Working to rule

Why does a workplace grind to a halt when everyone follows the written rules exactly?

In a work-to-rule, nobody stops working. Nobody walks out, nobody goes slow in any way they could be disciplined for, and nobody breaks a single instruction. Everyone simply does exactly what the written procedures say, in the order they say it. And output falls, sometimes by half, sometimes to nothing. That is very strange. If the rulebook describes the job, obeying it perfectly should produce the job — perhaps a little slower, but not a stoppage. So what does the slowdown reveal about the rules?

Start with what a rulebook actually is. It is an attempt to compress a job into text, written in advance, by people who cannot enumerate every situation the job will meet. Ask what such a document can contain, and the answer is: the cases someone thought of, described in the vocabulary available when it was written. Everything else is either absent, or covered by a rule aimed at a different case.

But the unanticipated cases arrive every day, and the work still gets done. So something must be handling them. What? Only the people present, deciding: that this check is pointless on this machine, that the form can be filled after the job rather than before, that the sensible thing is to telephone the next shift rather than raise a ticket. That work is invisible for a precise reason — it succeeds. Nothing is recorded, because nothing went wrong.

There is a second source of the same slack, and it comes from how rulebooks grow. Each incident adds a rule; almost nothing is ever taken out. Over years the set accumulates instructions written for equipment long since replaced, and pairs that genuinely conflict. At that point "follow all the rules exactly" is not merely slow, it is occasionally impossible — and the daily act of choosing which rule governs is itself part of the discretion being withdrawn.

A third contribution is easy to miss. Rules describe tasks one at a time, but a working day is tasks overlapping. Much of the real effort goes into coordination the procedure never mentions: checking that the next stage is ready before you finish, warning someone downstream, holding a job back so two can go together. Follow each procedure strictly and in sequence, and the overlap disappears — the work becomes correct and serial, and serial is slow.

Put it together and the tactic stops looking paradoxical. A work-to-rule withdraws the unwritten contribution while leaving the written one perfectly intact. The size of the collapse is therefore a measurement — of how much of normal operation was resting on judgement and effort that the rulebook never captured and the organisation never paid for.

Which explains the surprise. Because that contribution is unrecorded, management's model of the operation is the rulebook, and the model appears to work, right up to the moment discretion is withdrawn. Safety researchers give this gap a standing name: work-as-imagined against work-as-done. And some of what is withdrawn could not have been written down even in principle. Michael Polanyi's observation that we know more than we can tell applies directly here — an operator who knows the machine sounds wrong today knows something real, acquired by practice, that no procedure can state.

The claim is not that rules are useless or that procedure is a fiction. It is narrower and more interesting: rules are systematically incomplete, and the incompleteness is not a defect that a better-written manual would remove. Any finite text specifying an open-ended activity leaves a residue, and the residue is filled by people using judgement. A rulebook works as a floor and a common reference, not as a generator of the work.

Three honest qualifications. First, a work-to-rule is an industrial tactic, and the literalism is deliberately maximal — a determined worker can find a defensible reading of a rule that happens to be slower. Part of the slowdown is chosen, not purely mechanical, and treating the whole of it as a pure measurement of the unwritten contribution overstates the case. Second, the tactic is usually combined with refusing overtime and voluntary extras, which is a straightforward withdrawal of labour and should not be credited to the incompleteness of the rules at all.

Third, and most awkward: the discretion being withdrawn is not automatically virtuous. Some informal shortcuts are unsafe, and a work-to-rule can equally reveal that the normal pace depended on corners quietly being cut. The same event supports two readings — the rules are inadequate to the work and the normal pace was only possible because the rules were being broken — and which one is true is an empirical question about a particular workplace, not something the tactic settles.

• How practitioners acquire the judgement no procedure can state, taken up in the walk-throughs on skilled diagnosis and deliberate practice.
• Why organisations that try to write the informal layer down usually find that formalising it destroys it.

Every term the collection defines is gathered in the glossary.