A bookkeeping cleanup is a sequence, not a project

Every firm has a cleanup that went sideways. Quoted at twenty hours, closed at sixty, and the client heard "it's almost done" four times. The bookkeeper wasn't slow. The work was done in the wrong order, so every step undid part of the step before it.

We run an accounting firm, and cleanups are about a third of our new-client work. The ones that stay on budget all run the same six steps, in the same sequence, and the sequence is the whole method. Here it is, with the pricing and the timing that make it a product instead of a favor.

Last week: four full-year cleanups before the 15th

A proof point before the method. The week before the September 15 extension deadline, our firm took four restaurant clients with the same owner, all of 2025 untouched, and closed all four years of books in Woven between Wednesday and Monday, in under ten hours of staff time across all four clients combined. One of them alone was about 2,500 ledger lines, including nearly 700 handwritten checks that had to be read from check images and matched to the payroll roster to tell a tip payout from a vendor payment. Every coded line ended up with a vendor on it, every loan tied to the lender, and the owner's questions went out as one list through the client portal instead of forty emails.

That pace is not because we worked harder. It's because the sequence below was run by software that reads the bank's own statements line by line, learns a categorization rule the first time and applies it to the next 400 transactions, and matches check images against a payee list so a human only looks at the exceptions. Four full-year cleanups in under ten hours used to be a month of staff time. Here's the order that makes it possible.

Step 0: the diagnostic, priced separately

Before anyone touches a transaction, someone spends two to three hours producing a one-page diagnostic: the last date the balance sheet was provably right, every bank and card account and its last reconciled month, the Uncategorized and Ask My Accountant totals, the Undeposited Funds balance and its age, loan balances versus the lender statements, AR and AP aging with anything over 90 days flagged, and a count of obvious duplicates.

Charge for it. A diagnostic priced at a few hundred dollars does three things: it tells you what the cleanup will actually cost before you quote it, it tells the client the size of the problem in numbers they can see, and it filters out the prospect who wants a free assessment and a second opinion on their own DIY books. Every fixed-fee cleanup we've regretted was one where we skipped the diagnostic and quoted from a phone call.

Step 1: freeze the starting point

Pick the last date the balance sheet was right. Usually that's the last filed tax return, sometimes the last clean reconciliation. Everything before that date is locked. You are not fixing history; you are fixing forward from a known-good point. If the client wants earlier years touched, that's a separate engagement with its own diagnostic, because the tax preparer built returns on those numbers and changing them has consequences beyond your fee.

Write the freeze date into the engagement letter. It's the line that stops scope from drifting back into 2022.

Step 2: bank and credit card, nothing else

Reconcile every account to the statement, month by month, from the freeze date to today. Do not categorize anything yet. Do not fix names. Just get every transaction present and every month tied. Missing deposits and missing expenses surface here, because the statement is the truth and the books have to match it. When a month won't tie, the uncleared list against the statement tells you exactly what was never entered.

This step is where firms lose the most time when they skip it, because categorizing a transaction that later turns out to be a duplicate is work you'll do twice. It's also where a tool matters: pulling the statement and comparing it line by line against the ledger is mechanical, and it should be done by software that reads the statement, not by a person reading two PDFs side by side.

Step 3: kill the duplicates

Now that every account ties to a statement, the duplicates are visible: the check that was also entered as an expense, the deposit that was recorded from the invoice and again from the bank feed, the vendor bill paid twice in the books but once in reality. Remove them, or if the year is closed, reverse them in the current period with a note. Also merge the vendor list at this point. Eleven versions of the same utility company are a categorization problem waiting to happen in step 4, and a duplicate-vendor finder does this in minutes.

Step 4: categorize, from the chart down

Only now do you touch the P&L. Start by fixing the chart of accounts, not the transactions: collapse the 140-account chart the last bookkeeper built into one the owner can read, and map the old accounts to the new. Then categorize by rule, not by transaction. Sort Uncategorized by payee; the same payee almost always gets the same account, and a rule that codes it once codes the next two years. What's left after rules is the genuine judgment work, and it's usually a tenth of what it looked like.

Balance sheet accounts get the same treatment, and this is the part most cleanups skip: tie every balance sheet account to something outside the books. Loans to the lender statement, with the interest split out. Payroll liabilities to the payroll reports. Sales tax to the filed returns. Undeposited Funds to zero, because cash received and never deposited went somewhere, and it's an owner draw, an expense paid in cash, or a duplicate sale. Intercompany to the other entity's books. An account you can't tie to anything is a finding, and it goes in the memo.

Step 5: review the P&L for what it says

With the balance sheet tied, read the P&L month by month and look for the things that are wrong in a way only a human sees: a month with no rent expense, revenue that doubled in one month and vanished the next, an owner's personal spending sitting in Office Expense, cost of goods with no matching revenue. Each one is either a fix or a question for the client. Batch the questions into one list, once. Clients answer one list; they don't answer forty emails.

Step 6: the memo and the handoff

The deliverable isn't a clean QuickBooks file. It's a two-page memo: what was wrong, what was fixed, what couldn't be fixed and why, what the client's tax preparer needs to know (the prior-period adjustments, the loans that were misstated, the cash that was drawn), and what has to change so it doesn't happen again. The last section is where the recurring engagement gets sold, because the reason the books were a mess is almost never that the last bookkeeper was bad. It's that nobody owned the monthly close.

Pricing that holds

Fixed fee, priced from the diagnostic, quoted per month of catch-up with a floor. The per-month rate scales with the count of accounts and transaction volume, both of which the diagnostic gives you. Two clauses protect the fee: a change-order rate for anything before the freeze date or any account not in the diagnostic, and a clause that the fee assumes the client answers the batched question list within ten business days. Hourly cleanups punish the efficient firm and make the client anxious about every email; fixed fees from a real diagnostic are the version both sides like.

Why Q4 is when to sell them

A business owner who's been ignoring the books all year starts feeling the tax return in October. That's when the "can you just clean this up before year end" calls come in, and a firm that has a productized cleanup (diagnostic, fixed fee, six steps, memo) closes those calls at a real price. The firm that quotes hourly from a phone call does the same work for less money and more stress. And every cleanup done in Q4 becomes a monthly-close client in January, because the memo's last section already made the case.

Where the platform comes in

In Woven the cleanup steps map to the tools we built for them: statement import that reads the bank's own PDF and reconciles it line by line against the ledger, a duplicate finder and vendor merge for step 3, categorization rules that learn from the firm's choices for step 4, a loan sub-ledger that ties every note to the lender statement, and a per-client close board so the firm can see which of its cleanup clients has become a client whose books close on time. The point of a cleanup is that there isn't another one.