How to catch up on bookkeeping

Step 1 — Scope the backlog before touching anything

An hour of scoping saves days of drifting. The goal is a written answer to four questions: which periods are unrecorded, which accounts exist, which filings are outstanding, and what documents are actually available.

• List every bank account, credit card, loan, payment processor and cash float — including accounts closed during the period.
• For each, note the first and last month you are missing.
• List the returns and filings that are overdue or coming due, and their deadlines.
• Note which document sources exist: online banking, email, a capture tool, a drawer.
• Identify anything unusual — a loan taken out, an asset bought, a grant received, a change of legal structure.

Forgotten accounts are the classic scoping failure: the card used for one subscription, the account opened for a project, the processor that paid out twice. Each is a hole in the record that surfaces at the worst moment. Check bank correspondence and the payments already visible in other accounts to find them.

Step 2 — Gather every statement, for every month

This is a collection task, and doing it completely before starting any data work is what keeps the project moving. Download every statement for every account for every month in scope, into one folder, named consistently — account, year, month.

Consistent naming is not fussiness. It is what lets you see at a glance that August is missing from one account, and it is what makes the converted rows traceable back to a document later. Ten minutes of naming discipline now removes an hour of confusion at reconciliation.

Prefer PDFs from online banking over anything else. If the bank offers CSV or OFX exports for the period, take those too — a native export needs no conversion at all, and using it where it exists is simply faster. Extraction is for what the bank will not give you in a data format, which for older periods is most things.

flowparse.io

When statements are missing

Some will be. Ask the bank first — most provide historical statements through online banking or on request, including for closed accounts, sometimes for a fee. Pay the fee. Reconstructing a month without its statement costs far more than any bank charges for it.

If a statement is genuinely unobtainable, work from what exists: the closing balance of the month before and the opening balance of the month after bracket the period, card portals often show transaction history beyond the statements you kept, and payment processors retain their own records independently.

Whatever you do, document it. A note saying what was requested, when, what was received and what was used instead turns an unexplainable gap into an evidenced one — which matters a great deal if the year is later reviewed by an accountant, a lender or an auditor. That expectation is covered further in the audit preparation guide.

Step 3 — Convert the year in bulk, not one file at a time

Thirty-six PDFs converted individually and pasted into a spreadsheet is where catch-up jobs acquire their errors: mismatched columns, a file processed twice, one silently skipped.

Convert them as a batch and consolidate in the same operation. Smart Merge takes up to a hundred statements from different banks and produces one workbook with unified columns, duplicate detection across overlapping periods and a source-file reference on every row — so a surprising figure in March can be traced to the exact statement that produced it without re-deriving anything.

Then check completeness before going any further, because everything downstream inherits this data. Each statement's transactions must reconcile to its own opening and closing balance — the balance checkruns per account automatically and names the rows where the arithmetic breaks. Then check the sequence: each month's closing balance should be the next month's opening balance, with no gaps. That second check is what catches a missing month, and it takes five minutes.

flowparse.io

Step 4 — Build the transaction spine

You now have every transaction of the year in one place. Before categorising, get the structure right, because fixing it later means redoing the categorisation.

Two of those rows cause most of the damage in catch-up work. A running balance exported as the transaction amount puts a plausible wrong number on every row, and a date read in the wrong convention moves transactions between months — or between tax years, if it happens near a period end.

Overlaps, duplicates and transfers

Three things reliably corrupt a consolidated year, and all three are worth checking deliberately before categorising.

Overlapping statements. A statement covering January to March next to one covering March to June double-counts March. Duplicate detection handles this, but check the period coverage yourself as well — the check is a minute and the error is a year.

Genuine duplicates that are not duplicates. Two identical coffee purchases on the same day at the same merchant are two real transactions. Any de-duplication that removes them is wrong, which is why matching has to consider more than amount and date.

Transfers between own accounts. Money moving from the current account to the savings account is neither income nor expense, and if both accounts are in the dataset it appears twice with opposite signs. Identify and pair these early — misclassified transfers are the most common reason a catch-up year overstates both revenue and costs.

