Is the Data Any Good

A clock-in system reports confidently whether or not it is working. These four checks take an afternoon and are worth repeating quarterly.

Refusal rate

Refused punches as a proportion of attempts, per site, per week.

The headline number, and most deployments do not track it.

A site far above the others has a boundary or a coverage problem, and checking the geometry before speaking to anyone is the correct order.

A fleet-wide rise usually follows a phone operating system update that changed how location permissions behave.

Accuracy distribution

The accuracy figures recorded with punches, per site.

A median of eight metres and a median of sixty are different systems wearing the same name.

Track the tail: the proportion of punches with accuracy worse than the radius, which are the ones that succeeded or failed by luck.

Where that proportion is high, the site needs a different method, not a bigger circle.

Fallback usage

How often the primary method failed and something else was used.

By site and by cause: no fix, no signal, no battery, no phone.

Rising usage means the primary method is degrading, frequently for a reason you can fix.

Zero usage is also a signal — it usually means no fallback exists and people are working unpaid or being marked late.

Correction turnaround

Time from a worker submitting a correction to it being resolved.

This is their experience of the whole system.

Days means corrections pile up until payroll, which is when errors become pay errors.

A rising figure means the approver is overloaded or the refusal rate has climbed.

Reporting them

On the same page as any attendance analysis, so a reader knows what weight to apply.

Per site, because sites differ far more than people do.

With the radius stated, since it explains most of the variation.

Acting on them

High refusal rate: check the radius and the coverage, in that order.

Poor accuracy: consider a second method rather than enlarging the boundary.

High fallback usage: the same.

Slow corrections: name an approver and a deadline.

In every case the first hypothesis should be that the system is failing, not that people are. It is right more often, and being wrong the other way costs the cooperation the whole arrangement depends on.

System first, people second

The order that keeps the numbers usable.

A high refusal rate is a geometry or coverage problem until proven otherwise.

A person with many refusals usually works at the site with the problem.

Check the site, then the phone platform, then the individual.

Reversing that order is right occasionally and costs the cooperation the record depends on every time it is wrong.

Check the difficult case

Use the consulting workflow to frame one representative test. The useful evidence is the record created when an employee corrects an entry and an administrator exports it.

Independent reference

For an external point of reference, see the Office for National Statistics. Use this established source as an external check before turning the principle into a workplace rule.