Most social programmes can point to activity: people attended, sessions were delivered, satisfaction surveys came back positive. Far fewer can show that those activities caused a real improvement in the outcome the programme was actually meant to change. The gap between activity and impact is where a great deal of well-intentioned funding quietly disappears.
Correlation is the easy trap
If people who join a programme end up doing better than people who do not, that is suggestive — but it is also exactly what would happen if the programme mainly attracted people who were already more motivated, better connected, or further along to begin with. Rigorous evaluation exists to separate the effect of the programme from the effect of who chose to show up for it, usually through a comparison group that looks as similar as possible to participants in every way except the programme itself.
Build measurement in from day one
Evaluation added at the end of a programme, once funding is already spent, is usually too late to answer the questions that matter — there is no baseline, no comparison group, and no way to separate the programme’s effect from everything else that changed over the same period. The organisations that produce credible evidence design measurement alongside the programme itself: a clear, specific outcome the programme is actually trying to shift, a baseline measured before it starts, and a comparison that makes it possible to say, with some confidence, what would have happened anyway.
Evidence should change decisions, not just reports
The point of evaluation is not the evaluation report — it is the decision that follows it. That means being genuinely willing to scale down or redesign an initiative the evidence does not support, however much institutional energy has gone into it, and equally willing to invest further in something the evidence supports even if it was not the flagship idea going in. A sector that only ever publishes the evaluations that flatter its existing programmes is not really evaluating anything.