Statistical Project Management

    A comment on Hal Macomber's site to the topic of "Deadlines," brings up the issue of using statistics in the assessment of project performance and forecasting.
    David Green in a post on Hal's site suggested Thucydides had the notion that it is "impossible to calculate accurately events determined by chance."  First the terms "impossibility" and "accuracy" need units of measure before being useful. Second it may have been Thucydides was sleeping during the analytical statistics class.
    Statistics and statistical process control, statistical pattern recognition, statistical noise reduction, statistical thermal dynamics, statistical stress analysis and myriad of other statistical modeling all depend on predicting future behavior from "events of chance." The missing piece from that quotation is "to what confidence level can you know the future from events of chance?"
    This statistical analysis approach is the basis of Monte Carlo simulations mandated for DID 81650 compliant programs (programs greater that $20M). There are several tools available for most project management applications (Risk+, @Risk, PertMaster, Crystal Ball). Each depends on the user defining the underlying statistical profile of the task durations, having a properly defined network of activities, and an understanding of how the correlations between these activities impact the statistical behavior of the project.