A quarterly check on your own expense data

Some errors don’t exist in any single record. They only appear in aggregate — a supplier split across five spellings, a category that quietly absorbed everything ambiguous, a month with suspiciously few entries. No amount of per-receipt review finds them, because every individual receipt is fine.

So once a quarter, look at the whole set instead. Six checks, an hour at most, and they find a different class of problem from anything a review queue surfaces.

1. The vendor list, alphabetically

Not sorted by size — alphabetically. Every vendor, with transaction count and total.

Alphabetical order puts variants of the same supplier next to each other, and they jump out: Northwest Tools, Northwest Tools Ltd, NWTS HOLD 4412. Each of those entries is individually correct, and together they mean your vendor reports understate your largest supplier and your categorisation rules are only catching a third of their receipts.

Merge the variants, pick a canonical name, and note the aliases so future arrivals map to it.

2. Transaction count per month

A simple series: how many expenses per month, for the last year or two.

You’re looking for dips. A month with noticeably fewer entries than its neighbours usually means missing data — a period where capture lapsed, a card that stopped syncing, a batch that failed to import. Totals hide this, because a month with 20% fewer receipts still looks like a plausible total. Counts don’t.

Spikes are worth a glance too; they’re often duplicates from a double import.

3. The catch-all category

Every chart of accounts has one — Miscellaneous, Other, General, Uncategorised. Read it, in full.

It accumulates everything ambiguous, and it grows quietly. What’s in there is usually a mix of genuinely one-off items and two or three recurring things that never got a proper home. The recurring ones deserve their own category and a rule; once written, they stop landing here.

A catch-all that’s growing as a share of total spend is a signal your categories no longer match what you actually buy.

4. Unmatched transactions and unmatched receipts

Both lists, in full.

Transactions with no receipt are expenses you can’t support with a document. Some are recoverable at a quarter’s distance; fewer than at a week’s, which is the argument for doing this reconciliation weekly too. Whatever remains needs a note recording what it was, while there’s any chance of knowing.

Receipts with no transaction are usually cash purchases, a personal card used by mistake, or a duplicate. The personal-card ones are reimbursements someone is owed, and they’re easy to lose entirely.

5. Category totals against last quarter

Put this quarter’s category totals next to last quarter’s and look for the movements you can’t explain.

A category that doubled, or halved, is either a real change in the business — in which case you know why — or a categorisation drift. Drift usually means a vendor started posting somewhere new: a rule changed, a vendor’s descriptor changed and stopped matching a rule, or a first-time receipt established a wrong pattern that then repeated.

This check is the one most likely to find a systematic error that’s been running for weeks.

6. Retrieval, tested rather than assumed

Pick three expenses you remember and find their documents. Not by browsing to them — by searching, the way you’d have to in three years with only a vague memory of the vendor and the year.

If you can’t find them in a minute each, your storage is working and your retrieval isn’t, and those are different problems. Usually the fix is naming and consistent vendor names rather than anything about the storage itself.

While you’re there, open one of the images with something other than the app that stores it. Confirms portability at the same time.

What to do with what you find

Two rules make the hour worthwhile rather than merely informative.

Fix upstream. A miscategorised vendor gets a rule, not a corrected entry. A rule fixes every future instance; a corrected entry fixes one and you’ll be back next quarter.

Write down what you changed. Which variants you merged, which rules you added, which categories you split. Next quarter you’ll find something similar and the note tells you whether it’s a recurrence or something new.

Why quarterly

Frequent enough that a systematic error gets caught within a few months of starting, rather than at year end when it’s a year of data. Infrequent enough to be a real look rather than a ritual.

Weekly reconciliation and a working review queue handle the per-item errors. This is the pass that catches the ones that only exist when you stand back — and those are the errors that make reports wrong while every entry in them is right.