Five simple principles for scaling up in aid

There is a lot of discussion in aid on scaling up small-scale successes in aid to reach many more potential beneficiaries. But what things can be scaled up? Here are some principles so simple that they would be embarrassing except that they are routinely violated in aid. (1) Scale up success not failure

The only reason for mentioning this is that the aid business has a strange habit of trying to scale up again things that have already failed. PROGRESA is the great success story of scaling up something after you had determined it was successful.

(2) Don’t scale up what you think is most important, scale up what you do best

There are lots of important issues, so why not choose the one that you do best? And let the people who are good at the other issues work on the other issues? Yet both official agencies and NGOs are often pulled away from what they do best by well-meaning politicians and funders who are focused only on final goals.

(3) You can scale up only what requires cheap, abundant inputs; you cannot scale up something that depends on expensive, scarce inputs

This is one of my problems with the Millennium Villages – they at least partly depend on world-class experts flying in to solve idiosyncratic problems of each village. World-class experts are a scarce resource that you can’t scale up.

(4) Things that you make routine are among the easiest to scale up

It worked for Henry Ford, McDonald’s, and WalMart, why not in aid? One of the secrets to success of the large vaccination campaigns that reduced child mortality was that relatively unskilled medical workers (in abundant supply) could give vaccinations as a routine activity. Of course, not everything can be made routine. For a more complex discussion about social service delivery in general, see the great paper by Lant Pritchett and Michael Woolcock.

(5) Evaluate whether you are still successful after scaling up

Scaling up often changes the nature of what you are doing, so evaluate whether the scaled-up version works as well as the original version.

I'm sure readers have other principles to suggest -- please do so!