Concept: Some project topologies degrade the reliability of ES metrics early in a project’s timeline. ES Longest Path (ES-LP) alleviates the difficulty.
Figure 1
Practice: A schedule’s “shape” affects forecast reliability. Serial schedules produce more reliable forecasts than parallel schedules. A purely serial schedule eliminates non-critical activities that can degrade forecasts. Conclusion: serialize the schedule and improve the reliability of ES forecasts.
Walt Lipke’s Longest Path serializes schedules. It first identifies all serial paths. It next calculates a duration forecast for each time period in each path. It does not use the actual pattern of PV and EV, however.
Instead, it removes void periods, starting at the first period and proceeding without interruption to the final period. The LP duration forecast is then calculated. Void periods are factored back into the calculation. Anomalies are identified and removed. What results is the longest path through the schedule (the green cells in Figure 1).
Research by Lipke (2014b) has shown that LP improves the reliability of duration forecasts.
LP forecasts share the advantages of other ES forecasts. They are applicable to a wide variety of projects and they focus on time—one of the key measures of project success. LP forecasts uniquely offer the following benefits.
Pro:
- Improved Accuracy. LP addresses the key concern that has been raised about ES metrics. LP forecasts avoid potential pitfalls associated with false and misleading scenarios, especially if they occur at the beginning of a project. ES forecasts have already been shown to be superior to other types of duration forecasts. LP forecasts have, in turn, been shown to be superior to standard ES duration forecasts. So, LP forecasts represent the “best of breed”. Where forecast accuracy is paramount, LP forecasts are the best fit.
- Ease of Application. LP metrics function like standard ES metrics. The techniques and thresholds used for standard ES metrics can be used equally for LP metrics. Burndown charts, performance indexes, and duration forecasts function in the same way for LP that they do for standard ES metrics. Once you have LP metrics, they are easy to apply.
Con:
- Calculation Complexity. LP calculations are complex. Consider the steps required to produce LP metrics:
- Exhaustively identify all serial paths.
- For each series, identify and remove void periods.
- For each series, calculate the ES and SPIt.
- For each series, calculate the LP duration forecast.
- For each series, factor into the forecast void periods.
- For each period, select the longest forecast from among all of the paths.
- Identify anomalies and replace them with non-anomalous forecasts.
In real schedules, the complexity in any one of these step makes manual calculation of the project’s Longest Path infeasible. Automation is necessary.
- Scope of Metrics. Although the scope of metrics covered by LP is wide, it does not encompass all ES metrics. In particular, it appears that LP does not measure Schedule Adherence. Walt Lipke has stated that “ES(L) cannot be used in identifying tasks performed out of sequence …because it is not the point on the PMB where PV = EV”.* Schedule adherence (and the associated metric, P-factor) will be addressed in a future post. For now, it is enough to note that schedule adherence appears to be outside LP’s scope.
Note
*The Schedule Adherence quote is from Walt Lipke, Personal Communication, 3/1/2016.
Reference
Lipke, W. (2014b). Testing Earned Schedule Forecasting Reliability. PM World Journal, III (VII). |
