The number everything sits on
An opening balance is unusual among accounting figures: it is the only one that arrives without workings. Every other balance is the result of something you can inspect — transactions, adjustments, a prior reconciliation. The opening balance is simply asserted, usually by someone who is no longer available to ask.
That would not matter if it were reliably right. In practice it is the figure most likely to contain something old and unresolved: a plug from a reconciliation that was never finished, an item in transit that never cleared, a difference someone rounded away when the accounts had to be filed. And because every later balance is built on it, an error there propagates silently through everything you do afterwards.
The fix is cheap and rarely done. The bank states an opening balance on every statement, and that assertion comes from a third party with no interest in your accounts being tidy. Anchoring to it takes an afternoon and converts a number you inherited into one you can defend.
Why it goes unchecked
Three reasons, and they are all understandable. Checking an opening balance feels like re-doing someone else's work, which is nobody's favourite task and looks like duplication to whoever is paying. It is also usually the first thing on a list of many, arriving exactly when there is most to set up and least appetite for archaeology.
The third reason is the real one: there is no moment that forces it. Nothing breaks when an opening balance is wrong. The books still balance, the reports still run, and the error simply travels forward as a constant offset. It surfaces years later, during due diligence or an audit, when explaining it costs far more than checking it would have.
The asymmetry is stark. An afternoon at the start, against an investigation into a decade-old figure at the worst possible time — and by then the statements may be harder to obtain and the people involved gone.
Why the bank is the right anchor
In any reconciliation, the useful record is the one produced by someone with no stake in your conclusion. A handover spreadsheet reflects what the previous preparer believed, and their belief may have been reasonable and still wrong. A bank statement records what the bank did with your money, which is a different kind of claim entirely.
It is also complete in a way internal records often are not. A statement lists every movement through the account, including the ones nobody told the bookkeeper about — the direct debit set up by a colleague, the card payment from a forgotten account, the standing order that outlived its purpose. Those are precisely the items an inherited balance tends to be wrong about.
And it is verifiable later. A figure traced to a statement page can be re-checked by anyone at any time; a figure traced to a spreadsheet can only be trusted. When the question eventually gets asked — and on an opening balance it usually does — the difference between those two is the whole answer.
How to establish it
1 — Find the statement covering your date
The statement whose period contains your start date, or the nearest one that begins before it.
2 — Convert it, plus a few months forward
One to three months is usually enough to explain the position and confirm items in transit cleared.
3 — Take the stated opening balance
Use the bank's own figure as the anchor rather than the handover number, and note where they differ.
4 — Chain the proofs forward
Each statement's closing balance must equal the next one's opening balance. A break means a missing period.
The chaining step is what turns individual statements into a proof. Each one proves its own arithmetic; joining them proves there are no gaps. Together they establish that the position on your start date follows from documented movements rather than from an assertion.
When your date falls mid-statement
It usually does. Accounting periods end on month ends and statements run on bank cycles, so the two rarely coincide. The method is straightforward: take the statement that begins before your date, and work forward within it.
Start from that statement's stated opening balance, sum the movements up to and including your date, and the result is the bank's position on the day you need. Because the statement as a whole has already proved its own arithmetic, this partial sum inherits that proof — you are cutting a verified sequence rather than trusting an isolated figure.
One detail worth attention: use the transaction date the bank uses for its running balance, not the value date, when they differ. Mixing the two produces a figure that is a day out and disagrees with the statement it came from, which is exactly the kind of small discrepancy that consumes an afternoon later.
Items in transit at the opening date
The bank balance and the book balance will differ, and that difference is legitimate. Payments issued but not yet cleared, receipts recorded but not yet credited — these belong in the books before they appear at the bank, so the two records are correctly out of step.
List them individually rather than as a total. A single reconciling figure is unfalsifiable: it agrees by construction and tells you nothing. An itemised list is testable, because each item should appear in the bank data within the following weeks, and any that does not is a real finding.
The items that never clear are the most useful thing this exercise produces. A payment recorded years ago that never left the account usually means a cheque was never presented, a payment was cancelled without reversing the entry, or something was recorded twice. Each of those is a genuine correction to the opening position, and none would surface any other way.
Acquired entities
An acquisition is the highest-stakes version of this problem. You are taking on a set of balances prepared by people whose incentives during the sale were not identical to yours, and those balances become the base for everything you report afterwards.
Convert statements from a few months before completion as well as after. The pre-completion movements are what explain the balances you inherited — a receivable that was collected just before completion, a payment run that was delayed, a balance that moved unusually in the final weeks. None of that is visible in the closing position alone.
