Reading a time report when the numbers disagree with your memory
· 6 min read
The first week of tracking almost always produces an argument with yourself. The record says four hours twenty in the editor and it felt like most of the day. Or it reports nine hours and you are sure you finished at five.
Before adjusting anything, it is worth working out which side is wrong, because it is not always the record and it is not always you.
Reasons the record may be understating
Idle detection. Most tools stop counting after a few minutes without keyboard or mouse input. Reading a long document, thinking, or being on a phone call while looking at the screen all register as absence. Anyone whose work involves sustained reading loses genuine time this way.
Work away from the Mac. Meetings, calls on a phone, paper, whiteboards, travel. None of it is visible to a desk-based record, and for some roles it is the majority of the day.
A second machine. Each Mac keeps its own record. Tracking time across two Macs covers combining them.
A gap. The Mac slept, the app was not running, or an update reset something. When time tracking stops recording covers diagnosing that.
Reasons your memory may be overstating
Duration neglect. People remember events and estimate their length badly, and the estimate is biased by intensity. A difficult forty minutes is remembered as longer than an easy ninety.
The interruption tax feels like the work. Twenty minutes of writing, interrupted four times, feels like two hours of writing. The elapsed time may well have been two hours; the writing was eighty minutes, and the record is measuring the second thing.
Counting the whole day as work. The span from first email to last is not hours worked. Lunch, coffee, the walk, the conversation. A nine-hour span containing six hours of work is normal and does not feel like it.
The absorbing work compresses. Time that goes well is remembered as shorter than it was, which is the opposite bias to the one above and produces the reverse disagreement.
Working out which
Take one day where the disagreement is largest and reconstruct it deliberately.
Write down what you remember doing, in order, with your estimate for each. Then put the record beside it.
The interesting output is not the total but where the differences sit. If they cluster in one application, you have found a specific effect: probably reading time being trimmed, or an application you switch to briefly and often. If they are spread evenly, the record is probably right and your sense of the day is uniformly inflated, which is the common finding.
Do this once. It is not worth repeating; the point is calibration, and after one careful comparison you will know which direction your estimates run.
What to change
If reading time is being trimmed, know that the record is a floor rather than a measurement. Your real desk time is somewhat higher. That is a stable bias you can adjust for mentally, and it is more useful than a tool that counts you as present whenever a window is open.
If work happens away from the Mac, note it by hand. A receipt with a pencilled line saying “two hours, site visit” is a better record than either half alone.
If your memory was simply wrong, that is the useful outcome, and it is what the exercise is for. Most people’s estimate of their own working time is off by hours in one direction or the other, and knowing which is worth more than the individual day. Why your estimate of your own workday is wrong by hours covers the research-shaped version of this.
What not to do
Do not edit the record to match what you remember. The value of an automatic record is that it was not produced by your judgement, and adjusting it to agree with your judgement removes exactly that property. Annotate it instead, alongside, so the measured part and the remembered part stay distinguishable.
Do not conclude from one surprising day that the tool is broken. Check three or four days first. A consistent discrepancy has a cause worth finding; a single day is usually a gap or an unusual day.
Where Punchcard sits
Punchcard records the frontmost application by name and subtracts idle time based on input. That means it errs towards under-reporting rather than over-reporting: leaving the desk stops the count, and so does reading without touching anything.
That is a deliberate choice. A record that flatters you is worse than one that is slightly stingy, because you would act on it. The consequence is that the total is a floor, and for people who read a lot on screen it is a noticeably conservative one.
It also does not know what any of the time was for. It sees an application, not a document, so six hours in a browser is six hours in a browser. What automatic time tracking can and cannot know sets the limits out in full, and the difference between time spent and time worked covers the distinction the disagreement usually turns on.
Questions
Should I lower the idle threshold so reading counts? It is a trade with no right answer. A longer threshold counts your reading and also counts your lunch. Knowing which way your tool errs is more useful than tuning it.
The record says I worked more than I thought. Is that good? It is information. The common version is a day that felt like six hours and measured eight, with the difference in small pieces across the evening. Whether to act on it is a separate question.
Can I see a per-hour breakdown to check a specific afternoon? Depends on the tool. A daily summary answers “how long”, not “when”. If you need the timeline, that is a different kind of tool. Why a daily summary beats a live dashboard covers the trade.
How long before the numbers stop surprising me? Usually about two weeks. The first few days are the calibration; after that the record and your sense of the day converge, which is itself the main benefit.