible expenditure.

The file begins with the technical owner, not accounting: project objective, novelty, technological uncertainty, activities and people involved, timesheets or time allocations, materials consumed and the criterion separating the work from ordinary commercial activity. Accounting then bridges those records to the expense accounts. A trial-balance subledger labelled “R&D” does not, by itself, establish eligibility.

Part IX · Closing · phases 10–11

Chapter 44

Distribution of the result

The net profit now exists as a balance in account 121. But it is not at anyone's disposal: until distribution, it is a figure, not an available sum.

Distribution is the last step of the closed financial year and, at the same time, the first of the next one — because it happens after the financial statements are approved, that is, in the following year.

The order, which is legal

Three acts must occur in this order, and none can be skipped: the financial statements are prepared, the general meeting of shareholders approves them and decides on the distribution, and only afterward is the distribution recorded and the dividends can be paid.

A business owner who takes money out of the company before this order has not received dividends: they have taken a loan from their own company, with all the tax consequences that follow. The distinction seems formal and is among the most expensive in practice.

A dividend is not a withdrawal of money. It is the outcome of a decision that cannot be made without approved financial statements.

The legal reserve comes first

From the profit, the legal reserve is set up first: 5% of the gross profit, until the reserve reaches one fifth of the subscribed and paid-up share capital.

It is an unavailable reserve — it cannot be distributed and can only be used under the conditions set by law. Its purpose is creditor protection: part of the earnings stays in the company, as a buffer.

In accounting terms, the setting up of the reserve passes through an intermediate account, 129 Profit Distribution, which is then closed into 121. This detour seems unnecessary, but it is what makes the distribution appear distinctly in the income statement, not as a simple transfer between equity accounts.

What remains, in figures

Out of 80,000 lei of gross profit, the shareholder actually receives 52,684.80. The rest went to corporate income tax — 13,280 —, to dividend tax — 10,035.20 —, and to the legal reserve — 4,000, which nevertheless remains in the company.

It is worth telling a business owner clearly: between the profit in the balance sheet and the money that reaches them, there are two successive taxes, not one. This is the figure on which the decision to distribute or reinvest is made, and it appears nowhere in the financial statements.

The loss

When the financial year ends with a loss, the mechanism is symmetrical and simpler: the debit balance of account 121 is carried into retained earnings, where it awaits coverage from future profits.

It is neither erased nor ignored. A company with a loss carried forward cannot distribute dividends before covering it — the rule is economic common sense: you do not divide up a gain that has not yet recovered a previous loss.

Interim dividends

The law also allows distribution during the year, based on interim financial statements. It is a real facility, but with a condition that is often forgotten: if at the end of the financial year the result turns out to be smaller than what was distributed, the difference must be returned.

It is, therefore, an advance on a profit that has not yet occurred. Treated as such, it is useful; treated as a final distribution, it produces, at year end, a receivable from the shareholders themselves, which no one had anticipated.

What comes next

With the distribution, the financial year is complete: recorded, verified, declared, closed, and distributed. What remains is the part for which everything else was done — the documents that state, in one page, what the company did in a year.

How much can actually be distributed

Net profit is not automatically the amount that can be divided up. From it, three things are subtracted, in order — and what remains is called distributable profit.

Distributing beyond this limit is not an accounting error, but an act with consequences: the amounts must be returned, and those who approved it are liable for it. This is why the general meeting's decision must be based on approved financial statements, not on an estimate.

The reserve cap, calculated

The legal reserve is not built up indefinitely. It accumulates until it reaches one fifth of the subscribed and paid-up share capital, and from that point on the obligation ceases — along with its deductibility.

The last two rows show the rule: 5% of gross profit is set up, but no more than the space remaining up to the cap. For a company with high profits, the cap is reached within a few years, and the reserve ceases; for one with small profits, the obligation accompanies the company for a long time.

The legal reserve is not an expense and does not leave the company. It is part of the earnings that the law prevents from being distributed — a buffer for creditors, not a loss for the shareholders.

Distribution tests applicable in 2026

Law 239/2025 also introduced new minimum share-capital thresholds for SRLs: 500 lei for newly incorporated companies and at least 5.000 lei when reported net turnover exceeds 400.000 lei. Existing companies have the statutory transition period to comply, while companies increasing capital by 31 December 2026 benefit from a reduced publication fee. Keep this control in the same file because capital affects both the net-assets limit and the legal-reserve ceiling.

Part X · The financial statements · phase 12

Chapter 45

The Income Statement

We arrive at the documents for which all the rest was done. Forty-four chapters of entries, checks, and closings come together in two pages — and the first of them answers the question with which the book began: did you earn or did you lose?

