Forecasting Accuracy of In-Progress Activity Duration and Cost Estimates

Abstract

When a project activity has already started, tracking information such as percent complete and current activity duration and cost can be easily retrieved. This information can be used to update the project schedule to anticipate the eventual project duration and cost more precisely. But hardly any studies exist that compare how more accurate or reliable activity tracking information can be compared to the initial (planned) estimates, let alone what mathematical forecasting expressions are the most accurate. Therefore this paper quantifies said forecasting accuracy by extracting over three thousand activities with partial tracking information (which have already started, but are not yet complete) from a real project dataset. Two expressions for forecasting the activity duration and cost are tested by comparing their performance against initial (planned) and final (actual) values. The contributions to the body of knowledge are fourfold. First, it is shown that activity tracking information considerably outperforms planned estimates. Second, using two expressions can significantly minimize the deviations of time and cost estimates. Third, remaining activities’ duration and cost estimates can be closely modeled with Lognormal distributions as a function of the activities’ percent complete. And fourth, variability decreases linearly as activities approach their end. These findings allow project managers to better anticipate and model the duration and cost variability of ongoing activities and improve the forecasting accuracy of the project duration and cost estimates.