Nobody missed it yet

The part of the argument nobody runs

Almost every published account of a four-day week reports the same two things: output held, and people were happier. The accounts disagree about almost nothing. That unanimity is the first thing worth being suspicious of, and the suspicion is not that anybody is lying.

A shorter week is not a change to how hard people work. It is a subtraction. Four days contain fewer hours than five, and the hours that leave take work with them. Something has to stop. Every successful transition is therefore a list of things that stopped, and every argument about whether it worked is an argument about that list — except that the list is almost never written down, and when it is, it is almost never looked at again.

This site is built on one that was. One organisation of about ninety people moved to four days over eleven weeks. As the transition ran, somebody wrote down each thing that stopped and the week it stopped: thirty-eight entries. Then — and this is the unusual part — the list was kept, and for the next seventy-eight weeks each entry was marked up with what had become of it.

Four things can happen to a removal

Nobody asks for it back. Fourteen of the thirty-eight. Nobody requested it, nobody reinstated it, nothing went wrong that anybody traced to its absence. This is the strongest evidence a shorter week has, and it is more than a third of the list. A substantial proportion of what an organisation does is done because it was being done.

It creeps back. Eleven. Nobody decided anything. The thing reappeared, usually in a slightly different form, at a median of seven weeks. Nobody noticed it returning, because nothing announced it.

Somebody asks for it back. Eight, at a median of three weeks. These are the easy ones. A request is a signal, it arrives fast, and it can be argued with.

Its absence shows up as a problem. Five, at a median of thirty-one weeks, and the slowest at sixty-four.

The four outcomes are not four degrees of the same thing. The first three are all forms of information arriving. The fourth is information failing to arrive until the cost has already been paid.

The finding

Each entry on the list also records how much noise the removal made at the time it happened: whether somebody argued against it, asked about it without objecting, or said nothing at all.

Of the nine removals somebody argued about, none later surfaced as a problem. Of the seventeen nobody remarked on, four did.

Put plainly: the things that turned out to matter were, at the time, the things nobody thought were worth discussing.

This is not a paradox and it is not a coincidence. A contested removal gets watched. Somebody objected, which means somebody has a stake in its absence being noticed, and when it starts to hurt they are the first to say so and they are believed, because they said it would. An unremarked removal has no such person. It stops, nothing happens, and the fact that nothing happens is taken as evidence rather than as the absence of a reporter.

The lags say the same thing a second way, and they line up in order. Where a removal was argued about, the median wait before anything happened was four and a half weeks. Where it was questioned but not opposed, six and a half. Where nobody said anything at all, eighteen. The quieter the removal, the longer the organisation took to learn anything about it.

The corollary is uncomfortable for both sides of the argument. Advocates get told that the quiet transitions are the risky ones. Sceptics get told that the loud objections are reliably the wrong thing to worry about.

Why the lag is the whole problem

Consider what a lag of thirty-one weeks does to an evaluation.

A standard pilot runs six months. Counting from the day the transition started, twenty-six weeks is up before three of the five problems have arrived: on this record they surfaced in weeks thirty-four, fifty-three and seventy-three, long after such a pilot would have reported, the report been written and the policy made permanent.

What a six-month trial would have seen is this. Eight removals asked back, nine crept back, two problems, and nineteen entries on which nothing whatever had happened. Read at week twenty-six, nineteen quiet entries look like nineteen things the organisation did not need. Five of them were not.

This is why the published accounts agree. They are not wrong. They are measuring an interval shorter than the thing they are measuring. Any evaluation window shorter than the longest lag in the system returns a verdict that is correct about the window and silent about everything after it, and nothing in the result itself tells you which you are looking at.

What came back, and how much of it

Twenty-four of the thirty-eight came back in some form, and the obvious reading is that the real saving is fourteen out of thirty-eight. That reading is wrong, and the record says so.

Of the twenty-four, six came back as what they had been. Twelve came back smaller — fewer people, less often, shorter. Six came back as something else doing the same job. Against the load each had carried before, the median return was forty-seven and a half per cent.

So the list understates the reduction. Most of what came back came back at under half its old weight, and the half that did not return is a real and permanent saving. A site whose name is Nobody Missed It Yet is obliged to say that the number cutting against its own title is the honest one, and that is this number.

It also sharpens the asymmetry. What comes back comes back cheaper. What surfaces as a problem surfaces at full price, and late.

What this does not show

It does not show that four-day weeks work, and it does not show that they fail. It is one organisation, one transition, one list kept by one person, and the distribution of lags on another list would be different. The method travels; the numbers do not.

It does not show that long lags are peculiar to shorter weeks. Any removal anywhere has this property. A four-day week simply forces thirty-eight of them at once, inside eleven weeks, which is what makes the distribution visible.

It does not identify which removal is which in advance. That is the point. If it were possible to tell, from the way a removal looks on the day it happens, whether it will be missed in fourteen months, none of this would be difficult. The record here says the obvious signal — how much fuss it caused — points the wrong way.

And it does not say that late-surfacing removals should not be made. Four of the five that surfaced were, on the record, worth the trouble they caused. The argument is about the evaluation, not the removal.

What is worth doing

Three things, and none of them is an argument for or against a shorter week.

Write the list. Not the policy, not the business case: the removals, one line each, with the week. It costs an hour. Without it there is nothing to mark up later, and the question of what the shorter week actually did becomes a matter of recollection.

Mark the quiet ones. The entries nobody said anything about are the ones with no reporter. They are the ones to go and ask about, because nobody will tell you.

Fix the review window to the thing, not to the quarter. A six-month review of a change whose slowest consequence takes fourteen months is not a short review, it is a different review. If you cannot run a longer one, the honest report says what interval it covers.

The order of the parts

There are six, and they follow the list rather than an argument.

Start with what a shorter week takes out, which is the thirty-eight themselves and what sort of things they were. Then the lag: the gap between stopping something and learning anything about it. Then the quiet removals, where the finding sits. Then what returned, and at what weight. Then the other arrangements — compressed hours, a nine-day fortnight, staggered cover, annualised time — and the different set of removals each one forces. Last, how to take something out in a way that tells you early whether you were wrong.

Nothing on any of those pages comes from outside the list. Thirty-eight rows, one organisation, nobody identified anywhere, and that will not change.