Abstract
Predicting the cost at project completion is a necessary task, especially on large projects, but it is a process that is tainted with inaccuracies. This project reviews earlier literature on project cost forecasting and consequently develops a comprehensive model for relatively accurate forecasts. The model incorporates a comprehensive array of analytical tools in the application of earned value analysis concepts, viz., cost indexes, moving averages, time series, linear and non-linear regression, suitable weighting, and confidence analysis using central limit theorem. A sample set of costs at project completion was created, from which the final cost at project completion was derived through confidence analysis. The study was carried out on an example project with forecasts delivered regularly at each reporting period. The sample size of candidate values for cost at project completion keep increasing as the project progresses. This ensures that all periodic efforts through the project duration are given equal weight in forecasting. After completion of project, a postmortem was conducted to determine the exact confidence level at which forecasts should have been made at each reporting period. This led to the development of a declining confidence gradient to be used at each reporting period as the project progresses. A prorated declining gradient is practical from a forecast perspective. It was found that the declining gradient decreases from 100% at very early stages of forecast to 25% at the penultimate reporting period for the project analyzed. The application of declining confidence levels yields the most accurate forecast. Further research is suggested to evaluate a broad spectrum of projects.