Ask any bookkeeper where their recoverable hours go and the answer is rarely the accounting. It is the reconstruction: working out what a payment was for, chasing a document that no longer exists, and emailing a client in November about a transaction from March that neither party remembers.
The frustrating part is that the information was free at the time. A client who spent eight seconds adding a note at the point of purchase would have eliminated the exchange entirely. They did not, because nobody told them which eight seconds mattered.
Most firms hand clients a receipts policy that amounts to "keep your receipts". That instruction is not wrong, it is just insufficient — it specifies the artefact and omits the metadata, and the metadata is what cleanup work consists of.
The three fields that cause most cleanup
Across a typical client file, reconstruction work clusters around three absences.
Business purpose. Not the category — the reason. "Meals" is a category. "Client lunch, Hartley account, 3 attendees" is a purpose. The category can often be inferred from the merchant; the purpose never can.
The apportionment basis on shared costs. A client claims 60% business use on a phone. You have no way to evaluate or defend that number, and neither will they in two years. The fix is not a different percentage, it is one line recording how it was derived.
The link between a payment and its document. Most clients keep both and connect neither. You receive a folder of 180 images and a bank export with 240 lines, and matching them is a manual join on amount and approximate date.
Everything else in a cleanup — miscategorisation, duplicates, missing items — is either cheap to fix or detectable automatically. These three are expensive because they require the client's memory, and memory is the asset that depreciates fastest.
The specification to hand clients
This is short enough to fit on one page, which is the point. A two-page policy is not read.
| Column | Required | Notes |
|---|---|---|
| Date | Always | Transaction date, not statement date |
| Merchant | Always | As printed on the document |
| Amount | Always | Total paid |
| Tax component | If registered | From the document, not calculated |
| Category | Always | From the dropdown list we provide |
| Business purpose | Always | One line, written on the day |
| Business use % and basis | If shared | e.g. "70% — call log for March" |
| Project/property/job | If applicable | Short code, from the list |
| Receipt image link | Always | Linked from the row |
Nine columns, four of which are automatic if they scan rather than type.
Two implementation details matter more than the list itself.
Provide the category list; do not let them invent one. Build it from the lines of the return you will actually file, lock it with data validation, and put it on a separate tab. Client-invented categories are the root cause of the mapping work you do every year.
Insist the image is linked from the row, not stored in a folder. A Drive folder of 180 photographs named IMG_4471 is not a filing system. A row with a link is.
What to do about the clients who will not do it
Some proportion of any client base will not maintain a spreadsheet, and it is usually not the clients you would predict. Treating this as a compliance problem produces friction and no records.
Segment them instead, and price accordingly.
Tier 1 — structured. They maintain the sheet. Your monthly work is review and exception handling. These clients are profitable at a fixed fee.
Tier 2 — captures but does not organise. They will photograph receipts consistently if the act of photographing is the whole job, but will not fill in columns. This is the largest group, and the right answer is a capture tool that fills the columns automatically. A scanning app that extracts merchant, date, amount, and tax, assigns a category, and links the image into a shared sheet turns a Tier 2 client into something close to Tier 1 without changing their behaviour. Scan2Sheet is built for exactly this arrangement — the client scans, the row lands in a sheet you both have access to.
Tier 3 — shoebox. They will do nothing until January. Do not pretend otherwise. Price the reconstruction explicitly as a separate service with its own fee, so the cost sits with the client whose choices create it rather than being absorbed into a fixed fee quoted on the assumption of Tier 1 behaviour.
The segmentation itself usually changes behaviour more than any policy document, because it is the first time the cost of disorganisation has been made visible to the person generating it.
The review process on your side
With a shared sheet, the monthly cycle becomes four filters rather than a read-through.
- Blank purposes.
=COUNTIF(Purpose,"")— chase these while the month is recent, not at year end. - Blank or non-list categories. Data validation should make this impossible; if rows appear, someone has pasted data in, which is worth knowing.
- Rows above a materiality threshold. Set one per client. Review these individually; sample the rest.
- Possible duplicates. Same merchant, same amount, within a few days. A conditional formatting rule or a helper column catches most.
Then one reconciliation check: does the sheet's row set correspond to the bank feed for the period? Unmatched in either direction is the only question that matters, and it is the one most likely to reveal something real — a personal card used for business, or a business payment never recorded.
The handover that actually saves time
At year end, what you want from a client file is not a tidy spreadsheet. It is the ability to answer any question about any number without contacting them.
Test it before you need it. Pick five transactions at random from the middle of the year and try to answer, from the sheet alone: what was this, why was it a business cost, and where is the evidence?
If all five answer cleanly, the file is in good shape and your year-end work is mechanical. If two of them require an email, you have found the gap while the client can still close it — which is the entire argument for running the test in September rather than discovering the answer in January.
Setting a client up
The onboarding conversation is ten minutes and covers four things: the category list and why it matches their return, the purpose field and that it must be written the same day, how receipts get captured, and the monthly review you will do and what you will chase.
Make the first month's review a conversation rather than a correction. Show them the three rows you could not interpret and ask what they were. Clients generally find this more persuasive than any policy, because it demonstrates the cost of the gap using their own data — and once someone has experienced not remembering, the eight seconds stops feeling optional.