What a trial balance is, and what it does not check

A trial balance is a list of every account in the books with its balance, split into two columns, debits on one side and credits on the other, with both columns totalled. If the two totals are equal, the trial balance balances.

That is the entire document. It is not a financial statement, it is not shown to anyone outside the business, and it is not the accounts. It is a working paper produced at the end of a period as a checkpoint before anything else is prepared, and its purpose is narrow and specific.

It exists because of how double-entry works. Every transaction is entered with equal debits and credits, so the sum of all debit balances across all accounts must equal the sum of all credit balances. That is arithmetic rather than an achievement: given correctly formed entries, equality is guaranteed. Which means the trial balance is not testing whether the business is doing well or whether the books are right. It is testing whether the entries were correctly formed.

It is also the bridge to the statements. The income and expense balances on it become the profit-and-loss statement; the asset, liability and equity balances become the balance sheet. So it is the point at which the raw ledger becomes something readable, and it is checked first precisely because everything downstream depends on it.

The accounts are conventionally listed in a standard order: assets, then liabilities, then equity, then income, then expenses. Each carries its balance in the column matching its natural side. Assets and expenses normally carry debit balances. Liabilities, equity and income normally carry credit balances.

That word normally is where the useful reading begins. An account sitting on the wrong side is not automatically an error, and it is always a question worth asking. A bank account with a credit balance means the account is overdrawn, which may be exactly right or may mean payments were recorded that never cleared. A supplier account with a debit balance means you have paid more than you owe, which happens with advances and also happens when a payment was recorded against the wrong supplier. An expense account with a credit balance usually means a refund or reversal exceeded the original, which is occasionally legitimate and often a duplicated correction.

So the first pass over a trial balance is not the totals at all. It is scanning for accounts on the unexpected side, and for balances that are implausibly large or small relative to the period. An owner who knows their business can spot in thirty seconds that a figure is wrong by an order of magnitude, which is a check no arithmetic performs. The totals are the last thing to look at, because they are the least informative part.

A balanced trial balance establishes a small number of things, and they are worth stating precisely so that nothing more is inferred from them.

It proves that every entry recorded has equal debits and credits. That rules out one-sided entries, entries where the two amounts differ, and entries posted with an amount omitted.

It proves that the arithmetic of extracting and totalling balances is correct, or at least that any errors in it happen to cancel, which is unlikely enough to ignore.

And it proves that no account has been dropped from the extraction, since a missing account with a non-zero balance would break the equality.

Those are real controls and they catch a genuine and common class of mistake, particularly in manual books where entries are written by hand and totals are added by hand. In software the first two are largely guaranteed by construction, because a system that refuses to save an unbalanced entry cannot produce an unbalanced trial balance. That does not make the document pointless in a computerised environment, but it does change what it is for: the value shifts from the totals, which will always agree, to the scan of the balances, which is where a human can notice something implausible.

This is the part that matters, because a balanced trial balance feels like verified books and is not. Four distinct error types survive it completely, and they are traditionally named because they are common enough to have earned names.

An error of principle: the entry is balanced and posted to a fundamentally wrong type of account. A machine purchase debited to repairs expense instead of to fixed assets balances perfectly and misstates both profit and assets, and it will keep misstating them for years because depreciation was never set up.

An error of commission: balanced, right type of account, wrong account. A payment credited to the wrong supplier, or an expense in the wrong head. Totals are unaffected.

An error of complete omission: the transaction was never entered at all. There is nothing to be unbalanced. This is the one that matters most, because it is how unrecorded cash sales remain invisible: the trial balance balances, the bank reconciles, the invoice sequence has no gap, and the revenue is simply absent.

An error of original amount, or a compensating error: a sale entered as five thousand when it was fifteen thousand is balanced and wrong on both sides. Compensating errors are two separate mistakes whose effects on the totals happen to cancel, which is rarer and harder to find precisely because the totals give no hint.

Add reversed entries, where the correct two accounts are used in the wrong order, and duplicated entries, where a whole balanced transaction is entered twice. Both balance.

An out-of-balance trial balance in a manual system is a defined problem with a defined search, and the classical tricks are worth knowing because they narrow it quickly.

First compute the difference. If it is divisible by nine, suspect a transposition: two digits swapped, such as 540 entered as 450, always produces a difference divisible by nine. If the difference is exactly twice a transaction in the books, suspect that an entry was posted to the wrong side, since moving an amount from one column to the other shifts the difference by twice the amount. If it equals a single transaction exactly, suspect a one-sided posting, and look for that amount. If it is a round figure, suspect an addition error in the totals.

Then work backwards through the chain: re-add the columns, check that each ledger balance was copied across correctly, check that each account balance was calculated correctly, and only then re-examine individual entries, which is the most expensive step.

What should not happen is the shortcut. Posting the difference to a suspense account to force agreement makes the document balance and destroys the information it existed to produce, and the fiction then sits in the opening position of the next period where it will never resolve. If a difference genuinely cannot be located, it should remain visible and stated, with a note of what was investigated, rather than absorbed. A suspense balance carried forward is a permanent admission with the appearance of a clean set of books.

The trial balance is one checkpoint among several, and understanding what each one covers is how you know what is still untested.

It tests the internal arithmetic of the ledger. Bank reconciliation tests the books against an independent record, which is a stronger test because the bank statement is not produced by your business. A physical stock count tests the stock figure against reality. Comparing a purchase register against portal data tests your records against your suppliers'. Reading the accounts with knowledge of the business tests plausibility. Each catches things the others cannot, and the trial balance is the weakest of them, because it is the only one comparing the books solely against themselves.

That framing gives the honest limit for any accounting tool as well. Software can guarantee balanced entries, produce a trial balance instantly, flag an account carrying a balance on the unexpected side, and refuse to close a period with an unexplained difference concealed in a suspense head. Those are genuine and they remove the arithmetic entirely. What no system can do is know that a machine was posted to repairs, that a payment went to the wrong supplier, or that a cash sale was never entered, because in every one of those cases the entry it received was perfectly well formed. Detecting them needs comparison against something outside the books, or a person who knows what the numbers ought to look like.