Receipts are the worst-behaved documents in any office. They arrive as phone photographs, thermal prints that fade, PDF attachments, screenshots of an app, and forwarded confirmation emails with no attachment at all. There are a great many of them, each one carries barely any information, and they are all shaped differently.
The good news is that the small amount of information is genuinely small: six columns covers almost every expense claim. The work is in the collection and the checking, not the reading — which is the opposite of where most people put their effort.
Decide what the sheet is for first
Receipts get processed for three different reasons, and they need different columns. Deciding which before you start saves redoing it.
Reimbursement. Paying someone back. You need who, how much, and a date. Tax detail is optional.
Tax reclaim. Recovering input VAT or sales tax. Now you need the tax amount, the net, and the supplier's tax registration number — and a receipt missing the registration number is not reclaimable, so knowing which ones are missing it is the whole point of the exercise.
Cost analysis. Finding out what is being spent on what. Now the category matters more than the tax detail, and consistency of category matters more than anything.
Most teams need the first two. Build for both from the start: the tax columns are cheap to capture at the same time and expensive to go back for.
The columns, and why six is enough
Source file
Claimant
Receipt date
Merchant
Currency
Net amount
Tax amount
Gross amount
Category
Payment method
Ten named, six that do real work.
Source file first, as with every extraction on this site. A receipt sheet is exactly the thing someone queries nine months later, and the column is the difference between answering in ten seconds and re-opening a folder of photographs.
Claimant is the column specific to receipts. Invoices belong to the company; receipts belong to a person, and a receipt sheet without a claimant column cannot produce the one report it exists to produce.
Payment method earns its place for an unobvious reason: it is how you catch the receipt that has been claimed and paid on the company card. Duplicated expense is the commonest real loss in this process and that column is how you see it.
Net, tax and gross as three columns, not one total. It costs two columns and it gives you an arithmetic check on every row — net plus tax equals gross — which is the cheapest verification available on documents that mostly arrive as photographs.
And leave a cell empty when the receipt does not state the value. Plenty of receipts do not break out tax. A blank is a true fact about the receipt; a number you apportioned is a calculation you invented, and in six months nobody can tell which is which. A filter on blank tax is also precisely your list of not-reclaimable receipts.
Collection is the actual problem
Ask anyone who does this monthly where the time goes and it will not be the typing. It is chasing eleven people for the receipts they have in a coat pocket, and converting nine different file formats into something processable.
Four things help, all cheap:
One destination, and a deadline. A single mailbox or shared folder, with a monthly cut-off that is enforced. More of this process's pain is caused by a lack of a deadline than by a lack of tooling.
Photograph it at the moment of purchase. Thermal receipts fade, genuinely and quickly — a receipt photographed in the restaurant is legible and the same receipt photographed in three weeks may not be. Most phones have a document mode that flattens and squares the page; it makes a real difference to what can be read later.
Name the file on arrival. 2026-08-14-claimant-merchant.pdf sorts
chronologically, searches usefully, and makes the Source file column meaningful on
its own.
Get the PDF when there is one. Rail tickets, software subscriptions, hotels and airlines all email a proper PDF receipt. Claiming a screenshot of the confirmation email instead is throwing away a perfect document in favour of a bad picture of it.
Get everything into one format before extracting
This is the step that makes a mixed pile tractable, and it is boring.
Most extraction works on PDFs, so photographs need converting. Both desktop systems do it without installing anything: on macOS, select the images in Finder and use Quick Actions → Create PDF, or print them to PDF; on Windows, select the images and use Print → Microsoft Print to PDF. Phone scanning apps produce PDFs directly, which is the reason to use the document mode rather than the camera.
Two specifics worth knowing. HEIC photos from an iPhone are not accepted by much; converting to PDF solves that at the same time. And one receipt per file is much better than a sheet of six photographed together — a combined image is one document as far as anything reading it is concerned, and it will produce one row.
If your receipts are already PDFs, check what is inside them before planning anything. Extract Text from PDF tells you per page, in the browser, whether a document has a text layer. A PDF receipt emailed by a supplier usually does; a PDF wrapped around a photograph does not, and converting scanned PDFs to Excel is then the relevant route.
Extract, then check by arithmetic rather than by eye
With a clean folder and a column list, extraction is the ordinary field-extraction problem: extracting specific fields from PDFs covers how to name a field so it has one answer, which matters here because receipts are inconsistent about almost every word they print.
The checking is where receipts differ from invoices, because most of them started as photographs and misread digits are a real possibility. Five checks, all formulas:
- Net plus tax equals gross, on every row. The rows that fail are your whole review list.
- Sort each amount column. A lost decimal point is obvious at the top of a sort and invisible in the middle of a sheet. Receipts make this error more often than any other document type because of the photography.
- Claimant totals against what each person claimed. If people submit a claim total, that is an independent check of their rows.
- Duplicates. Claimant plus date plus merchant plus gross should be unique. One conditional format finds the receipt submitted twice and the one also paid on the company card.
- Dates inside the period. Sort them; anything outside the month is either a late claim or a misread.
Then keep the images. On receipts the photograph is the evidence, and the Source file column is what makes it reachable.
Categories: fewer than you want
A practical note, because this is where receipt sheets decay.
Use a short, fixed list of categories and a dropdown rather than free text. Twelve categories is plenty and twenty is too many — the sheet becomes unanalysable not because the categories are wrong but because "Travel", "travel" and "Taxi/travel" are three different things to a pivot table.
Put the list in its own sheet, validate against it, and resist adding a category for a single receipt. The one-off goes in "Other" with a note, and if "Other" grows, that is the signal to add a category deliberately.
Common questions
Can I extract receipts in bulk? Yes, and that is the right way to do it: one folder, one column set, one row per receipt. The workflow is the same as converting multiple PDFs into one spreadsheet — receipts are just unusually short documents.
What if a receipt does not show the tax separately? Leave the tax cell empty. Do not compute it. Many jurisdictions require the tax to be stated on the document for it to be reclaimable, so the empty cell is the answer to a real question rather than missing data.
What about foreign-currency receipts? Capture the currency and the amount as printed, in separate columns, and convert in the spreadsheet with the rate in its own column. A receipt converted at an unrecorded rate cannot be checked by anyone later.
Do photographs work as well as PDFs? No, and it is worth being plain about that. A PDF receipt from a supplier contains text; a photograph contains pixels that have to be interpreted, and the interpretation is occasionally wrong in plausible ways. Get the PDF where one exists; photograph carefully where one does not.
Should receipts and invoices go in the same sheet? No. The columns differ — receipts need a claimant, invoices need a due date and a PO reference — and most columns would be empty for half the rows. Two batches, two sheets. Consolidating invoices into Excel is the invoice-side process.
How many receipts before automating is worth it? Lower than people think, and the reason is not the typing. It is that receipt processing is interruption-sensitive work that gets postponed to a single unpleasant block, and the postponement is what loses the receipts.
The short version
Receipts are an easy reading problem and a hard logistics problem. Fix the collection — one destination, one deadline, photographed at purchase, named on arrival — and capture net, tax and gross separately so the sheet can check itself.
Then the extraction is the same job as every other document stack on this site, and how it works describes that part.