Why eleven weeks matters

What compression does to visibility

A procedure removed on its own is a local event. One team notices, argues about it for a day or does not, and the matter closes. There is nobody for whom the removal is one of a set, because there is no set.

Thirty-eight removals in a quarter create a set whether anybody intends it or not. That has three consequences:

1. There is something to write down. A list of one is not a list. A list of thirty-eight feels like a document, and somebody keeps it.
2. There is something to compare. The only reason it is possible to say that quiet removals surfaced later than loud ones is that both kinds happened in the same quarter to the same organisation.
3. There is a single date to count from. Every lag on this site is measured from a week inside one eleven-week window, which is why the lags can be put beside each other at all.

None of that is available to an organisation that removes a procedure every few months. The information exists; it is spread so thin that nobody can see it.

What compression does to risk

The same density that makes the record readable also concentrates the exposure, and that is the uncomfortable half.

If a removal has, say, a one-in-eight chance of turning out to matter, then removing one procedure a quarter means being wrong about one thing every two years — comfortably inside the organisation's ability to notice and fix. Removing thirty-eight at once means being wrong about four or five things simultaneously, with the consequences arriving at unpredictable intervals over the following year and a half.

On this record that is precisely what happened. Five surfaced, at weeks nineteen, twenty-four, thirty-four, fifty-three and seventy-three from the start. They arrived singly and they were handled singly, and nobody connected the one in week seventy-three to the transition at all until the list was read back.

The removals were simultaneous. The consequences were not, and that asymmetry is the reason the transition looked finished long before it was.

The window is the wrong length for the review

Eleven weeks is also, awkwardly, about the length of a reporting period.

A transition that takes a quarter invites a review at the end of the quarter, and a review at the end of the quarter sees: the removals done, the work still going out, nobody complaining much. Of the twenty-four entries that eventually produced an event, nine had produced it by week eleven. Fifteen had not.

The transition is therefore over, in the only sense visible to anybody managing it, while roughly two thirds of what it did has not yet happened.

This is not a criticism of anyone's diligence. It is a structural mismatch between the length of a change and the length of its consequences, and nothing in the change itself signals the mismatch. The last thing a transition does is finish looking like a transition.

What a slower transition would have cost

It is tempting to conclude that the removals should have been spread out. The record does not support that, for two reasons, and they pull in different directions.

Against spreading: a shorter week is not a sum of independent removals. The hours have to come from somewhere by a particular date, and removing six things a quarter for two years means running a five-day week badly for two years. Several people on this record said afterwards that the compression was the only part that had worked without argument, precisely because everybody understood that something had to go.

For spreading: the four knowledge-carrying removals that surfaced late were four separate, slow failures that happened to overlap. Spread across two years they would still have been slow, but they would have arrived one at a time into an organisation that was not simultaneously absorbing thirty-seven other changes.

The honest position is that compression buys clarity and sells margin. It makes the transition legible, the list possible and the finding visible, and it concentrates four unpredictable failures into a single eighteen-month window.

What follows

If the removals are going to be compressed — and on this record they had to be — then the thing that must not be compressed is the looking. Eleven weeks of removal wants seventy-eight weeks of attention, and the attention is cheap: it is a list, read back twice a year, by somebody asking who would have been the first to know.

What the people involved said about the pace

Asked afterwards, the view on compression was close to unanimous and it did not match the conclusion above.

Nobody wanted it slower. The reason given, repeatedly, was that a compressed transition made the trade visible: everybody could see that the hours had to come from somewhere, which made individual removals arguable on their merits rather than as encroachments.

Several people contrasted it with an earlier efficiency exercise at the same organisation, which had removed things gradually over about a year and had been, in the words of one account, resented throughout — because each removal arrived on its own, with no visible reason, and looked like something being taken away rather than something being traded for a day.

That is a point about consent rather than about risk, and it does not bear on the distribution of lags at all. It is included because it is the clearest thing the record contains about why compression happened, and because it suggests the choice is not really between fast and slow but between a trade people can see and one they cannot.