Estimating the effort required for software process activities continues to present difficulties for software engineers, particularly given the uncertainty and subjectivity associated with the many factors that can influence effort. It is therefore advisable that managers review their estimates and plans on an ongoing basis during each project so that growing certainty can be harnessed in order to improve their management of future project tasks. In this paper we investigate the potential of using effort data recorded for completed project tasks to predict the effort needed for subsequent activities. Our approach is tested against data collected from sixteen projects undertaken by a single organization over a period of eighteen months. Our findings suggest that, at least in this case, the idea that there are ?standard proportions? of effort for particular development activities does not apply. Estimating effort on this basis would not have improved the management of these projects. We did find, however, that in most cases simple linear regression enabled us to produce better estimates than those provided by the project managers. Moreover, combining the managers? estimates with those produced by regression modeling also led to improvements in predictive accuracy. These results indicate that, in this organization, prior-phase effort data could be used to augment the estimation process already in place in order to improve the management of subsequent process tasks. This provides further confirmation of the value of local data and the benefits of quite simple quantitative analysis methods.