A ruled notebook open on a kitchen table, its left page dense with handwriting, a pen in the fold and a mug beside it in low light

Published October 9, 2026 · The Timeness Editorial Team · fact-checked against the sources below

One test separates a work log worth keeping from one that is not, and it has nothing to do with the format, the tool or the template. It is when the entry was written.

A work log is a dated record of what you did and how long it took, written while it was happening rather than afterward. The one habit that decides whether it is worth keeping is timing. An entry written at the moment is a measurement. The same entry written on Friday is an estimate wearing the clothes of a measurement.

That distinction is not a matter of opinion. Two institutions that depend on accurate time records, a state bar and a national statistics agency, both say so in their own words.

What does a good work log entry contain?

Three things: the date, the duration, and a description specific enough that someone else could tell what the work was. The California advisory on bill padding shows why the third part matters, by setting two versions of the same entry side by side. One reads “meeting with client to discuss the elements of the separate statement of facts and the source of evidence for each element (1.8); research new opinion on the presumptions and burden of proof under Festo and progeny (2.5)”. The other reads “meeting with client re MSJ and research burden of proof (4.3)”.

Same work, same total. The advisory’s point is that the longer, itemized version is characteristic of entries typed while the task was happening, and the short combined one is characteristic of entries written in longhand and typed up later. The detail is not decoration. It is the fingerprint of when the entry was made.

A personal log does not need a client or a matter number, but it does need all three parts. An entry reading “admin, 2h” tells you nothing in November that you did not already know in June.

Why does a work log written on Friday go wrong?

Because by Friday you are reconstructing rather than recording, and reconstruction produces a tidier week than the one that happened.

The California advisory is unusually direct about this. It notes that contemporaneous records are almost universally acknowledged as the best practice, describes the common reality where a day or two or a month passes without entries, and then says what that does to the result: “the longer the time period between when the task was performed and the corresponding time entry is recorded, the greater the probability that the time entry may lack accuracy.” It adds that a reconstructed bill can look extremely precise, with exact times against each itemized activity, and that “the appearance of accuracy may be deceptive”. This is US guidance from a California bar committee about lawyers’ fee bills, not a rule that binds you, but the mechanism it describes is general.

The Bureau of Labor Statistics designs around the same problem from the opposite direction. The American Time Use Survey does not ask people how their week generally goes. Interviewers ask each respondent to recall their activities over a full 24-hour day, beginning at 4 in the morning on the day before the interview and ending at 4 in the morning on the interview day, reporting every activity in that window, and for most activities where they were and who they were with. A single previous day, recalled in order, is what a national statistics agency treats as reliable enough to publish. It does not ask about last week, and the reason is the one above.

You can see the signature in your own log. A reconstructed week is full of round numbers: 2 hours, 1.5 hours, half an hour. A recorded week is full of 25 minutes, 40 minutes, 1 hour 10. Real work does not land on halves.

How much does logging actually cost you?

Less than the week you are trying to reconstruct, and the arithmetic is worth doing once so the objection stops coming back.

Say your day contains six genuine switches of activity, which is modest for most people. That is 6 x 5 = 30 entries in a working week. If an entry takes 20 seconds to write, the week costs 30 x 20 = 600 seconds, which is 10 minutes. Across a 47-week working year that is 47 x 10 = 470 minutes, or 7 hours 50 minutes.

Now the alternative. Say reconstructing a week on Friday takes 25 minutes of working back through a calendar and a sent-mail folder. That is a stipulated figure, like the 20 seconds above, not a measurement. Forty-seven weeks at 25 minutes is 47 x 25 = 1,175 minutes, or 19 hours 35 minutes. On those two assumptions the reconstruction costs more than twice as much time, since 1,175 divided by 470 is 2.5, and produces the less accurate record. That is the whole case, and no part of it depends on which tool you use.

A routine that survives a busy week

Each step here is deliberately small enough to do without breaking what you were doing.

  1. Fix the categories once. Five to ten of them, named after the questions you have. Renaming them on Wednesday destroys the totals, which is the same discipline how to run a time audit recommends for a diagnostic week.
  2. Start the entry when the work starts, not when it finishes. A log kept at the end of each task is already a memory exercise, just a shorter one.
  3. Log the switch, not the task. You are not deciding what the work was called, you are recording that the subject changed. Naming it is a decision, and decisions are what make people stop.
  4. Include the unbillable parts. Admin, invoicing, proposals, reading. Leaving them out is what makes the log say you worked 22 hours in a 40-hour week.
  5. Close the day, do not rebuild it. Two minutes at the end to fix anything obviously wrong is fine. Two minutes is a correction. Twenty is a reconstruction.
  6. Total it weekly, read it monthly. A week tells you almost nothing. Four weeks tells you where the time goes.

What does a tax authority actually require?

Less than people assume, and the requirement is about completeness rather than format. The Internal Revenue Service tells US small businesses and self-employed filers that, except in a few cases, the law does not require any special kind of records, only that they clearly show income and expenses. GOV.UK says much the same to self-employed people in the United Kingdom, where the obligation is to keep records of business income and expenses and the medium is left open. Time tracking in a spreadsheet quotes both rules, and the US employment-side rule, in full.

So nothing here obliges you to use any particular tool, and nothing here is advice about your own tax position, which depends on your jurisdiction and your circumstances. What the rules have in common is the word accurate, and accuracy is a property of when the entry was made, not of what it was made in.

What happens to the log afterward?

It becomes three different things depending on what you need. Totaled by activity, it is the answer to where the week went. Totaled by client and filtered to chargeable work, it is the input to an invoice you raise somewhere else, and what are billable hours covers which entries survive that filter. Totaled across a quarter, it is the only honest basis for quoting the next piece of work. If you want the arithmetic run on start and finish times rather than durations, the hours calculator will do it, and a spreadsheet remains a perfectly good container for the result, with the caveats in time tracking in a spreadsheet.

The part none of those solve is the recording itself. That is where Timeness sits: an activity per client or per kind of work, one tap to start and one to switch, and a day timeline that fills in as the day goes rather than at the end of it. The month and year charts total it by activity, and each entry takes a note, which is where the description of the work goes. The app records time and stops there, so rates, invoices and anything owed to a tax authority stay where they already happen.

Whatever you record in, test the result the same way: count how many entries land on a round half hour. A reconstructed week lands on them constantly, and a recorded one rarely does. Freelance time tracking sets out what to do with the result. Once the log exists, how many productive hours in a day shows what an example week of entries adds up to.

Sources