Taking things out
Mark the quiet ones
Two questions, five seconds and an hour respectively, that between them address all four of the unremarked problems on this record.
The standing item for any other business
week 8 · still not missed
A written agenda for every one-to-one
week 6 · still not missed
The suggestion box
week 6 · still not missed
Weekly timesheets broken down by task
week 3 · still not missed
Not missed 4
Four removals were written down with the week each of them stopped. Four of them have not been asked for in the 78 weeks since. Four of the fourteen nobody has asked for, seventy-eight weeks on.
The removals that cost something on this record were the ones nobody discussed. That is not useful unless they can be identified at the time, and they can be, by two questions that require no judgement about whether any removal is wise.
Before a removal is agreed
"Who here has done this?"
If nobody present performs the procedure, the silence in the room contains no information about it.
The removal is not blocked. It goes on the list with a mark, and somebody asks a person who does it before the week is out. The question being asked is not "should we?" but "what does this actually involve?", which is a factual question with a factual answer.
On this record four removals were agreed in rooms containing nobody who performed them. Two of those four became problems — the standing slot and the reconciliation — and in both cases the missing sentence was a matter of fact rather than of opinion:
| removal | the sentence that was not said |
|---|---|
| the standing slot | "that hour is where ownerless jobs get picked up" |
| the reconciliation | "that is where the numbers get built, not reported" |
Either would have taken five seconds from somebody who did the work.
A month after the transition
"What did you used to do on the day that went?"
This catches the removals nobody proposed, which are invisible to every list and which produced the two slowest failures on this record.
Seven things stopped because the hours they lived in stopped existing — four of them in a Friday afternoon. Nobody decided any of them. They appear on the list only because somebody added them retrospectively, four of them more than a month late.
Asked of each team, this takes about an hour in total and produced five of the seven. The two it did not produce are the shadowing of new starters and the log of refused jobs, at lags of forty-seven and sixty-four weeks.
A narrower companion version, asked of anybody who joined in the last year: who showed you how to do this? If the answer is a procedure that no longer runs, a knowledge-carrying removal has been found that nobody logged.
What does not work
"Is anything missing?" Asked generally, it routes to people doing their current work, and the honest answer from all of them is no. The dangerous removals are precisely the ones that inconvenience nobody today, so the question is addressed to the one population guaranteed to have nothing to report.
Asking people to predict which removals will matter. Tried informally here after the fact: shown the list without outcomes, people picked the contested ones. That is the error restated, and knowing about the error does not protect against it.
Keeping the quiet ones. Thirteen of the seventeen were fine and several are the clearest savings on the list. Reinstating to be safe discards the main benefit to guard against four.
A longer meeting. The filter is the meeting itself. More of it selects harder for the same subset — the removals with a name, an obvious purpose and a visible owner, which are the ones that would have reported themselves anyway.
The quarterly form
After the transition, this reduces to one page read twice a year:
- Entries marked nothing that have produced no event.
- Against each, the next date on which its absence would be visible.
- Blanks in column two, at the top.
At six months on this record that page would have held about eleven lines, three of them future problems, one with a blank beside it. The blank is the entry that took sixty-four weeks to report itself, and a blank in that column is the clearest warning this method produces.
Why there is no system here
A column and two questions. No process, no template, no owner beyond a name, no software.
That is deliberate. Everything demonstrated on this record was produced by one person keeping a spreadsheet, and every elaboration would have made it less likely to survive the end of the project. The failure mode to design against is not inadequacy; it is abandonment.
A practice that takes twenty minutes twice a year has some chance of still happening at eighteen months, which is when it finally has something to say.
When the second question stops working
The retrospective question — what did you used to do on the day that went — has a shelf life, and this record suggests roughly what it is.
Asked informally at about a month, it produced five of the seven unproposed removals immediately. The two it missed are the two slowest failures, and both were things nobody thought of as activities: a week of sitting alongside somebody, and a few lines written down on a Friday.
Asked again at the eighteen-month read-back, the same question produced nothing new. By then the shape of the week had changed enough that people could not recall what the old one had contained, and the things that had quietly resumed were indistinguishable from the things that had never stopped.
So it is a question with a window. Early enough and it recovers most of the category; late enough and it recovers nothing, because the comparison it depends on has gone.
A month is about right on this evidence. Earlier than that and the unproposed removals have not yet been noticed as absences by anybody.