Variance analysis in earned value measurements when large uncertainties exist in underlying estimates

This paper addresses variance analysis in earned value measurements (EVMs) when budgeted costs are based on estimates that have large uncertainties. EVMs have traditionally been predicated on budgets that are based on "exact" or deterministic estimates. In these traditional cases, whenever the difference between the actual and budgeted values for an individual project element exceeds some nominal threshold, project personnel are typically required to analyze and explain the variance between budgeted and actual amounts as a precursor to bringing their work under control. Research projects and technology, product, and software development projects, however, often have large uncertainties in their underlying estimates. In these cases, differences or variances between the actual cost and budgeted cost of project elements can exceed thresholds normally applied in EVMs even though actual element costs are still within their estimated ranges. If variance analysis is required in these situations, much annoying and perhaps needless and meaningless work is involved. In this paper, situations like this are examined, and a sensible approach to variance analysis is proposed for them