Who else
Attendance and overlap are different numbers
Every system counts the first. Only the second buys anything. The gap between them is where a hybrid policy quietly fails.
Days this part is about, hour by hour
Thursday, the third week
0 of 7Two floors from anyone known
Wednesday, the first week
2 of 7Thursday, the first week
0 of 7Nobody from the team was in
three days · 21 hours logged · two of them were only possible there · two of the three days contained no such hour at all · three were made worse by being there
Three days chosen for the gap between being in and being together.
For teams moving between home and office, it helps to visit the Monitask product site and compare time, attendance, and work patterns without treating presence as the result itself.
For further reading outside this collection, see material published by CIPD.
A building can tell you how many people came in. It cannot tell you whether any two of them were the two who needed each other, and that second fact is the one the whole arrangement rests on.
The two numbers
Attendance. How many people were in. Countable from a badge system without anybody designing anything. Reportable, comparable, and the basis of nearly every hybrid policy review ever conducted.
Overlap. How many of the people you actually work with were in on the same day as you. Not counted anywhere, not reportable, and the only one of the two that predicts anything.
On the record here, attendance was high: thirty-four days of a required thirty-nine, which any reporting system would show as ninety-three per cent — excellent compliance.
Overlap was poor. The median number of the person's own team of nine who were in on the same day was three. On the worst nine days it was zero, and five of those were days when the building as a whole was perfectly busy.
Why high attendance can produce low overlap
Because attendance is a sum and overlap is an intersection, and the two behave completely differently when the days are chosen independently.
Nine people each in three days of five produces a building that is on average fifty-four per cent full every day — a healthy-looking number. It also produces, if the days are chosen independently, a meaningful chance on any given day that the two specific people who needed to talk are not both there.
In practice the choices are not independent, which helps: convergence on Tuesday and Wednesday raises overlap on those days sharply. It also guarantees that overlap on the other three days collapses to nearly nothing, which is where the fifteen empty days came from.
What this does to policy reviews
It makes them systematically wrong, in a predictable direction.
A review looks at attendance, finds it healthy or finds it short, and adjusts the number of days. Neither move touches overlap. Raising the requirement to four days adds a journey and adds attendance; whether it adds overlap depends entirely on whether the added days coincide with anybody, which is not a thing the rule governs.
So the lever being pulled is not connected to the outcome being sought, and the review concludes with a number that will be revisited in six months.
Measuring the one that matters
It does not need a system. It needs one question asked of each person once a quarter:
On your office days last month, how many of the people you work with most closely were there?
A median of three out of nine means something. A median of zero means the arrangement is producing attendance and nothing else, and it means it regardless of how good the attendance figure looks.
Why the two numbers diverge so sharply
A small worked example, because the arithmetic is counter-intuitive enough that it is worth seeing.
Nine people, three days each out of five, chosen independently. On any given day the building is just over half full, which looks healthy. But for any specific pair of people, the chance of being in on the same day is a good deal lower than that — and what the office is for is specific pairs, not averages.
Convergence improves this sharply on the converged days and destroys it everywhere else. On this record the two effects produced a median overlap of three of nine and nine days of zero, under attendance that every system recorded as excellent.
The question that produces the second number
It takes one line in a quarterly conversation and no system at all:
On your office days last month, how many of the people you work with most closely were there?
The answer is a small integer and people can give it immediately. It does not require anybody to be tracked, it is not comparable between people in a way that invites ranking, and it measures the thing the policy exists for.
What a low answer should trigger
Not more days. The sequence that follows from a low overlap figure runs in a different direction entirely:
- Name a day. The direct fix, and the only one that addresses the mechanism.
- Share plans weekly. Smaller, faster, needs nobody's approval.
- Sit together when in. Recovers the short-range part of the effect, which is most of it.
- Put something on the named day. So that it holds.
All four are available at team level and none of them requires the count to change. The reason they are rarely tried is that the figure that would prompt them is never collected.