Skip to main content

(EST-2877) Credible Estimates Require Probabilistic Models

In estimation and forecasting, a useful reference calculation is a project model solved with best single-point assessment inputs. However, this deterministic solution (e.g., cost or schedule) seldom provides an objective project estimate. Proper calculations under uncertainty require Monte Carlo simulation (MCS). In this, subject matter experts assess uncertain input parameters as probability distributions. The MCS model then produces cost, schedule and other outcome values as distributions.

MCS provides important benefits: 1) Output values are distributions, communicating uncertainty; 2) For the most part, unbiased inputs produce unbiased model outputs; and 3) The statistical mean outcome value is often substantially different than the reference value.

Stochastic variance (SV) is the correction from the deterministic solution to the mean MCS result. Contributors to SV include nonlinear equations, correlations, and options. When preparing a variance analysis “reconciling forecast to actual” the important SV term stands alone. A periodic variance analysis provides a useful report and project control tool.