Increasingly, development funding is directed to programmes aiming to make market systems more favourable for smallholders and low-income consumers of food. The development outcomes of these programmes are not self-evident. Programmes operate in dynamic markets full of uncertainties and surprises and depend on many other factors not under their control. Assessing whether a programme indeed contributed to development outcomes is challenging. Building on real-world experiences with theory-based evaluation in inclusive business programmes, this IDS Bulletin discusses approaches and methods for meaningful impact evaluation. It examines how these evaluations provided information that made programmes accountable to the donors while also helping the implementing agencies to learn and adapt their programmes. In this IDS Bulletin, the authors discuss the experiences of practitioners and academics in finding doable and creative ways to conduct impact evaluations of inclusive business programmes in the domain of food and agriculture. Inclusive business programmes that work in the area of food and agriculture aim to change current business practices of small and medium enterprises in a way that these include smallholders as producers or target poor consumers as consumers. The examples show a convergence in methodological approaches, with ‘What works for whom under what conditions’ as the key learning question. All use a combination of methods that complement and build upon each other. However, smart data collection and sharp analysis and synthesis alone are not enough. The evaluation process and outputs also need to be informative for the stakeholders involved. More interaction and sense-making between implementers and evaluators are needed. All experiences presented in this IDS Bulletin acknowledge that it is not easy to find ways to make learning useful for evaluation commissioners and implementing agencies. Under the right conditions, the presented approaches and tools might work and accelerate the learning loops for adaptive management. Crucially, three conditions appear as necessary for a good theory-based evaluation: 1. having interested ‘listeners’ as the audience of the evaluation; 2. applying rigour in anticipating and addressing validity threats to the conclusions; 3. sufficient resources for an appropriate mix of methods.