A work log is a dated record of what you did, written close enough to the event that it is still accurate. It exists to answer questions asked months later: what happened to that project, what did you ship this half, what should the invoice say, what does the person taking over need to know. Too often those answers are reconstructed instead of retrieved, and reconstruction is where details and emphasis drift.
What the log is actually for
Four recurring demands often arrive with little warning: a performance review covering six or twelve months, a status update needed before a meeting, an invoice or timesheet spanning several weeks, and a handover where any gaps become someone else's problem.
Each of these is a retrieval task against a period long enough that the details have gone soft. The log turns four uncomfortable reconstruction exercises into four searches.
Why reconstruction is systematically wrong, not just slow
Rebuilding a quarter from a calendar and a chat scrollback produces a specific kind of distortion. Calendars record meetings, so the reconstruction over-represents meetings. Ticket systems record what closed, so it over-represents finished work and drops the three weeks spent unblocking something that never got a ticket. And the last few weeks arrive in far more detail than the first few, purely because they are nearer.
The result is a version of your quarter that is accurate in outline and wrong in emphasis—usually understating exactly the work that was hardest to see.
The five-line format
One entry a day, written in under two minutes. Longer formats do not survive; this one does because it fits in the gap between finishing and leaving.
- Date. Not ‘today’—an actual date, so the entry survives being read out of order.
- What changed. One sentence in the past tense, with the thing named: ‘Baxter renewal moved to March; procurement asked for a revised scope.’
- Who was involved. Names, not roles. Six months later you will remember the person and not the job title.
- What is blocked and on whom. This is the line that makes the log worth reading back, because blockers are what you cannot reconstruct.
- One number or link. A ticket, a figure, a document, a percentage. Something checkable that anchors the entry to evidence.
The evidence that small entries matter
Teresa Amabile and Steven Kramer studied daily diaries from 238 people across 26 project teams in 7 companies, producing nearly 12,000 entries. They reported the work in ‘The Power of Small Wins’ in Harvard Business Review in May 2011 and in their book The Progress Principle that year.
Their finding was about motivation rather than record-keeping: progress on meaningful work was the strongest influence on people's inner work lives. The method is still relevant here. Nearly 12,000 contemporaneous diary reports gave the researchers a level of detail that a retrospective summary could not provide. That is evidence for the value of the raw material, even though the study did not test work logs as a productivity method.
A worked example
Two entries from a real-shaped week, to show what ‘specific’ means in practice:
- 2026-03-04—Rewrote the import retry logic after the third silent failure this month; failures now surface in the queue dashboard. With Priya. Blocked on ops for the alerting rule. PR 2214.
- 2026-03-05—Walked Marta through the renewal scope; she wants the training days itemized separately. Nothing blocked. Contract value unchanged at £48k.
- Compare with what reconstruction produces four months later: ‘worked on imports, some contract stuff’. Both describe the same week.
Keeping it alive without a ritual
The failure mode is not forgetting to write it. It is that writing it requires opening something, choosing where the entry goes, and formatting it—three decisions at the end of a day when you have none left.
So attach it to something that already happens and remove the decisions. Speak the entry on the walk to the car and let it be transcribed. Send it to the same place every time with no filing step. If the entry has to become a task—you promised Marta a revised scope—let it leave the log and turn into something that surfaces on its own, rather than sitting in a paragraph you would have to reread to notice.
The retrieval side matters as much as the writing. A year of entries is only useful if you can ask it a question in the words you have now—‘what did I do on the Baxter contract’—rather than the words you used in March.
What a work log will not do
It is evidence, not argument. A log full of accurate entries does not make a case for a promotion; it makes the case possible to write. Turning it into that case is a separate act, and a separate document.
It also will not fix a role where nobody agrees what your job is, and it will not make retroactive entries reliable—an entry written six weeks late inherits all the distortions described above, so it should be marked as reconstructed rather than logged. Finally, a log is only as honest as the day it describes. Entries written to look good in a future review produce a document that reads well and predicts nothing, which is worse than having none.
Five lines, under two minutes, written the same day: date, what changed, who was there, what is blocked, and one checkable number. The value shows up months later, when a review or handover asks for a period you would otherwise rebuild from a calendar—a source that tends to overstate meetings and understate less visible work.