Concept: Some project topologies degrade the reliability of ES metrics early in a project’s timeline. ES Longest Path (ES-LP) alleviates the difficulty.
Table 1
Practice: Walt Lipke’s Longest Path (LP) approach to duration forecasting avoids problems that beset other EVM (and some ES) forecasts. The way LP is calculated helps explain the technique and suggests how it can be practically applied.
From the outset, it must be admitted that LP calculations are more complex than other ES calculations. The first step is to identify all serial paths in the schedule. That task alone is daunting. In our experience, it is virtually impossible to identify all of the paths without automation. If done manually, some paths are invariably overlooked.
Once the paths are identified, the next step is to calculate a duration forecast for each serial task for each time period. The forecast calculations require several steps.
The Performance Measurement Baseline (PMB) used for the LP calculation is different from the one ordinarily used for ES calculations. It does not depict the actual pattern of Planned Value (PV) and Earned Value (EV). Instead, it posits an idealized pattern for PV and EV.
Here is an example that illustrates the approach. For path 7-10 from Table 1, the pattern of values used for the PMB is the following (where XX = void period):
Table 2a Table 2b
The idealized distribution eliminates void periods. The PV and EV start at the first period and proceed without interruption to the final period for each. There are no gaps. Based on this allocation of PV and EV, the ES and SPIt are calculated using customary equations. The next step is to calculate the duration forecast itself, called the LP Estimate at Completion for time (LP EACt).
Here is where void periods come into play. In a real schedule, many tasks start after the first time period (for an example, see Table 2b). It is also possible to have Down Time (i.e., interruption of PV) and Stop Work (i.e., interruption of EV). In Table 2b, there is a period of Down Time after the PV hits 50. All of these gaps create offsets that must be factored into the LP forecast calculation.
Table 1 illustrates the result of the LP EACt calculations for a sample case (Lipke (2015), p 34). Once the LP EACt calculations are done, the longest duration is selected for each time period from among all of the paths. In Table 1, the selections are identified by green highlighting.
At this point, the calculations have another twist. There can be anomalies in the LP EACt. Anomalies distort the standard deviation of the forecasts, and the distortion undermines the reliability of the forecast.
Take the forecast in period 3, for example. Normally, the standard deviation is constant with respect to a mean value. In Table 1, forecasts hover close to a deviation of .466. In period 3, however, the deviation jumps to 11.83 (Lipke (2012), p 7). The outlier casts doubt on the reliability of the forecast.
Fortunately, anomalies can be systematically identified and removed from the calculation. A decrease in the representative ES from one period to the next indicates that an anomaly has occurred and that an adjustment must be made. What is the “representative ES”?
Think of it this way. For LP, ES calculations are initially applied to the idealized pattern of PV and EV. They exclude void periods and so, represent the purely executable part of the schedule. By contrast, ES for the Longest Path, called ES(L), takes voids into consideration. It also aligns all path calculations to a common end point: the Planned Duration (PD).
The LP EACt is the key to achieving these ends. As already mentioned, the LP EACt is adjusted for void periods. And, it can be used to normalize the calculations to the PD. To normalize ES(L), the PD is multiplied by a common factor: the Actual Time (AT) divided by the LP EACt.* Thus, the ES(L) is calculated as:
ES(L) = PD * (AT / LP EACt). (see Lipke (2015), p 34)
Anomalies occur when the ES(L) drops from one period to the next. In those cases, the project’s overall EV increases, but the EV for particular paths decreases. The time periods in which such a decrease occurs are excluded from consideration for the LP.
Here are the ES(L) values associated with Table 1:
In Table 3, period 3’s ES(L) is 1.05, a decrease from the 1.48 of period 2. Thus, period 3’s forecast was not included in the LP. Instead, the next lowest ES(L), from path 3-8-10, is selected, implying that the corresponding forecast, 12.00, is the LP EACt for period 3.
Lipke puts the rule as follows:
LP is chosen as the longest forecast having a positive change in ES(L). (Ibid)
In conclusion, calculation of the LP is itself a long path, and the complexity of the calculations has an impact on the application of the technique. The next posting will address how to apply LP to your project.
Note
*By comparison, the following formula can be derived from basic ES equations: ES = PD * (AT / EACt).
[Edited 22 March 2016]
References
Lipke, W. (2012). Speculations on Project Duration Forecasting. The Measurable News, III, 1, 4-7.
Lipke, W. (2014a). Examining Project Duration Forecasting Reliability. PM World Journal, III (III).
Lipke, W. (2014b). Testing Earned Schedule Forecasting Reliability. PM World Journal, III (VII).
Lipke, W. (2015). Applying Statistical Forecasting of Project Duration to Earned Schedule-Longest Path. The Measurable News, II, 31-38.
Vanhoucke, M., & Vandevoorde, S. (2008). Earned Value Forecast Accuracy and Activity Criticality. The Measurable News, Summer, 13-16.
Vanhoucke, M., & Vandevoorde, S. (2009). Forecasting a Project’s Duration under Various Topological Structures. The Measurable News, Spring, 26-30. |
