In recent years many formal attempts, e.g. Baker&Pound, (1964), have been made to help R&D management assess the worth of R&D projects and select those most worthy of inclusion in the R&D programme. Most of these attempts are based on financial appraisals of a project's worth and require estimates of the costs and likely revenues that will result from the adoption of a given project. It is clear, therefore, that the validity of any measure of a worth for a project is dependent upon the accuracy of the estimates of cost and revenue associated with that project. Unfortunately the limited amount of evidence so far presented suggests, in particular, that cost estimates are extremely inaccurate.In this paper, therefore, we try to add some further evidence on forecast accuracy for a number of industrial R&D projects. We begin with a short review of the existing literature evidence and continue with an analysis of the past effectiveness of R&D work and, in particular, of forecast accuracy.