Do it early, while the seller is still cooperative. Requests for historical statements are answered quickly in the weeks after completion and slowly a year later, and the questions that make you want them tend to arrive on the slower timeline. For groups adding entities regularly, our multi-entity reconciliation page covers running this alongside a normal close.
Migrating between accounting systems
A system migration is a deliberate opening-balance event, and it is the best opportunity most businesses get to check the position they are carrying. The balances are being re-entered anyway; verifying them at that moment costs almost nothing extra.
It is also the moment when an old error is cheapest to correct. Once balances are loaded into the new system they acquire a second layer of authority — they came from the migration, and the migration was signed off — which makes questioning them later noticeably harder even when the underlying figure never changed.
A practical sequence: prove the bank position from statements first, then load, then reconcile the loaded balance back to the proved one. If they disagree, you know before anything is built on top. The alternative — load first, reconcile if time permits — reliably means the reconciliation does not happen.
Several accounts, several anchors
Most businesses have more than one account, and the opening cash position should be the sum of individually proved accounts rather than one combined figure. A single total cannot be decomposed later, and decomposition is exactly what any subsequent question requires.
Converting all accounts together keeps each one's provenance attached, so the combined position is auditable at any level. It also catches the account nobody mentioned, which is more common than it should be — a currency account opened for one supplier, a deposit account holding a rent bond, a card account that was never thought of as a bank account.
| Account type | Why it gets missed | What it can hide |
|---|---|---|
| Currency account | Opened for one supplier, rarely used | Unrecorded FX movements and charges |
| Deposit account | Only receives sweeps, feels inert | Interest never recorded in the books |
| Card account | Not thought of as a bank account | A month of expenses outside the ledger |
| Dormant account | Left from an old entity or venture | Standing orders still running |
| Merchant / payment account | Treated as a feed, not an account | Fees netted before settlement |
When the anchor disagrees with what you were given
This is the productive outcome, not the awkward one. A difference between the bank's opening balance and the inherited figure is information, and it usually has one of four causes.
| Cause | How to recognise it | What to do |
|---|---|---|
| Legitimate items in transit | Individually identifiable, clear soon after | List them; confirm each clears |
| An incomplete reconciliation | Difference is one or two whole transactions | Find the transactions; post them |
| A plug | Round or oddly specific, no components | Investigate the period it appeared |
| A genuinely missing account | Difference matches another account's balance | Add the account and re-prove |
Whichever it is, resolve it before building on top. An opening difference carried forward becomes a permanent constant in every subsequent reconciliation, and once it has survived two period ends it acquires a kind of legitimacy that makes anyone reluctant to touch it.
Reconstructing when there is no ledger at all
A harder version of the same problem turns up more often than it should: there is no opening balance to check, because there are no usable records. A business that ran on a shoebox, a company whose bookkeeper left mid-year, a set of accounts that were never completed. What exists is a bank account and a stack of statements.
That is more than enough to start. Bank data is a complete record of everything that moved through the account, which means a period can be rebuilt from it even when nothing else survives. What it cannot supply is classification — the statement knows an amount left the account, not whether it was a cost, a drawing or a loan repayment. Those come from invoices, contracts and conversations.
The efficient sequence is to build the spine first and classify afterwards. Convert every statement for the period into one table, prove each one against its own closing balance, and confirm the chain has no gaps. At that point you have an indisputable record of what happened, in order, with running totals — which is a far better starting position than most reconstructions ever reach.
Classification then proceeds by frequency rather than chronology. Sort by description, and the same counterparties appear in blocks: rent twelve times, the same supplier forty times, payroll monthly. Classifying a block at a time handles most of the value in a fraction of the effort, and what remains is a genuinely short list of one-off items that each need a document or an answer.
The realistic expectation matters here. A reconstruction from bank data produces a defensible cash record and a reasonable expense analysis. It does not produce accruals, stock, work in progress or anything else that never touched the bank, and it cannot tell you what an unlabelled transfer was for. Being explicit about that boundary with whoever asked for the reconstruction is what stops it being mistaken for a complete set of accounts.
What a lender or a buyer asks about the opening position
Opening balances are ignored right up until someone external has money at risk, and then they become one of the first things examined. Both lenders and acquirers ask a version of the same question: is the starting point real, or is it an assumption everything since has been built on.
