On why we’re so good at counting what we do, and so bad at counting what we prevent
There is a report sitting on someone’s desk right now — meals served, vaccines administered, households reached. Every number in it points to something that happened. But ,somewhere, a family didn’t spend three months’ income on a hospital bill because they got care early enough. Somewhere, a child didn’t develop the complication that would have shown up in a different report, a worse one. None of that is in the document. It can’t be. We don’t have a column for it.
This isn’t a usual data problem — messy spreadsheets, missing fields, bad Excel hygiene. It’s something more structural. Our entire measurement architecture is built to count presence. Doses given. Sessions held. Beneficiaries covered. We have structurally shyed away from counting absence — the disease that didn’t progress, the expense that never became catastrophic, the crisis that quietly didn’t happen because something upstream worked.
And because we can only fund and report what we can count, we have built an entire sector’s incentive structure around the visible half of the story. Lets deep dive in to certain constructs that shaped this narrative.
The Trap of “More Usage Is Good”
Take a familiar example from public health. A programme improves access to care in an underserved area. Utilisation goes up – more people are walking into clinics who weren’t before. So,everyone treats this as a win.But if we sit with the number a little longer it stops being simple. Increased utilisation could mean that need is met. It could also mean that more people are falling sicker now, or it could mean that a system which has no real gatekeeping is absorbing load it wasn’t designed for. Or it is an outcome of an action/ intervention that could have happened earlier such as prevention, awareness, basic sanitation, etc .
The number, “care utilisation up 40%,” can be a triumph or warning depending on what’s actually happening underneath it. Most dashboards don’t have the resolution to tell the difference. So we default to the reading that’s easier to celebrate.
We Applaud the Giving, Not the Not-Needing
The aid architecture — CSR budgets, foundation grants, government schemes — is built around the act of giving. You give a meal, a bednet, a cash transfer, a health camp. The organisation that gives more, more efficiently, to more people, is the one that gets funded again next cycle.
Nobody’s incentive structure rewards the version of success where, five years in, fewer people need the meal. If a programme succeeds so completely that demand for it drops, that’s often read as a shrinking footprint, a shrinking case for renewal. We have built a machinery that is extremely good at continuing to give aid, and mildly suspicious of any evidence that aid is no longer needed.
We forget that the best possible outcome of a well-run aid programme is often its own obsolescence.
Self-Sufficiency Doesn’t Look Like a Milestone
Part of the reason this tension is so persistent is that self-sufficiency in this case does not have a positive optics. There’s no ribbon-cutting moment for “this household no longer needs support.” It shows up as an absence over time — a family that stops enrolling, a village that stops requesting the health camp, a line item that quietly shrinks. In a sector trained to report growth, a shrinking line item is often perceived as a failure Compare that to distributing aid. Aid has a clean unit of measurement: how much, to how many, how often. It photographs well. It fits a quarterly report. Prevention and self-sufficiency, by contrast, ask you to prove a negative — to make the case that something bad didn’t happen because of something you did, which is a much harder argument to make convincingly, and a much harder one to fund on.
So funders, not unreasonably, gravitate toward what they can verify. And organisations,build programmes toward what gets funded. Nobody in this chain is behaving irrationally. The system as a whole still ends up optimising for continued need rather than resolved need.
Sitting With the Discomfort
There are no clean fixes here, measuring absence is genuinely hard — you’re trying to quantify a counterfactual, and counterfactuals are slippery by nature. You can build proxies: catastrophic health expenditure avoided, complication rates over time, repeat-need versus first-time-need ratios. But every proxy is an approximation of something you can never directly observe, and approximations are easy to game or misread if the incentive to look good outweighs the incentive to be right.
But what is essential is asking– asking every time a number goes up, what else that number could mean. Treating “more usage” as a question rather than an automatic win. Being open to the perspective that a shrinking need for a own programme might be the strongest evidence that the programme worked None of that solves the underlying measurement problem. But why not keep our perspectives balanced?
