Social policy research often depends on the application of generalisations from social science. Questions like ‘what works?’ assume that general principles can be translated from specific examples into other contexts. Pawson and Tilley argue that effective policy research has to depend on the idea of a ‘generative mechanism’, or relationships of cause and effect. Explaining issues in terms of causes, however, is problematic. Social phenomena tend to be multifaceted, and even relatively simple phenomena are likely to be influenced by a range of different factors; causal analyses have to be developed by interpretation, and the analyses are frequently wrong. Causal explanations often claim to do more than they can deliver: even if there is a convincing causal explanation, it does not necessarily imply any prescription for policy. There are ways of generalising, however, that do not depend on causal analysis. Phronesis develops principles experientially, setting them against empirical evidence, and it does not need to consider underlying mechanisms to be effective. Phronesis provides the basis for a critique of technocratic approaches, a rationale for action and a focus for the development of alternative methods and approaches. A dependence on phronesis cannot avoid all of the pitfalls associated with generalisation, but it is more flexible, and less presumptuous, than a causal approach.