One school morning, modelled

41 absences. One of them is not a child who stayed home.

Every tile is one family the office has to telephone before the register closes. From a desk they are identical, and they stay identical after the calls are made, because a call that reached somebody and learned nothing files itself next to the ones that worked. The column is the one this software refuses to lay flat.

Where 41 comes from. 600 pupils at an absence rate of 6.9 per cent is 600 × 0.069 = 41.4, so 41. The rate is the Department for Education’s measured figure for every school in England across the full 2024/25 academic year (Pupil attendance in schools). The school size is a premise, not a measurement, and it is printed here so it can be argued with. So is the window: the board lays the 41 calls evenly across an office morning from 08:30 to 09:20, which is where the time on each tile comes from when you point at it.

What this board does not claim. It puts no measured outcome on any tile. The only thing it says about a call is the split it draws, forty that close and one that does not, and the clock face a tile reports is its slot in the window above and not a time anybody dialled. This software has placed 20 real calls through CALL-E. A run built out of matched hard cases is not a sample of a real morning, and multiplying its proportions up to a school would state something nobody measured. The single open column is the shape of the problem, not a rate.

Back to the two calls, and the recordings

Whether the morning finishes before the register closes

At the default setting, it does not.

A cutoff is a wall, not a target. Every bar that goes through the plane is a morning that ended with families still unreached. The default cap of three misses by 72 minutes. Concurrency 6 is the first that fits, and it is one flag.

One bar a concurrency setting, from 3 at the tall end to 25 at the short end. The plane is the 75 minute cutoff; the two that go through it are 3 and 4.

Measured, not modelled. 11 real calls, timed from the turn offsets CALL-E itself returned, give a mean of 51.0 seconds. One 2 second poll interval a call makes 53.0 seconds a worker a call, and the rest is division: 500 absences against a 75 minute window. This drawing calls tools/throughput.py at build time rather than repeating its arithmetic, so it cannot disagree with the tool.

What the flag costs. The cap is the only brake there is, because a call that has been accepted cannot be recalled. Raising it to 6 means that many families dialled at once with no way to stop any of them. That belongs in a decision a district makes on purpose, which is why it is drawn here instead of left in a default.