Last year, the development nonprofit Evidence Action did something remarkable: It stopped fundraising for one of its programs because it had conducted a study that suggested it was less effective than it had thought. The program was called No Lean Season. It built on evidence that suggested that small subsidies to allow poor rural families in Bangladesh to migrate during periods when there was no agricultural work were a big help. No Lean Season gave Bangladeshis the money for a trip to the city, where they could find more reliable work.

It seemed like a great program — until Evidence Action’s study last year found otherwise, and prompted it to put further fundraising for No Lean Season on hold. Yesterday, Evidence Action announced that it was pulling the plug altogether after the study found other problems in the development program.

Here’s why this is noteworthy: Most development programs are not rigorously studied, and they’re often run by people who want to keep them going — as long as there’s any evidence they could be a good idea. Evidence Action is doing something different. It’s trying to be quick to test programs, and quick to admit if the programs don’t work or just aren’t likely to be the best possible place to spend money. It’s also being unusually transparent with the public about how it spends its money.

So, while it’s unfortunate that the program didn’t succeed in helping Bangladeshis, it’s exciting to see us moving toward an approach to development that is willing to cancel programs when it can do better elsewhere — even if, in cases like this one, it’s also painful.