Nobody stopped doing the work

The difference matters more than it sounds

There is a comfortable reading of this and a less comfortable one, and both are supported by the record.

The comfortable reading is that organisations accumulate a great deal of apparatus, that the apparatus is not the work, and that a forced subtraction finds this out fast. Fourteen things went and nobody has asked for them. That is a real finding and it is a third of the list.

The less comfortable reading is that apparatus is not decoration. Most of it exists because somebody was once caught out. A second reviewer on routine changes is there because an unreviewed change once went wrong. A log of refused jobs is there because the same job kept coming back. Removing the apparatus does not remove the risk it was built against; it removes the handling of the risk and leaves the risk where it was.

That is why the consequences of this kind of removal are slow. Nothing breaks when you stop a precaution. Things break later, when the thing the precaution was for happens, and whether that is three weeks or three years is a property of the risk and not of the removal.

Four things the apparatus was doing

Reading the thirty-eight entries against what each had been introduced for, the apparatus on this list was doing four distinguishable jobs:

• Catching errors. Second reviewers, duplicate records, counts, reconciliations. Eleven entries.
• Carrying knowledge sideways. Handover notes, induction folders, logs of decisions, shadowing. Nine entries.
• Keeping attention on something. Standing slots, circulations, reminders, planners. Twelve entries.
• Marking an occasion. Away days, Thursday biscuits, the suggestion box. Six entries.

The four groups behave differently, and if there is a usable rule anywhere in this record it is in the second group.

The knowledge-carrying ones are the dangerous ones

Of the five removals that surfaced as problems, four were from the second group. The reconciliation that nobody could reassemble. The shadowing that stopped and took three new starters with it. The contract reading. The log of refusals, where the same reason was worked out from scratch four times because nobody had written it down the first time.

This is not a coincidence and the mechanism is not subtle. Something that catches errors fails visibly: the error gets through and you see it. Something that keeps attention on a thing fails visibly too: the thing stops being attended to and somebody notices.

Something that carries knowledge sideways fails invisibly, because the failure is an absence in somebody's head, and nobody can see an absence. It shows up when a person who would have known is asked and does not know, and that depends entirely on when somebody happens to ask.

The failure mode of a knowledge-carrying procedure is a question nobody asks for a year.

What this does not establish

It does not establish that knowledge-carrying apparatus should be kept and the rest removed. Three of the nine knowledge-carrying entries are among the fourteen that nobody has missed, including two logs that appear to have recorded things nobody ever read back.

Nor does it establish that the error-catching apparatus is safe to remove. One of the five that surfaced was the quarterly reconciliation, which is squarely in the first group. Its failure was visible in the ordinary way — an auditor asked — but what made it expensive was that nobody had looked for eight months.

What it does suggest is where to point an hour of attention. If you are going to go and ask about a handful of removals rather than all of them, the ones that used to move knowledge between people are the ones where asking is the only way to find out.

The sentence to be careful with

"The work carried on exactly as before" is true of this record and it is the single most misleading sentence anybody could take from it.

The work carried on. The conditions under which the work would carry on next year were quietly altered, in thirty-eight places at once, and whether that mattered took between one week and sixty-four weeks to find out. Both halves of that are in the same record, and a report that gives only the first half is not wrong so much as unfinished.

Where the hours actually came from

Worth putting beside the apparatus argument, because the two are easily confused.

Thirty-eight removals did not supply eight hours a week each. Most supplied minutes. The deck was four hours a month for one person. The second quote on small purchases was perhaps ten minutes a week across three people. The wall planner was twenty minutes on a Monday.

Added together the thirty-eight come to something between five and six hours a week per person on the team's own estimate, which is most but not all of the reduction. The remainder came from the three things that could not stop being handled by fewer people, and from a general compression that nobody logged and nobody can account for.

That last part is the honest gap in this record. Somewhere between one and three hours a week per person went somewhere that no entry describes, and the most likely answer — that the remaining work tightened up — is also the one nobody can check.