Concept: Walt Lipke’s work is a rich source of metrics for Schedule Adherence (SA). Close examination of his work reveals hidden featuresthere are SA metrics supported by his work that have not yet been uncovered.
This post exposes some of the features as a first step toward additional SA metrics.
Practice: Let’s start with rework (R) and the index for its performance, the Schedule Adherence Index (SAI).
In Walt’s articles on rework [1] , R appears to be used in two subtly different ways:
 R is used as a label for tasks in which EV@AT – PV@ES is positive. Let’s call them the “Rtasks” (Figure 1, tasks in bold).They are out of sequence and are likely to entail rework.
 R is also used to represent the amount of value at risk of rework, i.e., the unusable portion of the value earned by R tasks. The amount is stimated using a mathematical model.
Figure 1
Rtasks are likely to entail rework because when work is done prematurely, [2] knowledge gaps are inevitable. To bridge the gaps, performers make assumptions about missing inputs. At some point, the missing information becomes available, and when it does, it often differs from the assumptions. That means what has already been done must be reworked.
The amount of unusable value is a function of schedule adherence, or, more precisely, schedule nonconformance. It is, strictly speaking, a function of 1 – P: that is, 1 less the measure of conformity. [3] That amount must be qualified further because it is affected by the remaining work: as the number of tasks drops, the pool available for rework shrinks. So, Walt proposes a mathematical model that adds a pattern of decline to the calculation. [4]
R, alone, does not offer a reliable indication of performance. As just mentioned, it depends on 1 – P. But, P by definition ends at 1.0, when the total EV = total PV. Inevitably, therefore, R drops to 0 at the end of the project. This behaviour makes it unreliable. As the project proceeds, the PM does not know if any improvement in R is due to remediating actions or to the inevitable rise in P.
To compensate for this effect, R is normalized to the remaining work. The result is the SAI. To sum up:
 R, as the volume of rework, is a quantifiable, discrete amount that is normalized to the remaining work, yielding the SAI.
 The SAI reliably reflects the impact of rework over the whole project timeline.
Next, let’s examine P, where there appears to be a parallel to the points made for R.
P covers two distinct sets of tasks:
 Tasks that have value planned as of ES where all the PV is earned as of ES.
 Tasks that have value planned as of ES where not all the PV is earned as of ES. These cases represent delays in the schedule. They are termed Impeded/Constrained tasks, or simply I/Ctasks (Figure 2, tasks in bold).
Figure 2
The tasks that have both planned value as of ES and earned value as of ES are included in the calculation of P. But, this means that these tasks are not included in the Rtasks, where the value is planned and earned after ES and as of AT. [5]
A potential subset of these tasks includes tasks that have planned value as of ES and have some, but not all, of the corresponding earned value. For these tasks, EV@AT – PV@ES is negative.[6] Such tasks are impeded or constrained because they have not fully achieved their planned value. Hence, they are termed the I/Ctasks.
P deviates from 1.0 solely because of the I/Ctasks. This takes some explanation.
The PV and EV for all tasks up to and including the ES time must fall into one of two categories: the PV and EV match for each task, or the PV and EV do not match for each task. If they match, P is 1.0. If they do not match, P must be less than 1.0.
The only way for this to occur is for EV < PV. Reason: In a given baseline, the PV for a task cannot drop from one period to the next. So, either the PV stays the same or it increases. In either case, a mismatch can occur only if EV is less—otherwise, it would match. If the EV<PV as of ES, then EV<PV as of AT, and that is the definition of I/C.
So, P varies solely based on the I/C tasks.
To parallel the consequences for R, it follows that:
 The volume of I/C tasks is a quantifiable, discrete amount that can be normalized to the remaining work.
 If so, there is an indicator that reflects the impact of I/C over the whole project timeline.
The next post explores the implications of these conclusions.
Notes:
[1] See: Lipke, W. (2011) Schedule Adherence and Rework. The Measurable News, Issue 1 (corrected version).
[2] It is important to keep the formal definitions in mind as they can conflict with ordinary parlance. To be specific, work is done “prematurely” when the value is produced after ES and as of AT. That’s because ES represents the time at which the value currently produced should be produced. Ordinary parlance is not so restrictive. For instance, value earned later than AT would ordinarily be considered early. But, that is not so, by the theory. And, value earned as of ES that results from earlier, incomplete input would be considered as a candidate for rework, but again that is not so according to the theory. There, the incomplete input would be considered I/C but its successor would not be called out as an Rtask. Such deviations between formal definitions and ordinary usage are often fodder for critics, but they are not definitive. For that, critics would need to adduce logical inconsistencies within the theory.
[3] For the math behind this claim, click here.
[4] Ibid.
[5] The value earned as of AT is counted up to and including the ES time. If ES falls before AT and value occurs in the gap, it is not included in calculation of P. Instead, it becomes part of the calculation of R. Note that work done after AT is not included in the calculation of P; so, if ES falls after AT and value occurs in that gap, the negative difference between EV@AT and PV@ES is not included in the calculation of P. That is because P measures adherence up to and including AT, but not beyond it. By the way, that also explains why the maximum value of P is 1.0—the EV can only be up to the same amount as the PV at AT—measurement does not go beyond that point for P.
[6] It should be obvious that for the tasks with EV@ES and PV@ES, the difference (ES@ES – PV@ES) is 0.
References:
Lipke, W. (2009). Schedule Adherence…A Useful Measure for Project Management.The Measurable News, Issue 3.
Lipke, W. (2012). Schedule Adherence and Rework. CrossTalk, NovemberDecember.
Lipke, W. (2011b) Schedule Adherence and Rework. PM World Today, July.
Lipke, W. (2011a) Schedule Adherence and Rework. The Measurable News, Issue 1 (corrected version).