The income statement is a flow statement: it covers a period, from January 1 to December 31, and shows what happened within it. It says nothing about what the firm has — that's the balance sheet's job.

Structure: two activities, two results

The form does not lump all revenues and all expenses into a single pile. It separates them into two activities, and this separation is the main piece of information the document provides.

The operating result tells you whether the business works. The net result tells you what's left after everything. A firm can have the first positive and the second negative — and then the problem isn't the business, but how it's financed.

On real figures

Here is the situation of a small firm, generated from its own accounting records — the same one from the chapters on ledgers.

The breakdown in the note is the document's second piece of information, and it deserves careful reading. At this firm, goods sold represent almost two-thirds of expenses, and salaries the rest. This structure tells you what kind of firm it is — a trader, not a service provider — without needing to ask anyone.

Why the money doesn't appear anywhere

This is the question every business owner asks on first reading: if I have a profit of 420.83 lei, where is it?

The answer is that it's nowhere, because profit is not an amount of money. It's a measure of the difference between value created and value consumed — and the two are recognized when they occur, not when they are paid.

The depreciation of 466.67 lei in the example illustrates this perfectly: it's an expense that reduced profit without taking a single leu out of the account. Symmetrically, an invoice issued and not collected produced revenue without bringing in any money. It's the same thing Chapter 1 was saying, now arrived at its final form.

How to read it, in five seconds

An experienced reader does not go through the document from top to bottom. They look at three things and usually form an opinion before reading the rest.

First, the sign of the operating result — if it's negative, nothing else matters much. Then the ratio between the operating result and revenue, i.e., the margin. Finally, the financial result: if it swallows a significant part of the operating result, the firm is working for the bank.

The comparison with the previous year, which appears on the form as a second column, adds a fourth piece of information and is the most valuable of all: not how much, but which direction.

What comes next

The income statement has answered the question about the period. What remains is the question about the moment: what does the firm have and what does it owe, at a given date. That's the balance sheet, and the equality that defines it is no coincidence.

What an outside reader looks at

The form has the same lines for every firm, but whoever reads it isn't looking for the same things. It's worth knowing what each one is after, because that shows which line has to hold up under scrutiny.

The structure of expenses tells you what kind of firm it is

The breakdown of expenses by nature — goods, materials, personnel, depreciation, services — is the document's second piece of information, and from it you can read the business model without asking anyone.

This ratio is useful over time too, not just between firms. An increase in the share of salaries without a corresponding increase in revenue means the firm hired ahead of orders; an increase in depreciation means it invested. Both are decisions, and the income statement shows them a year later than when they were made.

What can't be seen from it

The document says nothing about when the money comes in — that's the cash flow statement's job — nor about what the firm has, which is the balance sheet's job. But it has another, less obvious limitation: it doesn't distinguish what's repeatable from what happened only once.

A firm that sold a piece of equipment this year at a gain has a better result, but not a better business. That's why disposals of assets are presented separately, and anyone analyzing a firm first subtracts the non-recurring items — an operation not required by any form, but without which the comparison with last year says nothing.

Part X · The financial statements · phase 12

Chapter 46

The Balance Sheet

The balance sheet is a photograph. The income statement told the story of a whole year; the balance sheet shows a single moment — usually December 31, midnight — and states what the company has and to whom it owes.

It has two sides, and they are always equal. Not because the company is "balanced," but because the two sides describe the same thing from two angles: what exists, and where it came from.

The Two Questions

The assets answer the question "what does the company have": goods, inventory, receivables, cash. The liabilities and equity answer "what was it financed with": owners' money, undistributed profit, debts.

Every leu in the assets came from somewhere. Either the owners put it in, or the company produced it and kept it, or it is owed to someone. There is no fourth option — and that is where the equality comes from.

Assets = Liabilities + Equity is not a check that may or may not balance. It is the same reality counted twice: once by form, once by origin.

A Balance Sheet, Read

The figure of 80,000 lei in the note is the most important one in the whole document and, paradoxically, does not appear as a separate line in the usual reading. It answers the question "how much is the company worth to its owners," and its evolution from year to year is the most honest measure of success.

The Order of the Lines Is Not Random

Assets are arranged in increasing order of liquidity: first what is hardest to convert into cash — buildings, equipment — and last, cash itself.

Liabilities and equity follow the increasing order of enforceability: first what never has to be returned — the owners' capital —, then long-term debts, then those payable in the coming months.

Arranged this way, the two columns can also be read horizontally. An asset that is hard to convert into cash, financed from short-term debt, is the classic signal of a cash-flow problem that has not yet occurred. The comparison between the bottom of the assets and the bottom of the liabilities — current assets versus current liabilities — is the simplest financial health indicator there is.

