Check the Dates Before You Write the Story

2026-09-16 · Philip Choo · AI9OS

A deposit lands in an account. Two large transfers arrive in the same account soon after. The obvious reading writes itself: the deposit was funded by the transfers.

It was not. The deposit predated the transfers by three days.

Nobody who repeated the story was careless. The client wanted it to be true, because it fit what they already believed. The referrer had heard it from the client and passed it on as background rather than as a claim to check. The file itself had been named for the narrative before the dates were laid out, so opening the folder already told you what you were supposed to find. By the time anyone wrote a sentence of analysis, the sequence had been assumed by everyone in the chain, and assumed is not the same as checked.

Agreement is not evidence

This is the part worth sitting with, because it is the part that makes the mistake survive as long as it does.

A story that several people already believe does not need defending. Nobody asks a deposit's exact timestamp when the client is certain, the referral is confident, and the two facts obviously belong together. The question that would have caught it, does the deposit's date actually come after the transfers', is not a hard question. It is simply a question nobody had a reason to ask, because everyone in the room had already answered it by agreeing with each other.

Consensus is a social fact. Sequence is a factual one. They travel together often enough that it is easy to mistake the first for evidence of the second, and the only thing that tells them apart is looking at the actual dates on the actual documents.

Where it goes wrong, and why it is cheap to miss

Financial records rarely arrive already ordered by the axis that matters. A bank statement is ordered by posting date, which is not the same as the date value actually moved. A remittance advice carries the sender's date, which may differ from the date the receiving bank credits it. An registered instrument carries a lodgement date, which can lag the transaction it records by days or weeks. Stack a deposit slip, two remittance advices and a statement extract next to each other and each one is telling you about a different clock.

Reading four documents and getting four true dates is not the same as knowing what order the events happened in. That reconciliation step is exactly the one that gets skipped when the narrative already feels settled, because the felt sense of "this is obviously right" does not distinguish between a sequence you have checked and one you have merely not questioned.

The fix is mechanical, and that is the point

Lay every dated event from every document on one timeline before writing a single sentence of analysis. Not a mental timeline. A written one, with the source and the specific date field cited next to each entry, because different documents will disagree about what "the date" even means for the same event.

Then read the timeline the way a skeptic would, rather than the way the story wants you to. If the sequence does not support the inference, that is not a footnote to add later. It goes into the file note in terms: this is a sequencing question, not a finding, and it stays labelled that way until someone resolves it, not until someone forgets it was ever a question.

The distinction matters because a finding and a question get treated completely differently downstream. A finding gets built on. A question gets checked. Mislabel one as the other and you have removed the one signal that would have stopped the next person from building on sand.

Why this is worse in front of a judge than at the desk

Every other way of getting a fact wrong tends to surface early, because someone along the way asks an obvious question and the file does not have an answer. A sequencing error built into an agreed narrative surfaces late, because nobody along the way has a reason to ask the obvious question. The client will not ask it, since the story is theirs. The referrer will not ask it, since they are relaying what they were told. The analyst will not ask it, if the file was named for the conclusion before the dates were checked.

So the error travels intact through drafting, through review, through counsel's own read of the report, and it survives until the one moment specifically designed to test it: cross-examination, or opposing counsel with the same three documents and considerably more motivation to read them in the order they actually occurred. The finding does not fail quietly. It fails in the forum with the highest cost of failing.

This is the same family as the filename problem, on a different axis

A filename is a claim a person made about a document, and it travels into a schedule unchecked because checking it costs a step nobody remembers to take. A sequence assumed from an agreed story is the same shape of error, one layer up: it is a claim several people made about several documents at once, and it travels into an analysis for exactly the same reason. Neither failure looks like a gap when it happens. Both look like a slightly more complete picture than the file actually supports.

What to ask, before relying on any narrative built from more than one document

• "What is the specific date field on each document, and whose clock is it?" A posting date, a value date, a lodgement date and a sender's date are four different things wearing the same label.
• "Has the full sequence been laid on one timeline, with sources cited, or has it been read off the story?" These produce different documents and only one of them can be checked by someone else.
• "Where the sequence is uncertain, does the file say so, or does it say nothing?" Silence on a sequencing question reads as confidence. It should read as an open item. A transfer out of a joint account is the classic case, and it is one of the dead ends in what a matrimonial asset trace actually costs you.

If you are the one relying on the answer, ask the version that actually tests it: "Show me the three documents this rests on, in date order, with the field you took each date from." A firm that has done the work can produce that in minutes. A firm that has not will produce the narrative again, more confidently the second time.