I was reading this profile of Esther Duflo in the New Yorker last night when I realized that rural development and the tech sector aren't too different in their approach to problems. In each one you're trying to convince the customer that what you're selling - immunizations, software - will make their lives better. Sometimes you're not reaching out to the customer that well, in which case you try experiments to make things better. In tech, this is pretty low cost - you segment the customer base, show the treatment group a revised page and then measure the differences. In development this is more expensive - you have to create a randomized controlled trial, divide your population in two and then measure the results later. The basic principles are the same, though. Most successful web designers and aid workers/economists believe the following: 1. We can be doing a better job selling our product 2. Our assumptions could be right or could be wrong, but we have no way to tell without testing them 3. Good data is essential 4. Opinions without data are meaningless 5. Experiment constantly to collect more data and find things that work The success of this approach has become so clear, and such a part of my approach to problems, over the past few years that it's hard to imagine doing things differently, or having a discussion with people that rely on their gut. Even having discussions with people is pointless, when you can go look up some papers or evidence that proves or disproves your point, but most people aren't willing to trust the data over their intuitions. There are two types of people who have opinions about social promotion in schools: people who have looked at the data and people who haven't. Overwhelmingly the people who look at the data are in favor of social promotion (as are school administrators). Institutions push back against the data-driven approach, however, because of its implications. Underlying the hierarchical system at most companies is a belief that the people at the top are better at making decisions, or their assumptions are somehow better than the people just starting. If we're trusting the data, then your guess as the CEO is just as good as mine, as the intern; the people at the top must be comfortable ceding power. Plus we have all these biases, like the ability to discern patterns out of random data and anchoring, and if you're not familiar with cognitive bias by this point go read this. In both the tech sector and in development, there are companies that - for the moment - aren't randomizing, and are still doing okay and collecting money. That's going to change, though, and soon, because relying on the data gives you such a clear advantage. Google's made billions with it. J-PAL's gaining on everyone else in development, only because their experiments are demonstrating results. If you've got nothing else to go on, evaluate the future success of an organization by how it makes decisions, and how open it is to change.

Liked what you read? I am available for hire.