Step 5 — Categorise in bulk, not chronologically

Categorising a year one transaction at a time in date order is the slowest possible method. Sort by payee instead, and the work collapses: forty transactions from the same supplier get one decision, not forty.

Work down from the largest groups. The top twenty payees usually account for most of the volume, and once they are categorised the long tail is a fraction of what it looked like. Leave anything genuinely unidentifiable in a queue rather than guessing, and go through that queue once with the business owner — memory works far better when someone is looking at a list of specific dates and amounts.

Automated categorisation helps here and needs supervision, especially in catch-up work where the tool has no history to learn from. The mechanics of doing this well are in categorising bank transactions, and the wider automate-versus-review split in bookkeeping with AI.

flowparse.io

Step 6 — Attach the documents you do have

Now the receipts and invoices become useful, because there is something to attach them to. Work from the transaction list outwards: for each material payment, find the document that explains it.

Extract in bulk rather than typing — receipts and invoices convert into a schedule with merchant, date, total and tax that can be sorted and matched against the bank data by amount and date. A batch of two hundred receipts becomes a spreadsheet in one pass instead of an afternoon of retyping.

Prioritise by materiality. The large payments need their evidence; the many small ones matter less individually and are usually explained by pattern — a recurring subscription does not need twelve separate hunts once the first is identified.

flowparse.io

Missing receipts — what to do honestly

Some documents are simply gone. This is normal in catch-up work and it is worth handling explicitly rather than pretending.

First, reconstruct where you can: supplier statements list the invoices they issued you; email archives hold order confirmations; online accounts with major suppliers keep purchase history; card portals show merchant detail beyond the statement line. A surprising proportion of "lost" receipts can be retrieved this way in an hour.

Then record what remains undocumented, with the payee, date, amount and what you know. Whether an expense is deductible without a receipt depends on your jurisdiction, the type of cost and sometimes the amount — that is a question for your accountant, and neither this guide nor any software should answer it for you. FlowParse does not decide deductibility, does not compute tax and is not an archive; it structures what the documents say.

Step 7 — Reconcile month by month, forwards

Do not reconcile the year in one attempt. Reconcile January, prove it, then February. A difference found in a single month has a four-week search space; the same difference found across a year has fifty-two.

For each month and each account: the opening balance agrees with the prior month's close, every transaction is present and categorised, and the closing balance matches the statement exactly. When it does, move on. When it does not, find it — do not post a balancing journal to make it agree, because that journal is an unexplained figure that will outlive the project.

The usual causes, in order of frequency: a missing statement period, a transaction entered twice, a transfer treated as income or expense, a converted statement that lost rows, and an amount with a transposed digit. The method for working through them is in the reconciliation guide.

flowparse.io

Step 8 — Year-end adjustments

With the transactions complete and reconciled, the accounting judgement layer goes on top. In catch-up work this is usually where an accountant becomes necessary if one is not already involved.

The owner-transactions row is the one that most often needs a conversation rather than a rule. In a year nobody wrote up, personal and business spending have usually blurred, and untangling it is judgement — precisely the part no automation addresses.

Step 9 — File what is owed, and say what happened

Once the year is reconciled and adjusted, produce the accounts and file whatever is outstanding. If returns are late, penalties and interest may apply and are usually a matter for your accountant to handle with the relevant authority — the important thing is that filing late with correct figures beats filing on time with invented ones.

Keep the working papers with the filing: the consolidated transaction dataset, the reconciliation for each month and account, the list of undocumented items, and the note about any statement you could not obtain. If anyone reviews the year later, that pack is the difference between a short conversation and a reconstruction.

And keep the source documents themselves for the period your jurisdiction requires. An extraction tool is not an archive — FlowParse deletes originals immediately after processing — so retention is your responsibility and the statements are the primary record.

Several years behind? Do them oldest first

Each year's closing balances are the next year's opening balances, so working forwards means every year starts from a proven position. Working backwards means building each year on an assumption you have not verified, and an error in the earliest year propagates through everything already done.

Finish and reconcile one year completely before starting the next, even when deadlines make it tempting to jump to the most urgent. Half-finished years are how a two-year backlog becomes a three-year one.