What the Balance Sheet Does Not Show

This is worth saying, because otherwise the document is overrated. The balance sheet does not contain the market value of the company, but the historical cost of its elements, adjusted for depreciation and impairment.

It also does not contain the brand, the customer base, or the team's expertise — real assets, sometimes the most valuable ones, which do not meet recognition criteria. A services company with thirty good employees may have a smaller balance sheet than a warehouse full of goods.

That is why the balance sheet is not read alone. Read alongside the income statement and the cash flow, it says a lot; read in isolation, it says less about the company than it seems to.

What Comes Next

The two main documents are ready. However, they are not the only ones filed — and among those that accompany them, one answers the question that neither the balance sheet nor the income statement touches: where did the money go.

The Official Form Does Not Have Two Columns

The balance sheet in this chapter is presented in two columns, because that is the easiest way to understand the equality. The official form, however, looks different: it is VERTICAL, and the equality is not seen at a glance — it is reached at the end, through a chain of subtractions.

Line E deserves attention, since it is the only one that does not appear in the two-column form and is perhaps the most useful in the whole document: net current assets, i.e. what remains of short-term assets after paying short-term debts. Here, 50,887.50 lei — a company that can cover its near-term obligations more than three times over.

When line E is negative, the company has more to pay in the next twelve months than it can convert into cash in the same period. This is not an accounting error and does not appear anywhere in the income statement — but it is the first thing a bank reads.

The vertical form does not hide the equality, it postpones it — and along the way, it calculates an indicator that the two-column form does not produce at all.

Abbreviated or Complete

The form has two versions, and the choice is not free: it depends on the category the company falls into, according to size criteria — total assets, turnover, and average number of employees.

Micro-entities and small entities file the abbreviated balance sheet, with the large groups. Medium and large ones file the complete form, with over a hundred lines, where each group is broken down into elements. The economic content is the same; what differs is how much of it is visible.

The move from one category to another does not happen at the first exceedance: the criteria must be exceeded for two consecutive financial years. The rule exists precisely so that one good year does not change the reporting regime of a company that returns, the following year, to its usual size.

Part X · The financial statements · phase 12

Chapter 47

The Annexes

The balance sheet and the income statement are the documents that get read. Alongside them, three more are filed — less well known — and one of them answers a question the first two never touch at all.

That question is the one a business owner asks most often, and the one the accountant finds hardest to answer: if I have profit, where is the money?

The Cash Flow Statement

The cash flow statement tracks money exclusively: how much came in, how much went out, and from which activity. It has nothing to do with the moment the revenue or expense was recognized — here, only the timing of the actual cash movement matters.

Here is the statement of the company from the previous chapters, generated from its own records.

Operating cash flow is the figure banks read before profit. A large profit with negative cash flow means the company is selling, but not collecting.

This statement finally explains the paradox every business owner runs into. The profit from Chapter 45 and the cash change here are different figures, and the difference between them always has a name: depreciation, inventory growth, uncollected receivables, paid-off liabilities.

The Statement of Changes in Equity

The second annex is simpler and answers an owner's question: how did the shareholders' net worth evolve during the year.

It has one row for each equity item — share capital, reserves, retained earnings, the result for the period — and four columns: opening balance, increases, decreases, closing balance. The total closing balance must equal the equity shown in the balance sheet.

Its practical usefulness shows up at companies with capital movements: increases, distributions, loss coverage. At a company without such transactions, the only movement is the year's result — and the statement becomes a confirmation, not information.

The Explanatory Notes

The third annex is the least read and the richest. The notes explain what lies behind the figures: the accounting policies chosen, the valuation methods, the breakdown of the main line items, the commitments that don't appear on the balance sheet.

This is where you write which inventory issuance method the company uses, which depreciation periods it chose, how it calculated receivable adjustments, what litigation is ongoing, what guarantees it has given. These are exactly the decisions that the previous chapters said, every time, must be documented.

For an outside reader, the notes are the only place where the figures become comparable. Two companies with the same profit, one of which depreciates on an accelerated basis and the other on a straight-line basis, are not saying the same thing — and the difference is visible only here.

What Comes Next

The five documents are ready. Before signing, one more operation remains, which takes ten minutes and catches most serious errors: checking the consistency between them.

The Two Methods of Cash Flow

Operating cash flow can be built in two ways, and the result is identical — only the starting point changes, and with it, what you see along the way.

