What it adds up to: a practitioner's synthesis

Ten years on, the repertoire is wider and the ground less certain. Four practical conclusions follow for anyone designing or commissioning MEL today.

Four practical conclusions

1. The method question is no longer "RCT or not" but "what causal claim do we need, at what confidence, for which decision". A defensible design in 2026 states its causal logic (theory of change or system map), chooses the inference strategy that fits (experimental, quasi-experimental, theory-based or a combination), makes its evidence standards explicit (rubrics, quality-of-evidence tests, pre-registration where appropriate) and says who judged and how.
2. Learning is now expected, not optional — but it has to be evidenced. Learning agendas, pause-and-reflect cycles and adaptive management are routine language; what separates serious practice from rhetoric is a record of decisions that actually changed because of evidence.
3. Value and power are on the table. The 2019 DAC revision, the equity and decolonisation frameworks, the localisation targets and the trust-based philanthropy movement all ask the same question from different directions: whose values define success, and who owns the evidence. A MEL design that cannot answer that will be challenged.
4. The economics have changed. With USAID gone and European donors cutting, the market for large-scale external evaluation has shrunk, while demand for cheaper, faster, AI-assisted and locally owned evidence has grown. Cost-effectiveness, proportionality and evidence reuse (synthesis of what already exists) are back at the centre of the craft.