The one exception is scoping: gather statements for all years at the start, because bank retention windows are finite and the statements you need for year one may become harder to obtain while you work on year three.

How long it actually takes

Notice what drives the numbers: rarely the transaction count. It is the number of accounts, how quickly missing documents arrive, and how much judgement each unexplained payment needs. That is why the conversion step — the part that looks like the bulk of the work — is usually the smallest slice of the time once it is done in bulk.

A worked example: one client, three accounts, twelve months

A consultancy has not written up its books for a full financial year. There is a current account, a company card and a payment processor, roughly 900 transactions, an envelope of receipts and a self-assessment deadline eight weeks away.

Morning one — scoping. An hour listing the accounts, the months missing on each, and the filings due. A fourth account surfaces: a savings account that received two transfers and paid one supplier. Without the scoping hour it would have been discovered during reconciliation, halfway through.

Afternoon one — gathering. Statements downloaded for all four accounts, twelve months each, named by account and month. Two card statements are missing from the download window and are requested from the bank; they arrive three days later, which is why gathering happens first rather than last.

Day two — conversion.Forty-six PDFs converted as one batch and consolidated into a single workbook with unified columns and a source-file reference per row. Each statement is checked against its own closing balance; one fails, is re-processed, and turns out to have had two rows lost behind a mid-page summary box. Then the sequence check: every month's closing balance equals the next month's opening balance, and one gap identifies a statement that was never downloaded.

Day three — categorising. Sorted by payee. The top eighteen payees cover about 70% of the transactions and take an hour. Transfers between the four accounts are identified and paired — nine of them, which would otherwise have inflated both income and costs by five figures. Around sixty transactions go into a queue for the client.

Day four — documents and the queue. The receipt envelope is scanned and converted into a schedule, then matched to payments by amount and date. Twenty-two payments have no document; eight are recovered from supplier portals and email, and the remaining fourteen are listed with payee, date and amount for the accountant to advise on. A thirty-minute call clears the categorisation queue while the client looks at specific dates rather than trying to remember a year.

Day five — reconciliation and handover. Twelve months reconciled in sequence, each agreeing to the statement. Two differences appear and both are found within minutes because the search space is one month: a duplicated entry in April and a card payment recorded in the wrong account in September. Working papers — the consolidated dataset, the monthly reconciliations, the undocumented list and the note about the missing statement — go to the accountant for year-end adjustments.

Five working days, most of it judgement and chasing rather than data entry. The conversion — the part that looks like the whole job — took a few hours, and the completeness check earned its place twice before lunch on day two.

Pricing a catch-up job (for bookkeepers)

Catch-up work is the most commonly under-priced service in bookkeeping, because the effort is front-loaded and the scope is invisible until the documents arrive.

Three habits fix it. Price it as a project separate from any ongoing monthly fee, so the front-loaded effort is paid for. Charge for discovery separately if you cannot see the records in advance — an hour to scope the backlog is legitimate work and prevents quoting blind. And define scope in writing by account and period, with a clear statement about what happens when documents are missing.

Then batch the mechanical work across clients rather than doing it per client as jobs arrive. Statement conversion for six catch-up clients in one sitting is far faster than six context switches, which is the practical argument for batch processing in a bookkeeping practice.

Stopping it happening again

Backlogs are almost never caused by laziness. They are caused by a process where the first step was slightly too hard, so it got postponed, and then postponing became the norm.

• Connect every available bank and card feed, so transactions arrive without anyone doing anything.
• Pick one route for receipts and invoices and use only that route.
• Put a recurring ninety-minute appointment in the calendar and treat it as a real meeting.
• Reconcile weekly rather than monthly — four small tasks beat one big one.
• Download statements for non-feed accounts the day they are issued.
• Ask for a monthly figure you care about, so someone notices when the books stop being produced.

The last one is underrated. A business that looks at a monthly number — margin, cash, a cost ratio — notices within weeks when bookkeeping stops. A business that only looks at year-end notices twelve months later, which is exactly how a year gets lost. The month-end checklist is the routine that replaces the crisis.

Common mistakes

Frequently asked questions

Keep reading