The indirect method is the one that clarifies the "I have profit, I have no money" paradox, because it lists the very causes: depreciation, which reduced profit without taking out cash; inventory and receivables, which tied up cash without touching profit; liabilities, which brought in cash without generating revenue.

The Statement of Equity: What It's Actually For

At a company with no capital movements, the statement of changes in equity looks like a formality — a single line moving, the year's result. This is exactly the case of the company in our example.

The value of the statement shows up at companies with movements: a capital increase, a dividend distribution, a loss coverage, a revaluation. There, it answers a question the balance sheet doesn't touch — not how much the shareholders' wealth is, but WHAT it changed FROM during the year.

And the difference between these two questions is exactly the difference between a photograph and a film, which this part began with: the balance sheet shows where things ended up, the statement of equity shows by what path.

Part X · The financial statements · phase 12

Chapter 48

The four cross-checks to verify before signing

This chapter is a list of four checks. Together they take less than ten minutes, can be done on paper, and catch most serious errors in a set of financial statements.

It deserves to be treated as a ritual. Not because it's legally required, but because the signature on the last page creates personal liability, and those ten minutes are the cheapest thing in the entire exercise.

First: assets equal liabilities

Total assets must equal total liabilities, to the leu. Not approximately, not with an acceptable rounding difference — exactly.

As seen in chapter 46, the equality holds by construction: if it doesn't come out right, an account has fallen into two rows or into none. The most frequent cause is a new account, which appeared during the year, that the mapping doesn't know — and it doesn't announce itself, except through this discrepancy.

Second: the same result in two places

The result of the financial year appears in two documents: as a line in the balance sheet, among equity, and as the last line of the income statement.

The two must coincide. When they don't, the explanation is almost always the same: one of the documents was prepared on the trial balance before the closing of result accounts, and the other on the one after.

A correct balance sheet next to an income statement showing zero is not a matched pair of documents. It's the same document prepared twice, on two different trial balances.

Third: cash flow ending balance equals balance sheet balance

The amount at which the cash flow statement ends must be exactly the balance of the cash and bank accounts in the balance sheet. In the example from the previous chapter: 46,915 lei in both places.

This is the check that catches omissions in the flow — a forgotten bank account, a foreign-currency cash register not converted, a letter of credit treated separately. All produce the same discrepancy, and none of them shows up any other way.

Fourth: the change equals the sum of the three activities

The difference between year-end cash and beginning-of-year cash must equal the sum of the operating, investing, and financing flows. In the example: 46,915 minus 45,000 makes 1,915, and 1,915 plus zero plus zero also makes 1,915.

This check catches a different class of errors: misclassified transactions. A loan repayment recorded under operating instead of financing does not change the total change — but it completely changes the interpretation of the document, and if the sum doesn't come out right, it means a transaction was counted twice or not at all.

What they don't catch

It's worth saying, in the spirit of chapter 33: the four cross-checks are internal-consistency checks. They confirm that the documents agree with each other — not that what they say is true.

An overvalued inventory item, an unadjusted doubtful receivable, an expense recorded in the wrong year — all pass all four. For these there are other tools, described in their proper place: the inventory count, aging analysis of balances, comparison with supporting documents.

The cross-checks are the last gate, not the only one. Their purpose is to guarantee that, if an error still exists somewhere, it's a substantive error — not one that anyone could have caught by placing two figures side by side.

What comes next

With the signature on the financial statements, the financial year is closed. But the obligations do not end there: what remains is what happens after everything has been filed — the files, the traces, and the boundary between what software does and what always remains a human's responsibility.

What the validator checks, and what it doesn't

Before filing, the set goes through an official validator that rejects inconsistent forms. It's a good net, and a limited one, and its limit must be known: the validator looks inside the form, not outside it.

From this follows the correct order of checks: first the cross-checks with the accounting — the four in this chapter — then the validator. In reverse, a passed validator gives false reassurance, and the errors it cannot see remain undiscovered until an inspection.

The fifth and sixth cross-check

The four in the title are the blocking ones. There are two more that do not stop the filing and that catch equally serious errors.

It's worth noting why the second cannot be checked automatically by the validator: it does not have the prior year's financial statements. The only one who has them is the one who prepared them — which is why this check remains an obligation of the company, not of the system.

Order matters: cross-checks with the accounting first, the validator afterward. It is the last gate before filing, not the first check of correctness.

Part XI · The trace · phase 14

Chapter 49

Archiving

The bookkeeping is finished. The financial statements are signed and filed, the returns have receipts, the fiscal year is closed. This part is about what remains — and about the day, two or five years from now, when someone starts asking questions.

That day comes more often than people think: a tax inspection, a loan application, the sale of the company, a dispute with a partner. And the question is always the same: show me the document.

