The current ACIAR (Australian Centre for International Agricultural Research) guidelines for impact assessment of agricultural development projects see impact assessment as being useful for both accountability to stakeholders and as a learning tool to find out what works, what doesn’t work and why. The methodology involves the use of conventional economic evaluation and the estimation of a money metric based on measuring outcomes in terms of economic surplus changes attributable to directed actions and activities. On the question of accountability to stakeholders, this paper suggests that the money metric may not be the best outcomes-based measure of performance against development goals and that other performance indicators ought to be considered. The paper also suggests exploring other approaches to assess accountability including qualitative (narrative) methods as well as process-based accountability. On the question of using impact assessment as a learning tool, the paper suggests this might be quite useful for more traditional non-adaptive research, but is less useful for adaptive research projects involving participatory action research (PAR). With PAR projects, learning about what works, what doesn’t work and why already occurs as an integral part of the research process. The paper concludes with some thoughts about project evaluation of an ACIAR-funded project with which the authors are involved in northwest Cambodia focusing on upland crop production and marketing.