Subjective Estimates and Scheduling Decisions

Network scheduling models typically assume that estimates of individual activities are stable and are not improved or degraded as progress is made in a project and the time to undertake a particular activity comes near. The formulation of scheduling models is usually compatible with this assumption about estimates and excludes the effect of actions such as rescheduling which might be taken in response to changes in estimation error.

The present paper examines the nature of subjective estimates in an actual development project. Error characteristics as determined from prior estimates and actual outcomes are analyzed using a multiple regression model. Several variables are significant in their association with changes in estimation accuracy as activities are reestimated. Contrary to previously reported results there is a strong learning phenomenon. Implications for future work are discussed.