What to Report, and to Whom

A clock-in system generates a great deal of countable activity. Only some of it says anything about whether the arrangement is working.

The five

Refusal rate, per site, per week.

Fallback usage, by cause.

Correction volume and turnaround.

Accuracy distribution, split by phone platform.

Punch success in the five minutes around shift start, compared against the rest of the day.

All five are about the system. That is deliberate.

What each says when it moves

Refusal rate rising at one site: the radius, the coverage, or a changed entrance.

Rising everywhere on one platform: an operating system update changed permissions.

Fallback usage rising: the primary method is degrading.

Correction turnaround slipping: the approver is overloaded, usually because refusals rose.

Shift-change success below the daily figure: a queue, and people are about to start punching from the car park.

What not to report

A league table of punctuality, which measures queueing and coverage as much as behaviour.

Individual refusal counts, which mostly reflect which site someone works at.

Minute-level lateness, for the reasons in its own note.

Anything that ranks people using a measure this noisy.

Who gets what

The owner: all five, weekly.

Site managers: their own site's refusal rate and corrections.

Payroll: corrections outstanding before cut-off.

The workforce: the refusal rate, published.

That last one is unusual and it is the point. Publishing it says plainly that failures belong to the system, and it is the single most effective signal available.

The annual view

Did refusals fall as sites were fixed?

Did payroll corrections fall?

Is the collection still limited to what was agreed?

Has the notice stayed true after updates?

Would you deploy it the same way again?

That last question produces the useful decisions, and nobody asks it unless a process does.

Report the system, not the people

A rule that shapes every number here.

All five measures describe whether the system is working.

None ranks anyone.

Because the noise — coverage, phone model, queue position, site geometry — swamps any signal about behaviour.

And because a workforce that sees itself ranked on this data will work around it, which removes the signal entirely.

Turn the principle into a test

For an example that can make this requirement testable, consult project time tracking application. Treat the page as a starting point rather than proof and reproduce the workflow with real roles and failures.