I’ve spent eighteen years in design, twelve running the function at a company I co-founded, and I learned the hard way that the decision which matters most is usually what not to build. Customers asked us to put analytics inside our data platform. Instead I shipped one default view covering what most people needed and sent power users out to the tools they already paid for — a choice that saved years of engineering we’d have spent losing to specialists.

With OpenSRP, we had to design electronic medical records for health workers who’d never held a smartphone, I built the first version out of things they already carried: paper ledgers and the passport cards families keep for their children. It worked. It has collected 1.6 million health services in 80+ implementations.