What to tell people before you start

"Something has to stop, and we are going to write down what"

The first sentence does two jobs.

It sets the expectation that a shorter week is a subtraction rather than an efficiency, which on this record was the single thing everybody understood and agreed on, and the reason the compression into eleven weeks went through without argument.

And it makes the list a normal object rather than a bureaucratic imposition. A list introduced at the start is how the thing is done. A list introduced in month three is somebody checking up.

"If you think a removal is a mistake, say so, and we will ask you

again later"

The second half is the part that is usually missing.

Objections on this record were the best-behaved signal available: nine of them, all reporting inside twelve weeks, four producing fast corrections and two producing versions better than anything argued for in the room.

That only works if objecting is worth something. An objector watches and reports because reporting has a point. Where losing an argument is remembered, people stop making them, and the nine fast reports become nine silences — which converts the best-behaved category on the record into the worst.

The explicit form, said when an objection is overruled: you may well be right; if you are, what will we see and when? And then asking them at that time.

"Nobody is being measured on how many things stay removed"

This is the one that protects the request channel.

Eight people asked for things back on this record, at a median of three weeks, and none of them was treated badly for it. Four of the eight came back smaller than they went.

A request is the cheapest way an organisation can find out that a removal was wrong: fastest, clearest, and arriving with a diagnosis attached. It is also trivially easy to suppress — it is enough that asking be slightly awkward, that the transition be reported as a success, or that whoever asks be the person who could not cope.

Suppressing it improves the headline number immediately and converts the cheapest signal available into the most expensive one.

"We will look again in a year, and some of this will turn out to be

wrong"

Said at the start, this is unremarkable. Said at month twelve it sounds like an admission.

It does three things:

• It makes the twelve-month read-back a planned event rather than somebody reopening a closed matter.
• It sets the expectation that some removals will be mistakes, which on this record was five of thirty-eight, so that the first one does not read as a failure of the whole thing.
• It gives whoever holds the list a reason to still be holding it.

What not to say

"Nothing will change about the work." It will. On this record the work itself was untouched and thirty-eight things around it stopped, and several people who were told nothing would change experienced the second part as a contradiction of the first.

"We will put back anything that turns out to be needed." Mostly true and quietly wrong in the way that matters. Fourteen entries were never asked for, eleven came back without a decision, and five produced problems that took between fourteen and sixty-four weeks to surface. "Anything that turns out to be needed" promises a detection system the organisation does not have.

"This is about working smarter." On this record it was about working less, and the hours were found by stopping thirty-eight things. The euphemism costs credibility at exactly the moment the list is being introduced, and the list is the thing that has to survive.

The shortest version

We are going to stop doing a number of things. We will write down what, and whether anybody objected. If you think one is a mistake, say so — we will ask you again in a few weeks. If you want something back, ask; nobody is counting. And we will read the list again in a year, when some of these will have turned out to be wrong.

Five sentences. Everything this record produced depends on some part of them being true.

What was actually said on this record

For comparison, since the four statements above are a reconstruction of what would have helped rather than a transcript.

The first was said, more or less in those terms, and it was the thing everybody remembers. The framing of a shorter week as a trade — a day for a list of removals — is credited by several people with the absence of argument about the compression.

The second was not said. Objections were heard and answered on the merits, which is better than most places manage, but nobody was told they would be asked again, and nobody was asked again. The four objectors who turned out to be right volunteered it.

The third was not said either, and did not need to be: the culture happened to be one where asking for something back was unremarkable. That is luck rather than design, and a transition relying on it in a different place would have had eight silences instead of eight requests.

The fourth was not said, no read-back was planned, and the eighteen months of outcome data on this site exists because one person kept a file.

Three of the four conditions that make this record possible were accidents.