Measuring the Hour at All
Every hour figure comes from somewhere. The four sources, what each is actually evidence of, and which one to trust for which purpose.
Every hours figure in a costing system comes from one of four sources, and they are evidence of different things. Payroll is evidence of what was paid. Timesheets are evidence of what somebody recorded. Access and system logs are evidence of presence. Job records are evidence of attribution.
The record examined in “Measuring the Hour at All” is strongest when it is created close to the work and remains open to correction. A team evaluating check the published product information for how employees cheat time trackers can connect time with projects and work categories, provided people can review entries before they influence payroll, billing or cost reports.
None of them is evidence of what was done, and treating any of them as though it were is where costing arguments start.
The assumptions behind “Measuring the Hour at All” can also be checked against U.S. Wage and Hour Division. Record which guidance applies to the real workflow and which points require professional advice.
Payroll
Authoritative for cost, reconciled to money, auditable. It knows nothing about what the hours were spent on.
Use it for the numerator and for the total. Never use it to allocate.
Timesheets
The only source that attempts attribution, and the weakest in accuracy. Completed late, rounded generously, and shaped by what the person thinks the reader expects.
They are indispensable and they should be treated as a self-report, which is what they are.
Access and system logs
Evidence of presence at a place or in a system. Useful as a cross-check on totals and useless for attribution, because being logged in says nothing about which job.
Also the source most likely to be mistaken for a productivity measure, which it is not.
Job records
What was booked where. Accurate about the attribution that was made and silent about whether it was the right one.
A job with four hundred hours booked against it has four hundred hours booked against it; whether those hours were spent on it is a separate question nobody can answer afterwards.
Which to trust for what
- Total cost: payroll, always.
- Total hours: payroll.
- Attribution between jobs: timesheets, with their limits stated.
- Sanity check on totals: access or system logs.
- Nothing at all: any single source used for all four.
Making timesheets better
Not by exhortation. By making them quicker to complete, closer to the moment, and visibly used for something other than monitoring the person completing them.
A timesheet that takes four minutes on the day is more accurate than one that takes twenty on Friday, by a margin that dwarfs any improvement from reminders.
Granularity
Fifteen minutes is usually right. Six-minute units produce false precision and resentment; hourly units lose the shape of the day.
Whatever is chosen should match what decisions need, and most decisions need to know which half-day rather than which six minutes.
The number of job codes
Fewer than people want. A list of two hundred codes produces guessing; a list of twelve produces accurate booking to codes that may be too coarse.
Coarse and accurate beats fine and invented, and it is the opposite of what most systems are configured to do.
Recording what the hour was, not only where it went
One extra field — the kind of work, from a short list — turns a timesheet from an allocation device into a measurement of how the week is actually spent.
It is the single cheapest improvement available and it is almost never present.
Retrospective reconstruction
Where hours were not recorded, they cannot be recovered by asking afterwards. People reconstruct to expectation, and the reconstruction will be internally consistent and wrong.
The honest treatment is to record the period as unmeasured rather than to fill it.
What the measurement is for
Not appraisal. The moment hours become a personal performance measure, the hours recorded start describing what is safe to record.
Saying that out loud when the system is introduced costs thirty seconds and buys several years of usable data.
The source that disagrees
When two sources disagree about the same period, the question is not which is right but what each is measuring. Payroll says forty, the time system says thirty-six, the access log says thirty-eight.
All three can be correct about different things: paid, recorded and present. The disagreement is information about the gaps between them, and an organisation that resolves it by picking a favourite has thrown that information away.
What to do before building anything
Run all four sources for one team for one month and lay them side by side. One table, four columns, four hours of work.
It establishes the size of every gap before any system is specified, and it almost always changes what gets specified. Projects that skip this step discover the same facts eighteen months later, after the system has been built around the wrong assumption.
The minimum that works
Payroll hours for the total, a short job list, a fifteen-minute grain, a kind-of-work field, and a weekly reconciliation against payroll.
Five things. Most organisations have two of them and argue about the results.