The hours
Why the good hours cluster
Seventeen of the thirty-one hours that needed the building fell on one weekday. That is not a coincidence and it is not about the weekday.
Days this part is about, hour by hour
Wednesday, the second week
3 of 7Planning day, the densest of the quarter
Thursday, the third week
0 of 7Seven hours of solo work at a hot desk
Wednesday, the fourth week
3 of 7The new starter's first day in the building
three days · 21 hours logged · six of them were only possible there · one of the three days contained no such hour at all · two were made worse by being there
Three of the twelve days drawn in full on this site.
Where work is split across locations, managers can see the official product overview to make project effort and recurring bottlenecks clear enough for a transparent capacity discussion.
An independent source worth checking for this subject is SANS Institute.
The whole quarter, by the day of the week it landed on
34 days in the building across five weekdays · 23 of them on just two of the days, Wednesday and Tuesday · the hours that needed the building fell hardest on Wednesday, seventeen of them · on Monday and Friday nothing at all required being there
Thirteen weeks under a three-day policy that named no days.
Thirty-one hours in the quarter could only have happened in the building. Seventeen of them were on a Wednesday. Eleven were on a Tuesday. Three were on a Thursday. None at all were on a Monday or a Friday.
| Weekday | Days in | Hours logged | Hours that needed the building |
|---|---|---|---|
| Monday | 2 | 14 | 0 |
| Tuesday | 11 | 77 | 11 |
| Wednesday | 12 | 84 | 17 |
| Thursday | 8 | 56 | 3 |
| Friday | 1 | 7 | 0 |
If the hours were distributed by chance across the days actually attended, Wednesday would have had about eleven. It had seventeen. The clustering is real and it has a dull, mechanical explanation.
It is not about Wednesday
Nothing about the middle of the week produces collaboration. The clustering is about who else was there, and the rest of the team had converged on Tuesday and Wednesday for exactly the same reason this person had: those were the days that felt least likely to be wasted.
The feedback loop writes itself. Wednesday is slightly fuller, so Wednesday is slightly better, so more people pick Wednesday, so Wednesday is fuller still. Three weeks of that and the week has a shape nobody chose and everybody is now bound by.
What happens at the thin end
Monday and Friday are the same mechanism run backwards.
On this record Monday was attended twice in thirteen weeks and Friday once, and between them they produced zero hours that needed the building. Not because those days are inherently worse, but because nobody else was there — and the first kind of hour requires a second person.
That is the part worth holding on to. An empty day in the building is not a day with less collaboration in it. It is a day with none, structurally, regardless of how willing you are.
Why the policy makes this worse rather than better
A three-day rule that names no days looks like flexibility and acts like a coordination problem with no coordinator.
Each person picks independently, each one reasons correctly, and the aggregate is two days at ninety per cent occupancy and three days at twenty. The building is simultaneously too full and too empty, the rooms run out on exactly the days worth being there, and the person who picks Thursday for sensible personal reasons gets a quarter of the value for the same journey.
Nobody designed that. It is what independent sensible choices produce when the thing being chosen has increasing returns to agreement.
The one-line version
The valuable hour needs somebody else in the room. A policy that counts days does not put anybody in the room with you. A policy that names days does, and that is the whole of the difference between the two.
What the clustering does to everybody's incentives
Once the pattern exists it stops being a preference and becomes a constraint, and it does so for three separate reasons that reinforce each other.
The meetings move. Recurring things get scheduled where people are. By week eight, four of the team's five recurring meetings were on Tuesday or Wednesday, which means being elsewhere on those days now has a cost beyond missing people.
The social proof. Being in on a full day is unremarkable. Being in on a Thursday when four people are on the floor is visible in a different way, and several people reported a mild sense of having got the day wrong.
The thin day is genuinely worse. This is the real one. It is not a perception: on this record the thin days produced nothing, repeatedly, and anybody who tried them a few times learned to stop.
The thing the clustering hides
A clustered week looks, from above, like a week that is working. The building is full on two days, there is visible activity, and attendance figures are healthy.
What it hides is that the same policy is producing two completely different experiences depending on which days a person can make, and the difference is invisible because both people are recorded as having done three days.
| Converged days | Thin days | |
|---|---|---|
| Colleagues present | Most of the team | Nobody from the team |
| Hours that needed the building | 28 of 31 | 3 of 31 |
| Rooms obtainable | Rarely | Always |
| Desk near the team | Rarely | Irrelevant |
| Recorded as compliant | Yes | Yes |
The last row is why nothing in the system ever notices.