The monthly file

The archiving unit is not the document, but the month. A complete monthly file contains everything an outsider would need in order to reconstruct that month without asking anyone.

A reconstructed file contains what was found. A file put together on time contains what was supposed to exist. The difference between the two is exactly what an inspector is looking for.

The year's file

On top of the twelve monthly files, an annual one is added, with the documents concerning the entire fiscal year: the financial statements with all annexes, the declaration of responsibility, the inventory register and the inventory documents, the tax records register, and the resolution approving and distributing the result.

It is worth building as a closed whole, not as a box you keep adding to indefinitely. A closed fiscal year no longer changes — and its file should have the same property.

In an electronic archive, this property can even be proven: each file receives a cryptographic fingerprint, and the entire file set receives a combined one. Any later modification to any file changes the final fingerprint, so it becomes visible. It is the digital equivalent of the seal on a paper file, and it's harder to forge.

How long to keep it

Retention periods are set by the accounting law and have been amended several times in recent years. This is one of the few figures in this book that we don't write down: it should be checked against the text currently in force, not recited from memory.

What doesn't change is the underlying principle: documents must exist for at least as long as the tax authority can still establish obligations for that period. A document destroyed before that deadline expires leaves the company defenseless exactly in the situation the document was meant to cover.

There is also a category with its own rules, worth noting separately: documents used to establish employees' pension rights. A former employee may need a certificate more than twenty years later, and the company is obligated to be able to issue it.

An electronic archive doesn't keep itself

A paper archive degrades slowly and visibly. An electronic one disappears instantly and silently — a failed drive, an expired storage account, a format nothing can open anymore.

That's why a backup that has never been restored is not a backup, but an assumption. The only proof that an archive works is its actual restoration, done periodically, as an exercise.

The second rule is just as simple and just as often broken: the copy must be kept in a different place than the original. An archive kept on the same computer as the bookkeeping solves accidental deletion and nothing else — not fire, not theft, not encryption.

What follows

The files answer the question of what was recorded. One question remains, which the documents don't answer: who did each thing, and when.

When the document is missing: reconstruction

The rule from chapter 9 says that without a document, nothing is recorded. It does not, however, address what happens when the document existed and was lost — and for that there is a procedure, with steps and deadlines.

What cannot be reconstructed are documents attesting to movements of cash or valuables — receipts, fiscal receipts, cheque leaves. For these there is no duplicate: their value lies precisely in their uniqueness, and the loss is treated as a shortage, not as a document to be recreated.

The electronic archive: three conditions, not one

"We keep everything scanned" is not an electronic archive. The law requires three things at once, and an archive that meets only two of them fulfills none of the purposes for which it exists.

From this comes a practical choice worth making consciously: keep the archive in open, widely used formats — PDF, XML, CSV — not in the internal format of the program you use today. It's the same logic behind encrypting the offsite copy with a standard utility rather than a proprietary format: recovery must not depend on the very program that was just lost.

And the destruction of documents upon expiry of the retention period is done based on a report, with management's approval. Without it, during an inspection, documented destruction cannot be distinguished from loss — and the consequences of the two are not the same.

Part XI · The trace · phase 14

Chapter 50

Who did what, and when

Accounting records economic transactions. It does not, by itself, record human actions: who entered an invoice, who changed a price, who deleted a draft, who logged in at three in the morning.

For that there is a second log, running parallel to the accounting one, and its purpose is frequently misunderstood — including, or especially, by those who install it.

It's not surveillance

Any team's first reaction to the idea of an operations log is that someone is watching them. The reaction is natural and, in the overwhelming majority of cases, wrong.

An audit log doesn't exist to catch anyone. It exists so that an inevitable question can be answered: why does this figure look like this. And the answer "it was modified on March 14, by user X, who noted reason Y" closes the discussion in thirty seconds.

An operations log protects, above all, the person who did things correctly. They're the only one who can prove it.

Without a log, in a three-person team, every error becomes a discussion about who might have done it. With a log, it becomes a finding, and most of the time the finding is that the work was done correctly and the problem lies elsewhere.

What gets recorded

Not everything. A log that records every click becomes unreadable, and an unreadable log is equivalent to no log at all. What gets recorded are actions that change something or that concern access.

Note that the list does not include ordinary reads. Who looked at a report is not useful information and, accumulated, it drowns out exactly the events that matter.

Two properties without which it serves no purpose

First: the log must be append-only. It is written to the tail and never rewritten. A log from which lines can be deleted proves nothing, because the absence of a line cannot be distinguished from the event never having occurred.

