Some years ago, two British supermarket chains needed to place a large order for replacement trollies. They had to decide what ratio of full-depth and shallow trollies to order, and how best to allocate them between stores.

One of them amassed mounds of demographic data and tried to construct an elaborate optimisation model. The second took a different approach: ‘We already employ people to collect trollies in the car parks — so why don’t we ring them and ask what they need?’ The second approach worked much better.

How often are problems best solved not centrally but by devolving decisions to people on the ground? Anyone working in the car park of a superstore is alert to factors no head-office model can know. And this approach is safer overall: if a number of trolley-pushers get their ratio wrong, their errors are likely to cancel each other out. If the centralised model is wrong, you end up with a large surfeit of one kind of trolley.

Solve for the specific and you may also solve for the general. The reverse does not apply. Or as Luca Dellanna recently wrote on his blog in what is perhaps the best sentence of the year: ‘Centralisation is only efficient when viewed from the centre.’ Luca is an expert in complexity theory and someone Dominic Cummings should hire. In fact every government department should have a complexity expert on hand to prevent the group-think which arises when you have a homogeneous group trying to impress each other with single solutions.

This problem occurs, I think, because humanities graduates think that policy is like high-school physics: a space where the same laws apply regardless of scale. But school physics is one of those rare cases where solving for the whole also solves for the specific. The confusion leads to what Nicholas Gruen calls ‘the war on context’, where you create off-the-peg policies or metrics detached from local reality. It is exacerbated by the assumption that public services should be egalitarian. You can sit in public sector meetings for years before you’ll hear anyone say ‘Why don’t we offer people a choice?’ (it took ten years for government to offer the option to pay your TV licence fee monthly rather than annually).

Universal metrics stifle experimentation and choice, and without these you never discover what people really want. There are first-class and second-class stamps for a good reason: you may want a letter to arrive overnight or you may simply want it to arrive in a few days. Speed isn’t always the decisive factor. Surely the same applies to A&E waiting times? There are some medical conditions which require urgent treatment and others where it might be preferable to offer someone a choice to ‘come back in five hours and the dermatologist will see you then’. To apply the same metric to all cases seems dumb.

The private sector is far from immune to this folly. Take the airline industry’s obsession with legroom. Legroom is a fine metric for lanky males — but for 90 per cent of women, and 50 per cent of men, elbow-room is much more important. The Boeing 787 was designed by Boeing for a family-friendly 2-4-2 seating configuration. Every airline has replaced it with the dreaded 3-3-3.

A large part of the reason I am opposed to HS2 is that I know the train interiors will be shit. Artists’ impressions will be glorious: cocktail bars, children’s play areas, dining-cars and abundant tables. Then some dismal accountant will get involved and replace them with ranks of austere-looking seats with backs like ironing boards, crammed into a monotonous configuration which pleases no one.