Most time logs die on day four.
Not because the person lacked discipline. Because the log asked for more precision than a working day can produce, and by Wednesday the honest options were falsifying it or abandoning it. Everyone abandons it, then concludes they are bad at tracking time.
The time log below is designed backwards from that failure. It has six columns, one recording rule, and a hard stop at two weeks. It is a less accurate time audit than a proper time study, and it will still tell you the thing you need to know.
Run it after a calendar audit, not instead of one. The audit tells you what was claimed; the log tells you what actually happened, and the gap between them is the whole reason to do this.
Why you need a time audit if you already have a calendar
The calendar is a record of scheduled time. Your week is not made of scheduled time.
Four things live outside it, and in a bad week they are most of it: the meeting that ran twenty minutes over, the preparation you did at 7am, the “quick question” that took forty minutes and a lost train of thought, and the block you defended on the calendar and then spent on someone else’s escalation.
A calendar audit cannot see any of that. It will tell you that you had 14 hours of meetings. It will not tell you that you had 14 hours of meetings and produced nothing else, which is the sentence that changes a conversation.
The six columns
Nothing else. Every column you add lowers the odds of reaching day ten.
| Column | What goes in it | Why it earns its place |
|---|---|---|
| Start | Clock time, rounded to 15 minutes | Precision beyond this buys nothing and costs compliance |
| Minutes | Duration, rounded to 15 | Same |
| What | Four or five words. “Q3 forecast”, “Priya escalation” | You are identifying it later, not writing a report |
| Type | One of: meeting, focused, reactive, admin, gap | The only analytical column, and it does all the work |
| Trigger | planned, someone asked, I drifted | This is the column that surprises people |
| Cut short by | Blank, or what interrupted you | Where fragmentation becomes visible |
Five types, and the definitions have to be tight or the log becomes uninterpretable by week two:
- meeting — anyone else was in it, scheduled or not. A 20-minute unscheduled call is a meeting.
- focused — single task, more than 25 minutes, no channel-switching. If you checked messages, it was not focused. This is strict on purpose; the count of genuinely focused hours is usually the finding.
- reactive — real work, but someone else’s queue set the order. Replying, reviewing, unblocking, answering.
- admin — expenses, tickets, scheduling, approvals. The work of having a job rather than doing one.
- gap — the 25 minutes between two meetings, the wait for a call to start, the ten minutes after something ran over. Do not fold these into anything else. Their total is a finding.
Trigger is the column people want to drop and should not. I drifted is not a moral category; it is the tell that you were too depleted to start the hard thing, and it clusters. Almost always it clusters directly after a long meeting block, which means it is not a discipline problem, it is a recovery cost — and that is a completely different fix.
The one rule that makes it survive
Log at transitions, not at the end of the day.
The end-of-day reconstruction is the thing that kills time logs. By 6pm you are tired, the morning is gone, and what you write down is a plausible narrative rather than a record. You will smooth over the interruptions, which are exactly what you are running this to find. After two days of that, you know the log is fiction, and you stop.
So: when you change activity, you write a line. It takes eight seconds. Keep the file open in a tab or a note on your phone.
Three concessions that keep it alive:
- A missed hour is a blank row marked
unlogged. Not a guess. Blanks tell you something (they cluster in bad afternoons); invented rows destroy the whole dataset’s credibility with the person who most needs to believe it, which is you. - Anything under 15 minutes is not logged unless it interrupted focused work — in which case it goes in the Cut short by column of the row it interrupted, not as a row of its own.
- If you miss a whole day, you carry on the next morning. You do not restart. A log with two missing days still answers the question; a log you restarted three times does not exist.
Why two weeks and not one, and not four
One week is inside the noise. Every week has a reason it was unusual, and with only one you cannot tell an anomaly from a pattern.
Four weeks is where compliance collapses. By week three the logging is degraded, and degraded data mixed with good data is worse than less good data, because you cannot tell which is which.
Two weeks is long enough to catch a fortnightly meeting and a full sprint or reporting cycle, and short enough that you can hold your attention on it. Pick two consecutive weeks with no holiday, no offsite, and no quarter close. If your role has a monthly cycle with a heavy week, run it once on a heavy week and once on a light one, separately, and do not average them.
Reading it
Two hours on a Friday afternoon, at the end.
Total by type. Meeting hours you already knew. The two numbers you did not: total focused hours, and total gap hours. Focused hours are almost always lower than people expect — the strictness of the definition is doing that, and it is doing it correctly. Gap hours are almost always higher.
Focused sessions by length. Count how many focused blocks lasted more than 45 minutes. If the answer for two weeks is in the low single digits, you have your finding, and it is not about volume. It is about placement, which is the fragmentation problem.
Trigger distribution, by day part. Look at where someone asked and I drifted concentrate. If reactive work owns your mornings, your best hours are going to other people’s queues by default rather than by decision.
Cut short by. Sort the column and count. One name or one channel usually appears far more than any other. That is your single highest-value conversation this month, and it is a conversation about how to reach you, not about a meeting.
Calendar versus log. Overlay the meeting hours from the log onto the hours the calendar claimed. The overrun is real and unbudgeted, and it is typically the difference between a load that balances on paper and a week that does not.
What to do with it
Pick one finding. One.
The log is a diagnostic, not a plan, and the failure mode after a successful log is doing five things at once and learning nothing about which one helped. Take the largest single item — the recurring meeting that overruns every time, the interruption source that owns your afternoons, the reactive work that has colonised your mornings — and make one change for four weeks.
Then, if you want to know whether it worked, run the log again for one week only. You have the columns and the habit by then, and one week is enough for a comparison when you already know the baseline.
Keep the file. Two logs six months apart make an argument that no single week can, and it is the version that holds up after a reorg has changed everyone’s memory of what the workload used to be.
Common questions
Should I use a time-tracking app instead? If you already run one for billing, export it and skip the manual log entirely — you have the data. Do not install one for this. The apps that categorise automatically cannot fill in Trigger or Cut short by, which are the two columns that produce the non-obvious findings, and the manual version teaches you to notice transitions in a way an automatic tracker does not.
My work is too fragmented to log. Then that is the finding, and it took you two days rather than two weeks to get it. Log one day properly, count the rows, and stop. A day with forty transitions does not need further analysis.
Do I show this to my manager? Show the summary, not the log. The raw file contains “I drifted” rows and personal detail that will be read uncharitably by someone skimming. Bring the totals and the one change you want to make.
What if the log shows I am not busy? Then it has told you the truth and saved you an argument you would have lost. More often it shows the hours are there and the shape is wrong: enough time in total, never enough at once. That is a real finding, it is fixable, and it is not the one you were expecting to make.