In due diligence, the specific concerns are predictable. Whether the cash position at the start of the reviewed period agrees to the bank. Whether every account is included, or only the ones the seller thought to mention. Whether any balancing figure exists that nobody can decompose. Whether the movements immediately before the period start explain the position or complicate it.
A business that can answer those in an afternoon is treated very differently from one that cannot, and the difference is not really about accuracy. It is about what the answers imply about everything not yet examined. A group whose opening position traces cleanly to bank statements signals a finance function that keeps evidence; one that cannot signals the opposite, and the review widens accordingly.
The practical implication is timing. Doing this work while under diligence is expensive, rushed and visible. Doing it in an ordinary month, before anyone asks, costs an afternoon and produces a document that sits in a folder until it is needed. Businesses that expect to raise, sell or refinance within a few years are the ones for whom the calculation is most obviously favourable — and they are also the ones most likely to postpone it.
Documenting the position
Write it down at the time, in one page. The anchor figure and which statement it came from; the items in transit, listed; any difference from the inherited number and what it turned out to be; the date you did the work and who did it.
That page is worth disproportionately more than the effort it takes. It converts every future question about the starting position from an investigation into a lookup, and it is the single most useful document to hand an auditor, an acquirer or a successor. Its absence is why opening balances become mysteries in the first place.
Keep the converted statement data alongside it, with the source column intact. The summary explains the conclusion; the data lets someone verify it without going back to the bank.
Date the note and name the person who did the work. Both sound like formalities and both get used: the date establishes what was known when, which matters if a later period is questioned, and the name establishes who to ask before anyone starts reconstructing from scratch. A note without either tends to be treated as unverified regardless of how good it is.
One habit is worth adopting alongside it. Repeat the exercise at each year end rather than only when taking on the books — not the full reconstruction, just the check that the brought-forward cash position still agrees to the bank. It takes minutes once the process is familiar, and it means an error can never travel more than one year before something catches it. Groups that do this find that opening balances stop being a category of problem entirely.
Store it where the accounts live, not where the project lived. Opening-balance work is usually done during an onboarding, a migration or a deal, and those all have their own folders that nobody opens again. Two years later the person looking for this note will be looking in the year-end file for the period concerned, which is where it should have been put in the first place.
A last point about what to write in it, because there is a temptation to record only the conclusion. Record what you looked at and did not find as well: which accounts were confirmed to exist, which periods were checked, what was explicitly excluded from scope. A note saying the position was proved is useful. A note saying which four accounts were proved, over which three months, with the fifth account excluded because it was closed in March, is the one that survives a question — because it answers the question a reviewer actually asks, which is not whether you were right but how far you looked.
None of this needs a template or a system. A page of plain text, dated and named, stored beside the converted statement data, does everything described here. The reason opening balances become mysteries is almost never that the work was too hard — it is that nobody wrote down that it had been done.
Five mistakes
Accepting the handover figure.It may be right, and you have no way of knowing without the bank's version. Checking costs an afternoon; not checking costs an unknown amount at an unknown time.
Plugging the difference. A single balancing figure agrees by construction and explains nothing. It also hides the finding that made the exercise worthwhile.
Only converting forward. The movements before your start date are what explain the position at it. Starting the data on day one leaves the anchor unexplained.
Forgetting an account. A proved position for three of four accounts is not a proved position. List the accounts before starting, from bank correspondence rather than from memory.
Not confirming items cleared. An in-transit item identified and never followed up is indistinguishable from an error you accepted. The check a few weeks later is the part that makes the list meaningful.
Who needs this
Accountants taking on a client
Anchor the first period to the bank rather than to a predecessor's file.
Businesses changing systems
Verify balances at the one moment they are being re-entered anyway.
Post-acquisition finance
Prove what you inherited while the seller is still responsive.
Anyone with an old difference
Establish where the number stopped making sense, and why.
For a group doing this across several companies at once, the same work runs in one pass with each entity keeping its own anchor — covered on the group finance page.
What this does not prove
It proves cash. It says nothing about receivables, payables, stock, accruals or any other balance — those need their own evidence, and a proved bank position should not be mistaken for a proved balance sheet.
It also does not validate the accounting treatment of what it finds. Knowing that 40,000 arrived on a particular day is a fact; whether it was revenue, a loan or a deposit is a judgement that has to come from the underlying documents.
And FlowParse is not an archive. The original PDF is deleted immediately after extraction, so the statements and their retention obligations stay wherever you keep them today. What you get is the working data, in a file you hold.
Anchor your first period
Convert one historical statement free — no registration — and see the opening balance, the movements and the proof.