What attendance data is not for

The three bad inferences

Hours present as effort. Someone at their desk for nine hours may have spent four of them blocked on a review nobody did. The clock cannot tell.

Punctuality as commitment. Arrival time is mostly a function of a commute, a school run, a bus route and a household. Rewarding it rewards the people with the shortest journeys.

Absence as reliability. A person with a chronic condition, a caring responsibility or a rough quarter generates a pattern that looks identical to disengagement in a monthly export.

Our own leaderboard, and why it needs care

PeopleMuster ships a monthly punctuality ranking. It names the top three and, in a section labelled Room for Growth, it names the lowest-ranked people with grades.

Saying that plainly is fairer than leaving it in a feature list. A product that publicly ranks people by arrival time is making a claim about what the workplace values, and it is worth deciding whether you agree before you switch it on.

In some workplaces this reads as ordinary recognition. In others it reads as surveillance, and the reaction is immediate and hard to undo.

So this is a judgement call an administrator should make deliberately, with the people it names in mind, rather than a default to leave running because it was there.

Questions worth asking before you switch any of it on

The same questions apply to rankings, reports and exports alike.

• Who can see this, and is that the smallest group that needs it?
• What decision will it be used for, and is the data good enough for that decision?
• Does the person being measured see what their manager sees?
• Can someone correct a wrong record, and how long does that take?
• If a number looks bad, what is the first step. A conversation, or a consequence?

What the data genuinely supports

Coverage. Whether enough people were present on Tuesday to run the shift. That is an operational question and the records answer it well.

Correction. A missed punch is a data problem with an owner, and a request flow is how it gets fixed rather than argued about.

Pattern noticing, at the level of a prompt. A change in someone's pattern is a reason to ask how they are. It is not a finding.

Everything past that needs a second source and a conversation.

Who sees what, and why that is the real control

Most of the damage comes from audience rather than collection. The same row is unremarkable to the person it describes and corrosive on a shared screen.

So the question to settle first is not what gets recorded. It is who can see it, in what form, and how far it travels.

A personal view of your own record is uncontroversial. A departmental export that circulates by email is a different object, and it keeps circulating long after the month it described.

Access here is governed by roles, with attendance and device sync as separate permission domains. Deciding that boundary deliberately does more than any wording change.

What people conclude when the tone is wrong

Measurement always teaches something, and it rarely teaches what the dashboard intended.

Measure arrival and reward it and people optimise arrival. They turn up on time and leave on time, including on the days the work needed an extra hour.

Publish a ranking and the people near the bottom mostly stop volunteering for anything visible. That is a rational response, and it is invisible in the data that caused it.

None of which argues for recording nothing. It argues for being able to say, out loud, what each number is for. A measure nobody can justify gets resented, and then it gets gamed.

Correction has to be easy, or the data rots

One practical thing does more for attendance data than any policy about it. Make a wrong record trivial to fix.

When correction is slow or embarrassing, people stop bothering. The record fills with small errors nobody disputes, and within a quarter the whole dataset is treated as approximately true.

When correction is a form that takes a minute and gets reviewed the same week, people fix things. The data improves because the cost of being accurate dropped.

That is also the fairest version. A person should be able to say the record is wrong, be heard quickly, and not have to make a case about it.

A correction queue that someone clears weekly is worth more here than any amount of careful wording about what the numbers mean.