The What Works movement continues to thrive - seven What Works centres are already up and running (or nearly), with new ones in the pipeline, providing access to the state of knowledge for policymakers and practitioners. The Alliance for Useful Evidence is also doing well, linking several thousand people and organisations in a community devoted to more intelligent use of evidence. The UK does seem to be building a more mature ecosystem for evidence, and over the next few months we’ll see some interesting new steps, like Manifesto Check, which we have supported and which will assess party manifestos as to how well they’re grounded in evidence.

I’m a great enthusiast for this movement. It’s still amazing how many decisions are made without reference to what the world already knows. But one thing sometimes gets forgotten by the enthusiasts. The crucial question is not 'what works', but rather 'what works for who, when, where and with who'. You could call them the 'Six Ws'. What works at one time and place may not in other times and places; it may work for one demographic group but not for another; and it may depend on one implementation capacity and not others. Systematic exploration of these dimensions of what works is vital if big mistakes are to be avoided.

Bill Clinton once commented that 'Nearly every problem has been solved by someone, somewhere. The challenge of the 21st century is to find out what works and scale it up.' It’s a comment that has been widely quoted. But it continues to mislead as much as it encourages. Yes, we need much more openness to adopting and adapting from elsewhere. But we also need smarter ways of doing so, and widespread skills to take the essential elements of ideas and know how to fit them to varied contexts. Just copying apparently good practice can easily backfire, and I’ve seen plenty of examples of apparently proven models failing to transfer, usually because there hadn’t been enough detailed attention to the factors that had made them work.

This is also why the ‘what works’ movement doesn’t just advocate mindless replication, but rather intelligent feedback loops at every level to assess the other four Ws. We need schools and hospitals with lots of data, but also plenty of opportunities for reflection on data, evidence and experience – for example through study circles. We need networks that link practitioners together to share ideas and innovate incrementally. And we also need the more formal feedback loops of experimental trials. All contribute to a collectively intelligent system, that can make sense of the options it faces.