There is a genre of work that happens in accounting offices every spring, and everyone in the trade knows it by feel even if they’ve never given it a name. A client’s year arrives in fragments: a bank feed that mostly matches, an invoicing system that sort of agrees with it, a folder of receipts, a spreadsheet titled reconcile_FINAL_v3, and a certain number of transactions whose explanation lives in someone’s memory of a phone call from March. The accountant then does what accountants have quietly done forever, which is archaeology. They dig, they date the layers, they reconstruct the civilisation.
They are good at it. That’s the strange part. The reconstruction usually comes out right, filings get made, nobody goes to jail. The problem isn’t competence. The problem is that a professional trained in judgment is spending a meaningful slice of their hours determining what happened, when what happened was never actually in doubt. Every one of those transactions was recorded, correctly, at the moment it was born — in the point of sale, in the invoicing tool, in the bank. The information existed. It just didn’t travel.
Here’s the mechanic, in almost every set of messy books we’ve seen, including our own before we fixed them: a transaction is born in one system, hand-typed into a second, and reconciled against a third. Each copy is made by a person, so each copy drifts a little. Multiply by twelve months and you get three slightly different versions of the year, and a professional billing professional rates to rule on which version is canon. The typing gets billed as expertise because it arrives bundled with expertise, and nobody ever unbundles the invoice to notice how much of it was data entry.
We ran a business this way once, and the part that stung wasn’t even the cost. It was the lag. When the books close three weeks after month-end, the number you finally get describes a company that no longer exists. You can’t reprice anything. You can’t catch the margin slide while it’s a slide instead of a cliff. A number that arrives in time is an instrument; a number that arrives late is a plaque commemorating where the money used to be.
The fix is unglamorous and has two halves. First the books get set up properly, once — a chart of accounts designed with your accountant, matched to how they actually file, because building automation on top of a miscategorised ledger just automates the mess. Then the systems where transactions are born get wired to feed the ledger themselves, so nothing is typed twice and the humans only see exceptions. That’s it. No dashboard rhetoric, no platform migration. Pipes.
What changes downstream is bigger than it sounds. Month-end stops being a project and becomes a checklist that mostly runs itself. The owner sees numbers they can act on while acting still helps. And the accountant — this is the part people don’t expect — usually becomes the fix’s biggest fan, because they get to spend their hours on the judgment they trained for instead of retyping a year. We’ve written before that the quiet months are the right time to do work like this, and books are the classic case: nobody rebuilds a ledger in their busy season, and nobody should try.
If any of this sounds like your file, the honest starting point isn’t a software pitch. It’s an afternoon inside your actual books, finding out where the transactions are born and where they go to get retyped. That’s the day we do free, and spring is a lot cheaper when it doesn’t begin with an excavation.