Of course, the big challenge — and the big argument for localism — is that you never actually *know* what sort of ‘response surface’ or ‘variability across places’ you are dealing with. Even if you do the evaluation (which seldom really happens — have a look at http://www.treatyforgovernment.com/), it’s near-impossible in the complexity of implementation to know what drove particular outcomes.

And it’s not only the ‘causal texture’ (https://journals.sagepub.com/doi/abs/10.1177/001872676501800103) of the place and the fitness landscape of the policy (see https://en.wikipedia.org/wiki/Fitness_landscape — but of course policy seldom evolves) which are important — it’s the *way* the policy is implemented locally, interactions with all the other local implementations, and how the policy is *communicated* locally.

All pretty good reasons for the intermediary role of the commissioner — see Gary Sturgess’ diagram which I included in https://medium.com/@antlerboy/what-is-commissioning-c77b5983bc12. Clearly some things need to be done nationally — but extension of your arguments suggest it should be (and has to be) the exception rather than the rule.