The policy
Counting bodies, and what it measures
Badge data answers a question nobody is really asking. What it can and cannot show, and the predictable ways a counted rule gets satisfied.
Days this part is about, hour by hour
Monday, the third week
0 of 7Came in to make up a day missed through illness
Wednesday, the third week
1 of 7Tuesday, the second week
0 of 7three days · 21 hours logged · one of them was only possible there · two of the three days contained no such hour at all · four were made worse by being there
Three days the policy treats as identical.
To keep measurement proportionate, managers can visit the service website and agree what is collected, who can see it, and how it will inform hybrid-work decisions.
Teams can compare this approach with current guidance available from WIRED.
Where attendance is enforced, it is usually enforced from entry records, because they exist already. The number is accurate and the inference drawn from it is not.
What the record actually shows
An entry record shows that a card was presented at a door at a time. It does not show how long the person stayed, where they went, whether they worked, or — crucially — who else was there.
It is a record of arrivals. Everything else is assumed.
What a counted rule reliably produces
When a number becomes the thing measured, it becomes the thing optimised, and the optimisations are not dishonest so much as obvious.
- The short day. In and out, counted in full. On this record one day of the thirty-four was four hours long and counted identically to a nine-hour one.
- The displaced day. A day taken in the building because the count needs it, rather than because the work does. Five days of this quarter were of this kind and four of them produced nothing.
- The convenient day. Whichever weekday is easiest, which is how the thin end of the week gets attended at all.
- The defensive day. The day taken because not taking it might be noticed. The clearest example in the record is a two-hour-forty journey through engineering works to a day that produced nothing.
None of those are cheating. Each is a rational response to being asked for a quantity rather than an outcome, by people who are otherwise doing the job properly.
Why the data gets used anyway
Because the alternative is to ask people, and asking produces answers that are soft, slow and arguable, while the badge system produces a percentage by Tuesday.
There is also a reasonable version of this: if a policy exists and is widely ignored, that is worth knowing, and attendance data is the only way to find out. The failure is not in collecting it; it is in treating it as a measure of whether the arrangement is working.
The one thing entry data could usefully show
Not compliance. Co-presence.
The same records that show who arrived could show, without naming anybody, how often two people from the same team were in on the same day. That is the number the policy is actually about, it is derivable from data already held, and it is almost never computed.
A team that found its co-presence rate was one day in five would have learned something it could act on. A team told its attendance was ninety-three per cent has learned nothing, and will be asked to improve it.
The practical caution
If attendance is being counted where you are, the useful stance is not evasion. It is to make sure the second number gets into the room alongside the first — because the counted number will be in the room regardless, and on its own it points at exactly one conclusion.
What a swipe does and does not establish
| The record shows | The inference usually drawn | Whether it follows |
|---|---|---|
| A card at a door at 09:12 | A day worked in the office | No |
| Three swipes in a week | Compliance with a three-day rule | Only in the sense the rule means |
| Rising swipe counts | The policy is working | No |
| Falling swipe counts | People are disengaging | No |
| Swipes concentrated midweek | Preference | Partly; mostly coordination |
The last row is the most useful thing in the data and the least examined. A sharp midweek concentration is a signature of uncoordinated choice under a count, and it is visible in the data every organisation already holds.
The measure that is derivable and never derived
Co-presence: how often two people from the same team were in on the same day.
It needs no new collection, no individual reporting and no names — a team-level rate is enough, and a team-level rate is not a measure of any individual, which is what makes it safe to compute.
The difference is what each figure prompts. One of them names a mechanism somebody can change on Friday afternoon. The other names a percentage, and the only thing to do with a percentage is move it.
If attendance is being enforced where you are
Three practical notes, none of which are about evasion.
Find out what is actually counted. It is usually entry, not duration, and it is usually a week rather than a day. Knowing which matters, because people frequently comply with a stricter rule than exists.
Assume the number will be in the room. It will, and arguing that it should not be is a losing position. The available move is to bring the second number in beside it.
Do not let the measure choose the days. The commonest cost of enforcement is not the enforcement; it is that people start picking days by what is easiest to comply with, which is how the thin end of the week gets attended at all.