Free guide
Choosing an Impact Evaluation Design
How to choose between randomised, quasi-experimental and other designs for evaluating the impact of a health programme or policy, with the assumptions each design needs and the questions that decide the choice.
PDFUpdated 30 September 2026Impact evaluation
What's inside
- The counterfactual problem at the heart of impact evaluation
- Randomised designs and when they are feasible
- Difference-in-differences, regression discontinuity, matching and interrupted time series
- The key assumption behind each design, in plain language
- A decision table to match designs to programme situations
- Adding costs to answer value-for-money questions
Key points
- 01Every impact evaluation needs a credible estimate of what would have happened without the programme.
- 02The best design depends on how the programme is rolled out and what data exist.
- 03Plan the evaluation before roll-out wherever possible; it widens the options.
- 04Collect cost data alongside outcomes to answer value-for-money questions.
Frequently asked questions
What is the difference between monitoring and impact evaluation?
Monitoring tracks whether a programme is delivered as planned and what outputs it produces. Impact evaluation estimates the change in outcomes caused by the programme, compared with what would have happened without it.
Is a randomised trial always the best design?
Randomisation gives the most credible comparison when it is feasible and ethical, but many programmes cannot be randomised. Well-designed quasi-experimental methods can provide credible evidence when their assumptions hold.
Can impact be evaluated after a programme has started?
Often yes, using routine or survey data with methods such as difference-in-differences or interrupted time series, but the options are narrower than when the evaluation is planned in advance.
Need this done properly?
This guide explains the method. If you would rather have the analysis done for you — with local data, independent quality control and a report you can defend — we can help.
Need this done properly?
Tell us about the decision you need to support. We will propose a scope, timeline and price.