1.3 Same effort, very different results

In the 2000s, a charity called PlayPumps had an idea that photographed beautifully. Replace hand pumps in African villages with children's merry-go-rounds. As the children played, the roundabout would pump water into a tank. It won awards and attracted large donations.

It did not work as promised. Each pump cost about $14,000 to install. To meet the charity's stated goal of supplying water to 10 million people with 4,000 pumps, children would have had to "play" for 27 hours a day. For the cost of one PlayPump, a chlorine-dispenser programme run by Evidence Action could provide clean water to over 9,000 people for a year.

PlayPumps is an extreme case, but the pattern is general. Ways of helping with the same goal can differ enormously in what they achieve.

Those numbers probably overstate the real gap. Measurements are noisy, and the interventions that look best are partly the ones that got lucky in the data. This is regression to the mean. Todd's own adjusted estimate is that the best measurable interventions in an area are roughly 3–10 times better than the average one, not 20–200 times. That is smaller, but still large. Choosing well can matter as much as trying hard.

So how do you choose? Throughout this course we will use three questions about each problem:

• How big is it? How many lives does it affect, and how badly?
• How neglected is it? How much money and talent already go into it? The less there is, the more an extra person or euro tends to add.
• How solvable is it? If more resources went in, how much progress would they buy?

These are rules of thumb, not a formula. In Unit 6 you'll use them to rank the problems yourself. The intro course spends a week on them.