Second: it must survive the database. If the log lives in the same database it audits, it disappears exactly in the situation it existed for — a corruption, a botched restore, a deliberate deletion.

The practical solution is duplication: a live, capped record for quick reference, and a separate, append-only file included in the backups. The two complement each other, and the cap on the first is justified precisely by the existence of the second.

A log that falls silent

It's worth closing with the least obvious observation in the chapter, because it applies to any control mechanism, not just this one.

Writing to the log is, by nature, secondary: if it fails, the main action must not be blocked. No one wants an invoice to fail to save because a log file is full. The consequence, however, is that a log that can no longer write looks exactly like one in which nothing happened.

That's why the ability to write must be actively checked, not discovered when you need proof. It's the same rule as with the backups in the previous chapter and with filing declarations in chapter 38: the absence of an error signal is not confirmation.

What's next

With the archive and the log, everything that could be recorded has been recorded. Only one chapter remains, and it's about a boundary: what, of everything the book has described, no program can do.

Part XI · The trace · phase 14

Chapter 51

What a program cannot do

This book was written starting from how an accounting program works, and fifty chapters have shown how much can be automated: entries composed from document types, VAT returns compiled from records, trial balances checked, declarations generated, controls run continuously.

The closing chapter is about the other side — and it is the most important one of all, precisely because the power of automation makes its limits easy to forget.

Three things it does well

They deserve to be listed first, so that the limit stands out by contrast. A program calculates without getting tired, checks without skipping a line, and remembers without forgetting.

All three are qualities a human being lacks. A tired accountant makes calculation errors; a rushed one skips a check; anyone can forget a deadline. Nothing that follows contradicts the fact that, on these three points, the machine is decisively better.

What it cannot do

A program answers questions about data. Accounting, at its decisive points, asks questions about reality.

The example that sums them all up

An invoice for five thousand lei for "consulting services." The document is compliant: it has all the required elements, the supplier has a valid VAT code, the amount matches the payment on the bank statement, and the program can record it correctly without any intervention.

The only question that matters appears nowhere in the data: was the consultancy actually provided? And if it was, was it related to the business's activity?

The answer cannot be read from any field, cannot be inferred from any correlation, and cannot be checked by any rule. Someone knows it — the administrator who ordered the service — or no one knows it, which is itself already an answer.

The consequences are real: the deductibility of the expense, the right to deduct VAT, and in extreme cases, the legal classification of the transaction. All of them hinge on information that is not in the system.

Why automation does not reduce responsibility

The opposite argument is often heard: if the program does everything, the accountant is less responsible. The reality is the opposite, and it is worth understanding why.

Automation does not eliminate decisions; it moves them earlier. The choice of inventory valuation method, the depreciation period, the materiality threshold, the treatment of an unusual transaction — all of these are made once and then applied automatically to thousands of transactions.

A wrong decision made manually affects one document. The same wrong decision, embedded in an automatic rule, affects an entire year. The power of the tool increases, rather than decreases, the consequences of the judgment that configures it.

The correct division of labor

Everything the book has described ultimately suggests a simple, stable division of labor between human and machine.

The machine does what is repetitive, verifiable, and tedious: it calculates, adds, compares, flags. The human does what requires knowledge of the reality behind the documents and what carries responsibility: classifying, judging, deciding, signing.

The boundary between them is not technological but logical. It will not shift as programs advance, because it separates questions that can be answered from data from questions that cannot.

Closing

The book began with a simple statement: accounting does not track money, it tracks profit. Fifty chapters later, the statement can be restated more precisely.

Accounting is the discipline of telling the truth about an economic activity, in a form that someone else can verify. Its rules seem arbitrary until you understand that each one solves a problem of trust: double-entry bookkeeping makes errors visible, the supporting document makes statements probable, the reversal entry preserves the trail of a mistake, the closing of a period makes the past stable.

A program can apply all these rules faster and more consistently than a human. What it cannot do — and what therefore remains the essence of the profession — is know whether the recorded event actually happened, and answer for the claim that it did.

When the machine misreads

Automatic data extraction from a document has changed day-to-day work more than anything else in recent years. But it has also changed the nature of errors: they are no longer typing errors, but errors of understanding — and they look different.

The classic example is a VAT rate read from the wrong line of the invoice: 11% instead of 21%. The amount is plausible, the arithmetic checks out if the base was read from the same place, and the document appears in order. It is discovered only at the VAT return stage, or during an audit — and, in the meantime, the same mistake has repeated on every invoice from the same supplier.

A human makes a mistake once. A machine makes the same mistake every time — which is easier to fix, but only after someone has noticed that it is making it.

