the open model · units
What a point is, and what it is not
This page exists because the app withholds something people reasonably want — a total for their own week — and the reason is a judgement rather than an oversight.
Two components, and only one of them is a duration
Execution is hours-equivalent. One unit of it is one hour a week of neutral, unpleasant-free work, marked up by at most 1.55× for effort and for how badly the task breaks up a day. That can be checked against a clock, and it is.
Cognition is a dimensionless weight. Not minutes, not hours, not convertible to either. Carrying three things at once is possible in a way that doing three things at once is not, which is the whole reason the model keeps them apart.
Add them and you get an index. It orders and compares. It does not measure. There is no honest sentence of the form your week came to this many units of load, and the app does not offer one.
The sentence we got wrong, and how
An early draft of the model said one point is about an hour a week of neutral work, and offered 40–90 points as a sane range for a family with young children. That was false and it was withdrawn.
Measured against the finished library, the same household came back at 536. The failure was structural rather than arithmetic: the cognitive half is computed per row and does not scale with the size of a row, so a household's cognitive total is approximately its average row weight times the number of rows it has. On the library of the day that was 0.937 × 374 = 350.4 against a measured 350.6.
The cognitive total was a count of library rows wearing the costume of a measurement. It is the most instructive mistake in this project's history, and it is why the rule below exists.
So the app never prints a total
No screen in Champignon quotes a points total, a per-row score, or a category score. What it may say is shares, differences between shares, and counts.
What survives the problem above is exactly the thing the app shows, and this was tested rather than assumed: a ratio is stable under the thing that inflates the total. Pruning one test household's list from 287 rows to 40 moved her share from 72.0% to 75.9%; twelve random halvings of the list gave 71.1–74.5%. The split and the gap survive an inflated denominator. The unit does not.
The trade, stated plainly
What you lose. You cannot open the app and see how your own hours add up. That is a fair thing to want, and for the execution half — which genuinely is in hours — a total would even be roughly meaningful.
Why it is withheld anyway. Three reasons, in order of force.
- It could not be right to the hour, and it would look as though it were. The minutes in the library are the library's estimates of marginal time — what a row costs that its neighbours don't already claim — not a measurement of your Tuesday. A figure printed to the minute claims a precision the instrument does not have and cannot get.
- Half the total is not in any unit at all. Showing an execution total and withholding a cognitive one would quietly say that the countable half is the real half, which is the exact error the product exists to correct.
- A total is the most weaponisable number in the app. Two people are looking at this screen together and one of them is usually already on the defensive. A share invites what would move this; a total invites look at my number.
What we are not claiming. That withholding it is costless, or that everyone will find the trade convincing. It is a judgement, and like the others it is published as one. If the beta says people cannot orient themselves without some sense of scale, the answer is more likely to be a better sense of proportion between rows than a grand total.
Where the weights are, if you want to check them
Every task in the library shows its frequency, its marginal minutes, its three dimensions and its computed execution and cognitive figures. Those are the instrument's calibration, published so it can be argued with — a task's weight is not a person's score, and no page on this site totals a household.