Reconstructing a year you didn't record
Someone has asked for a year you didn’t keep. There is no shoebox, or there is one and it covers March. This is a recoverable situation, mostly, and the recovery has a definite order.
What follows is the order. It is worth knowing up front which part of the job is genuinely reconstruction and which part is guessing, because they must not end up in the same column.
What survives without you
You held almost none of this year’s records. Other people held a great deal of it.
Your bank and card provider hold every payment. This is the spine of any reconstruction: complete, dated, amount-exact, and independent of anything you did or didn’t file. It establishes that money moved, to a merchant descriptor, on a date. It establishes nothing else — see what makes a receipt useful later on why a statement line is a different artefact from a receipt.
Vendors hold order histories. Anywhere you have an account has a list of what you bought, often with a downloadable document per order. This is the single richest source available and the most neglected, because it takes a login rather than a search.
Your email holds confirmations. Search by amount, by the word “receipt”, by “invoice”, by vendor name. Digital-native receipts were delivered to you and never filed, which means they are still there — when the receipt is already digital covers why an inbox full of them isn’t the same as having them.
Your calendar holds the reason. This is the important one and nobody looks. A calendar entry naming a client, a site visit, a trip, or a meeting is contemporaneous evidence of why a spend happened, written at the time, by you, before you knew you’d need it.
Your photo library holds dated images. Not just photographed receipts — pictures of a job in progress, a delivered pallet, a hotel room, all timestamped.
The order of work
One: get the statement lines into a list. Export or transcribe the whole period. Do not filter yet. You want the complete set of transactions, because completeness is what a reconstruction has going for it and selectivity is what makes it look assembled.
Two: mark the ones that need a record at all. Some proportion of the list is personal, or irrelevant, or obviously not the sort of thing anyone is asking about. Marking these is faster than identifying the ones that matter, and it shrinks the job substantially.
Three: go source by source, not line by line. This is the difference between an afternoon and a week. Working line by line means switching between email, portals, and calendar for every row. Working source by source means one login, everything that source can answer, then the next. Portals first — they produce documents, not just facts.
Four: fill the reason column from the calendar and the correspondence. Separate pass, and a different kind of thinking. You are no longer establishing that a spend happened; you are establishing what it was for.
Five: list what you could not establish. Explicitly, as its own artefact. A reconstruction with a short honest gap list is a far more credible object than one where the gaps have been quietly filled with plausible descriptions.
The line you must not cross
There are three grades of thing in the finished pile and they need to stay visibly distinct.
A record is a document created at the time by someone with reason to get it right: an invoice, a receipt, an order confirmation.
A reconstruction is a record obtained afterwards from a party who held it: a re-issued invoice, a downloaded order document, a statement export. Later, but not invented.
An inference is you deciding what a transaction probably was. “£31.40, hardware shop, mid-March — that’ll have been the studio shelving.” It may well be right. It is not a record, and if it is filed looking like one, the whole set stops being trustworthy.
Mark inferences as inferences. In the filename, in a notes column, however you like — but mark them. Whether an inference is acceptable at all is a question for the person who asked, and it is not one this site answers.
Keep or bin
KEEP OR BIN — a reconstructed year
· Invoice re-downloaded from a vendor
portal
→ KEEP. A real record, obtained
late. Best outcome available.
· Complete statement export for the
period
→ KEEP. The spine. Proves payment,
nothing more.
· Calendar entry naming a client on the
date of a spend
→ KEEP. Contemporaneous evidence
of the reason, which the receipt
never had.
· Statement line with a description you
worked out afterwards
→ NOT A RECEIPT, and not a record.
An inference. Label it.
· A spreadsheet you typed up that looks
like a ledger
→ BIN the appearance, keep the
content. Never let derived work
resemble source documents.
· Whether a reconstruction is acceptable
where an original is missing
→ ASK LOCALLY. Entirely a rules
question, and it varies.Do the portals before anything clever
If you only have one hour, spend it logging into the six or eight vendors you spend most with and downloading everything they hold for the period. It is dull, it needs no judgement, and it produces actual documents rather than inferences — which is a better return than any amount of careful detective work on statement lines.
Two practical notes. Order histories are frequently pruned after some period, and accounts you have closed take their history with them, so this source degrades — see records you don’t actually hold. And when the documents arrive, they need to land somewhere findable, or you will do this twice; finding it again covers that.
What this doesn’t settle
Whether reconstructed records satisfy anything. Whether a statement line will do where a receipt is gone. Whether an inference you have honestly labelled is any use at all, or whether labelling it is what disqualifies it. Whether there is a rule about missing records specifically.
All of that is decided by your tax authority or an adviser applying rules to your circumstances, and it varies. What this page can say is that a labelled gap is a defensible thing to hand over and a confident fabrication is not, regardless of which jurisdiction you are in.