An impact evaluation of a pro-poor market system development programme, ‘Making Markets Work for the Poor’ (M4P), poses several methodological issues for evaluators. M4P interventions intend to change the contextual conditions in which stakeholders take business decisions so that it triggers change processes in the wider social system and ultimately benefits poor people. An impact evaluation design for such a programme thus needs to be robust enough to adequately capture these systemic outcomes, acknowledging dynamic changes in intervention delivery as well as in market conditions over time. Theory-based evaluation can provide learning and accountability when it incorporates methods that allow a critical reflection on the key causal steps in an intervention’s theory of change. We present our learnings about indicators and methods to reflect on the importance of the contributions to market system change of a large M4P programme in Ethiopia.