Why confirmation is not a formality

The step where the operator confirms the data read seems like a checkbox, and by the hundredth invoice it becomes one. Yet it is the only point in the entire flow where a human looks at the document and the figure at the same time.

A well-designed flow does not require confirming every field, only the ones that matter and the ones with low confidence: the rate, the counterparty, the total. The rest is accepted tacitly — not out of carelessness, but because attention demanded everywhere is attention given nowhere.

And when a business chooses automatic posting — without confirmation — the decision must be made knowingly: time is gained, and exactly the moment when one would have noticed that the machine is systematically misreading is lost. The reasonable compromise is automatic posting only when ALL checks pass, as shown in Chapter 13, plus spot-checking of what has been posted.

What remains, in the end

The book ends where it began: with two questions. The first — how much have I earned — received an answer that can be calculated. The second — what do I have and to whom do I owe it — likewise.

What cannot be calculated is whether they are true. Their truth depends on documents that were read, classifications that were chosen, estimates that were made — and each of these is someone's decision. Accounting does not produce the truth; it produces a verifiable statement, and someone signs it.

Part XII · Professional judgment · cross-cutting layer

Chapter 52

Accounting policies, estimates, errors and materiality

The same 40.000 lei difference can represent three entirely different matters: a new accounting policy, a revised estimate or a prior-period error. The amount is identical; the treatment, affected period, account and note disclosure are not. Professional judgment begins with diagnosing the nature of the change, not with choosing debits and credits.

Classify the issue correctly before selecting the journal entry. Flawless bookkeeping applied to a wrong classification still produces misstated financial statements.

Hierarchy that prevents improvisation

Classification test in four questions

accounting policy is not a fiscal choice

Accounting policies must produce a true and fair view under the applicable accounting regulations; tax rules determine the taxable base. The values may coincide, but that does not merge the two calculations. An item may qualify as a fixed asset for accounting purposes because of its nature, use and the entity’s capitalisation policy, even when its value is below the 5.000 lei tax threshold applicable from 2026. Track the difference in the tax records register; do not “solve” it by distorting the accounting classification.

materiality: not a magical percentage, but a conclusion

Information is material when omitting or misstating it could influence users’ decisions. Size is only the first measure. Nature can make a small amount material: a transaction with a director, a covenant breach, turning a loss into profit, concealing fraud, crossing a tax threshold or affecting the right to distribute dividends.

How to correct without rewriting history

Under OMFP No. 1.802/2014, material prior-period errors are corrected through retained earnings, using account 1174. Immaterial errors may be corrected through current-period profit or loss when the entity’s policy consistently provides for that treatment. Financial statements already approved and filed are not rewritten in the accounting records; the nature and effect of the correction are disclosed where required. Any affected tax returns are assessed separately and amended when necessary.

Account 1174 is not a catch-all for old problems. Use it only after demonstrating that an error exists, belongs to a prior period and is material under the documented policy.

Professional judgment memorandum

Part XII · Professional judgment · cross-cutting layer

Chapter 53

Provisions, subsequent events and going concern

Year-end closing is not limited to counting items that exist. It also requires recognising obligations that do not yet have an invoice, final amount or certain due date, and using information arising after 31 December to understand the conditions that actually existed at 31 December.

Uncertainty is not a reason to omit an item. It is a reason to identify the obligation, assess the probability of an outflow and develop the best supportable estimate.

Decision tree for an uncertain obligation

Estimating a provision

The best estimate is not the most prudent amount or the amount the management hopes to pay. For a large population of obligations, the probability-weighted expected value can best describe the exposure. For an individual obligation, the most likely result may be the starting point, adjusted if other results have a significant probability. Risks, uncertainties, repayments from third parties and the effect of value over time shall be treated according to regulation, without doubling adjustments.

Review provisions at each balance-sheet date and adjust them to the current estimate. Reverse a provision to income when an outflow is no longer probable. Use it only for the obligation for which it was recognised; it is not a reserve for smoothing profit between years.

Events after the balance sheet date

The period between the balance-sheet date and the date the financial statements are authorised for issue is not informationally closed. Classify events arising in that interval by what they evidence: conditions existing at the balance-sheet date or conditions arising afterwards.

Register of subsequent events

A professional does not wait for events to reach accounting by chance. Send written enquiries to management, legal counsel, sales and treasury; inspect subsequent receipts and payments, minutes of governing bodies, new contracts, notices, court portals and lender correspondence. Record the event, date identified, condition existing at year-end, adjusting/non-adjusting conclusion, amount and disclosure.

