Time Tracking · Practice
Reading Time Reports Without Drawing the Wrong Conclusion
The number is not the finding. What to check before believing a report, and the readings that look obvious and are usually wrong.
A time report looks like measurement. It is an aggregation of approximate self-reported records with known biases, presented to two decimal places.
Which does not make it useless. It makes the reading a skill.
For a vendor-side perspective on remote employee monitoring software, see this guide from Monitask.
Before believing any number
How was it recorded? Live entry, timer, or reconstructed at the end of the week. Reconstruction is estimate. See why time data is almost always wrong.
What is missing? If there is no category for waiting, admin or rework, those hours are inside the project numbers, inflating them invisibly.
Did the method change? New tool, new categories, new policy — any of these creates a discontinuity that looks like a real change and is not. Mark the date on the chart.
Is the period comparable? Holiday months, a quarter with a launch, a period with three people off. Comparing across these produces conclusions about the calendar.
Who is missing? Partial compliance means the report describes the people who record diligently.
The readings that are usually wrong
"This project came in under estimate"
Usually a recording gap, not efficiency.
Check whether hours went somewhere else — a catch-all category, a different project, or unrecorded.
Surprisingly low is a prompt to investigate, not a result to celebrate.
"This person is less productive"
Different work. Almost always.
The person doing difficult, uncertain, or interrupt-driven work posts worse numbers than the one doing routine work.
And the report cannot see the difference. See tracking time against measuring people.
"Billable utilisation is down"
Check whether non-billable recording improved before concluding anything about output. Better recording of admin looks identical to less billable work.
"Meetings are 30% of our time"
Plausible, and check what got coded as meetings. Where the category list is thin, meetings absorbs anything collaborative.
"This client is profitable"
Only if non-billable time for that client was recorded. Briefing, revisions, chasing, relationship management. Frequently it was not, and the profitable client is the one nobody logged the extra hours for.
"The trend is improving"
Over how many periods, and did anything change in method? Three points is not a trend, and two of them may straddle a tooling change.
What reports are actually good at
Comparison between similar things. Two projects of the same type, same team, same period conventions.
Ratios. Billable to total, planned to unplanned, project to admin. Ratios survive consistent bias because the bias applies to both terms.
Trends over months, with method changes marked.
Finding the surprising category. The one holding far more hours than anyone expected is usually the finding, and it is usually a process problem.
Estimating support. Actual against estimate across several jobs. See estimating from your own data.
And prompting a question. The best use of a time report is deciding what to go and look at.
What they are bad at
Absolute claims. "This cost 340 hours" carries an unstated error bar.
Individual comparison. The error is larger than the difference.
Short periods. A week is noise.
Anything requiring precision the recording method cannot support.
And causation. The report shows that hours went somewhere. It does not show why.
Presenting one honestly
State the recording method and the known bias direction. "These figures under-record waiting time" changes how a room reads the chart. For broader independent background, see ILO guidance on working time.
Mark method changes on the timeline.
Show ratios alongside absolutes.
Show the catch-all category rather than hiding it in other. Its size is a quality indicator for the whole dataset.
Report by work, not by person, as the default.
And present findings as questions where they are questions. "Approvals appear to be costing two days a week — worth checking" is honest. "Approvals cost two days a week" is not, from this data.
A short checklist
- [ ] Recording method known, and its bias direction stated
- [ ] Non-productive categories exist and are being used
- [ ] Method changes marked on any timeline
- [ ] Periods genuinely comparable
- [ ] Catch-all category shown and its size checked
- [ ] Ratios alongside absolute figures
- [ ] Surprising numbers investigated rather than believed
- [ ] No individual comparisons
For solo trackers
The same reading discipline, applied to yourself.
Look at ratios, not hours. Billable share, admin share, effective rate.
Look monthly, not weekly.
Investigate the surprising month rather than concluding from it.
And do not compare your logged hours to eight a day. Four to six focused hours is normal. See time tracking for freelancers.
The short version
Check the recording method before believing any figure — reconstruction is estimate.
Surprisingly low numbers are usually recording gaps, not efficiency.
Ratios survive consistent bias; absolutes do not.
Mark method changes on any timeline, because a tooling change looks exactly like a real one.
And the best use of a report is deciding what to go and look at, not concluding.