The rule you are about to write

Most of these studies examine a decision made once, by people who then lived with its result. A policy is not like that. It is a decision made once and applied many times, by people who were not in the room, to cases its makers could not see. That separation – between whoever sets the rule and whoever meets the case – is what makes policy its own kind of decision, and it is where the characteristic failure starts.

Start where rules actually start. Something goes wrong. A supplier is paid twice; a date is promised that nobody checked; a change ships on a Friday and the weekend goes on recovering it. Sensible people meet, and a proportionate rule is written: two signatures above a threshold, delivery dates confirmed before they are given, no releases after Thursday. The rule is good. Everyone in that room could say exactly why it exists.

Which is the moment its reason is most at risk, precisely because it is obvious. Obvious things do not get written down. The rule is recorded because it has to be followed; the reason stays in the room, held in the shared memory of the week everyone spent clearing up. Then the rule ships – into the handbook, the checklist, the induction pack – and it travels perfectly. The room disperses. People move on, move up, leave. The reason does not travel at all.

What is left is a rule that can only answer for itself. Ask why releases stop on Thursday and you get because that’s the policy, or it’s how it’s done, or it’s always been that way. Those are not evasions. They are the honest report of someone keeping a rule faithfully, which is what they were asked to do. The fault is not in the person. It is in what was handed over.

The cost arrives in the one case the rule was never written for. A rule that comes with its reason equips whoever holds it: they can tell when a case is the kind the rule was made for and when it is not, and they can raise the second sort rather than forcing it through the first. A rule that arrives bare cannot do this. It has nothing to hold the case against, so every case is either compliant or non-compliant, and the judgement the rule was meant to support is quietly replaced by the rule itself. The first kind builds up the capacity of the people who keep it. The second feeds on it.

So the question worth asking while the rule is still being written – while the reason is still in the room and still visible – is what has to travel with it. Two things, in practice. The first is the reason in plain words: not the incident report, but the sentence that says what this rule protects and what it would cost to lose it. The second is rarer, and harder to write, because it asks you to imagine the rule outliving its usefulness on the day you are proudest of it: the condition under which it should be reviewed or dropped, set down now, while you can still see clearly.

Volumes change. Systems change. The supplier who was paid twice is no longer a supplier. A rule that carries its own review condition can be retired honestly, by anyone who finds it. A rule without one can only be broken, defended, or endured.

None of this settles whether to write the rule at all. Some situations want a rule and some want a conversation, and the instrument does not choose between them; that belongs to the people who will live with the result. What it refuses is the making of a rule as though it were free. A rule is a constraint placed on people you will not meet, in situations you cannot foresee, and the only question the instrument presses is whether what you are handing them helps them see – or stands in for seeing.

The rule you cannot account for – the approval gate no one defends, the standing report no one reads, the practice that survives every restructure – was once a rule like the one you are about to write. It was proportionate. It was written by someone sensible, in answer to something real, on a day when the reason was obvious to everyone in the room.


This is one of the moves the programme makes visible. The full account of what it is and how it works is here.