Going concern: a conclusion, not a tick

Financial statements are normally prepared on a going-concern basis. That basis is no longer appropriate when the governing bodies decide to liquidate or cease operations, or when no realistic alternative exists. Uncertainties that may cast significant doubt must be assessed and clearly disclosed. The director’s declaration required by the Accounting Law does not replace its supporting analysis.

Baseline scenario, severe scenario, response plan

Start the cash-flow forecast from reconciled bank balances and receivables and liabilities by due date, not from the annual income budget. Use approved assumptions for the base case. A severe but plausible case delays receipts, reduces sales or removes uncertain funding. For every shortfall, state the response, date, owner, amount and whether the action is under the entity’s control. Conclude whether a material uncertainty exists and which disclosure is required.

Part XII · Professional judgment · cross-cutting layer

Chapter 54

Related parties, shareholders, capital and financing

Cash moving between a company and its shareholders looks like any other cash in the bank statement. In substance it may be capital, a loan, a dividend, an expense advance, a reimbursement or an unsupported withdrawal. Each classification changes the rights, risk, maturity, tax and sometimes the legality of the payment.

The shareholder relationship is not a supporting document. “Money paid in” and “money taken out” describe the movement; the contract, resolution and economic substance determine its accounting nature.

First: map the related parties

Identification is not limited to direct shareholders. Map direct and indirect control, joint control, significant influence, directors and key management, relevant relatives and entities they control, as well as linked enterprises under tax legislation. Accounting, tax and company-law definitions do not perfectly overlap; apply the definition in the instrument requiring each test.

Capital or loan?

Interest between related parties requires three distinct analyses: accrual accounting over the financing term; arm’s-length pricing and tax documentation; and applicable deductibility limits for borrowing costs. Non-payment does not eliminate accrued expense or income when a contractual obligation exists. Recording interest does not prove that its rate is at arm’s length.

Money withdrawn by a shareholder

Distribution and payment test in 2026

A shareholders’ resolution cannot make an amount distributable when the law prohibits distribution. Before declaring a dividend, verify determined and approved profit, lawful loss coverage, reserves, net assets, contractual restrictions and interim-dividend status. The dividend tax is 16% for distributions governed by the regime effective from 1 January 2026, but the rate is only one of the tests.

For newly incorporated SRLs, minimum share capital is 500 lei. For limited-liability companies whose turnover exceeds 400.000 lei, the minimum is 5.000 lei, subject to the application and transition rules in Law No. 239/2025. Treat these as volatile statutory parameters: verify them before each decision rather than copying the prior-year file.

Confirmation and intra-group reconciliation

Confirm related-party balances bilaterally by currency, document and nature. Break differences down into in-transit transactions, exchange rates, one-sided invoices, set-offs, interest and classification differences. Agreement on the amount does not prove correct classification: one entity may show a loan while the other shows an advance, with totals agreeing perfectly.

Part XII · Professional judgment · cross-cutting layer

Chapter 55

Working papers and professional review

A trial balance can be correct even when nobody can still explain why. Working papers turn the result from a collection of files into a chain of assertions, evidence, procedures and conclusions that another professional can reperform.

A check proves that someone checked. A worksheet proves what was checked, against what criteria, with what evidence, by whom, when and with what conclusion.

Assertions: what a figure actually states

From risk to procedure

working materiality and uncorrected errors

Financial-statement materiality defines what could influence a user. Performance materiality is lower, reducing the risk that undetected and uncorrected misstatements together exceed the final threshold. Also set a clearly trivial level, but never use it to dismiss qualitative errors, fraud or related-party transactions.

Evidence: sufficient and appropriate

Sufficiency concerns quantity; appropriateness concerns relevance and reliability. One hundred copies of internal invoices do not compensate for missing evidence that a service was performed. Evidence obtained directly from an independent source is generally more reliable than an oral explanation; a controlled original is stronger than a file of unknown provenance; documentary evidence is easier to verify than memory. No hierarchy is absolute: the procedure must address the assertion.

Sampling without self-deception

Selection does not begin with “the first ten invoices”. Define the complete population and the purpose of the test. Test all individually material or specifically risky items. Select the remainder randomly, systematically or by risk, but limit the conclusion to what the method supports. A selection aimed only at exceptions may find problems, but it cannot support a statistical conclusion about the entire population.

Review: the preparer cannot see every assumption they made

The reviewer assesses the objective, source, link to the trial balance, logic of the procedure, conclusion and propagation of adjustments. The reviewer need not mechanically repeat every step, but examines judgmental areas, estimates, exceptions, unusual transactions and changes from the prior year in greater depth. A review note remains open until the response is evidenced and its effect followed through; “discussed” is not a resolution.