Satkhira sits where the Bay of Bengal is steadily winning. Salt water has pushed into the groundwater and the fields, so families lose their drinking water and their income in the same season. When that happens the first thing to go is a child’s schooling — usually a girl’s — and once a child stops attending, they rarely come back.

So the work is deliberately unglamorous and local. Put a learning hub inside the village rather than expecting an eight-year-old to travel to one. Hand the family a uniform and a set of books so school costs nothing they cannot spare. Teach mothers how to make water safe, because a child who is repeatedly ill cannot attend anyway. Train the teenagers in cyclone response, because they are the ones who will be there when it comes.

The number that matters most is not how many children enrolled — it is that 94% of them stay and finish primary school. Enrolment is easy to buy. Retention is the thing that changes what a life looks like twenty years on, and it is the figure we hold ourselves to.

This is also the same instinct that runs through the research: the point of a diagnostic model is a patient who gets treated earlier, and the point of a learning hub is a child who can still read at sixteen. Both are about getting a useful thing to the people who have the least access to it.