This paper illustrates problems in the evaluation of community development programs for health promotion. It is based on the authors' retrospective process evaluation of the Wanganui Community Alcohol Action Program (WCAAP), a recent example of a health promotion program directed at reducing alcohol‐related problems in a small New Zealand town. Described by its designers as a community‐based program, it included coordination of community organizations, education, publicity and increased enforcement of drinking laws. Discussion of the problems in the evaluation of such programs puts them within the context of the substantial body of previous social science research in both evaluation and community development. This, it is argued, is a body of knowledge with which health promotion researchers need to be conversant. Community development programs usually stem from a process of negotiation between interest groups with differing objectives. This results in a changing definition of both the problem to be solved and the nature of the solution, making evaluation difficult. Community development is also likely to be seen as a new solution providing a panacea for old problems. This can lead to such programs being too ambitious. From the point of view of experimental design these programs are likely to have a number of technical problems. This paper argues that these problems are so significant that it is often unwise to attempt large scale evaluations of community development programs. Rather, attention should be concentrated on critically assessing the policy‐making process and disseminating previous knowledge about such programs.