Forecasting the final cost and schedule results
There are numerous reasons it might be advisable to employ some minimal form of earned value on any project. Perhaps the single most compelling reason is to enable the project manager to “statistically” calculate the final cost and schedule results on the project…from as early as the 15 percent completion point. Such predictions have been proven to be reliable, and are based on two factors: a detailed bottom-up project plan, and subsequent actual performance against the project plan.
One word of caution on the use of top-down forecasting: it must always be done with care and intelligence. There may well be overriding issues that must be considered. The single most reliable forecasting technique for any project is likely the bottom-up forecast, where every future work task to be accomplished is re-examined and re-estimated, taking into consideration all known factors.
But there is one problem with detailed bottom-up estimates: they take both time and energy to prepare. The very same people who should be performing the project work must stop their effort and re-estimate the remaining tasks. While they are doing these detailed estimates, the project's primary work effort is halted, or at least disrupted. And besides, nobody ever sets aside budget for work called “re-estimating.” These are “freebie” tasks that management simply directs to somehow magically happen: anytime overnight, after-hours, on weekends, etc.
Two performance indices resulting from earned value data are needed to statistically forecast the project final cost and schedule results. The first and most valuable is the Cost Performance Index (CPI), which represents the relationship between the value of the work-in-process and/or the performed tasks, compared against the actual costs for doing such work. Obviously, if one spends more money than one performs in work accomplished, such performance is called an “overrun.”
Over the past three decades, a scientific body of knowledge has been accumulated on the use of earned value data to predict the final cost and schedule results on a project. One individual who has been particularly articulate in disseminating this doctrine cites the following empirical evidence to support the utility of using earned value data to provide estimates at completion:
Given: Contract more than 15 percent complete.
1. Overrun at completion will not be less than overrun to date.
2. Percent overrun at completion will be greater than percent overrun to date.
Conclusion: You can't recover!
Who says: More than 700 major
DOD contracts since 1977.
Why: If you underestimated the near [term planning], there is no hope that you did better on the far-term planning.1
There are two factors that may put into question the application of these findings to all projects in the world.
First, all of the contracts cited were for major systems acquisitions for the United States Department of Defense (DOD). Some people feel that conditions for government procurement are esoteric and do not apply to commercial projects. Perhaps. However, there is another school of thought that suggests that project management fundamental principles apply universally, regardless of industry.
Figure 1. Monitoring Earned Value Performance
Second, and perhaps of greater consequence, most of the contracts cited were of a cost-reimbursable type. We all know that senior management takes a greater interest in the final cost results whenever they underwrite the risks of cost growth, i.e., on any fixed-price or lump-sum type job.
These two factors may well alter the universal application of the cited data. Nevertheless these findings, to some extent, may have broad applications to all projects. Again, such data must be used with intelligence and care.
Final Project Results are Determined by Three Factors
While the earned value performance indices can be most useful in predicting the final results on any project, the validity of these indices are subject to three critical variables. Each of these deserves some discussion because they can have an influence on the usefulness of these indices in predicting the final cost and schedule outcome.
Factor 1: The Quality of the Project Plan. Not all project plans are created equal. Some companies and individuals are quite good at preparing project plans; others are not, and would rather fire first and aim later. The quality of project plans will vary, and will influence the final project results.
The competitive environment under which a given project plan is created will likely influence or bias the project strategy. Competition will typically result in risk taking by the creators of the plan. Any individual providing an estimate of the time or resource requirements for given tasks will typically offer their estimated values reflecting the degree of competition that may exist. Competition for an award will bring in management posturing, and will result in a different standard in the final proposal.
Factor 2: Actual Performance Against the Plan. Once the project plan has been approved and implemented another important variable will come into play. That is the actual performance factor, or achievement rate against the plan. How has the performance been against the plan? Is the project meeting, exceeding, or falling behind its own plan, and at what performance rate? Such values can be quantified and monitored for the duration of the project's life cycle.
A cost and schedule performance factor, once established, can be used to statistically forecast the final results for the project employing earned value.
Factor 3: Management's Determination. The third factor is also most critical and can alter the final project results. If a given management is closely following its earned value performance to the authorized plan, and does not like, or cannot accept the final forecasted results, to what extent will management take aggressive actions with the remaining work to alter the final outcome?
Final project results are not necessarily preordained. Final project results can be altered…but only with aggressive management action. The critical issues are: (1) to what extent will project performance data be monitored? (2) to what extent will it actually be believed by management? (3) what actions will be taken to alter the life style of the project for the remaining work? and finally, (4) to what extent will all discretionary tasks be eliminated, budgets reduced, risks taken to bring the final projected results down to acceptable levels? Aggressive management actions can alter the final projected outcome.
Forecasting Final Cost Results
As was emphasized in our previous article (“Monitoring Performance Against the Baseline,” September 1995 PM Network), the performance for any project that employs the earned value concept allows it to measure actual performance against two standards: planned schedule and cost efficiency.
The first measurement is actual performance against a planned schedule, or the accomplishment of the work scheduled. Did the project accomplish what it set out to accomplish…and in the same time frame? This is a value that can be quantified by a 1.0 standard, as is displayed in Figure 1.
The second and related issue is that of cost performance. Cost performance can also be measured with use of a 1.0 standard, representing the relationship of work accomplished versus the actual costs being spent. This is also displayed in Figure 1, and the formula for the two standards are displayed at the bottom of the figure. While over the years there have been numerous methods suggested for statistically forecasting the final cost results on a project using earned value data, most center upon a single formula. They typically focus on the performance of three variables, as of a point in time in a project duration:
1. The value of the work remaining (measured as the total project budget, less the earned value actually accomplished);
2. The work remaining divided by a performance efficiency factor (example: the cumulative CPI, or the SPI, or some combination of the two indices);
3. Plus the sum of actual costs incurred. Likely the most common and respected of all such statistical methods is displayed in Figure 2, The “Cumulative CPI” Estimate at Completion. The formula for this method is displayed at the bottom of the figure, and incorporates the three variables described above. This statistical formula has the most scientific data to support its reliability.
Note the use of only cumulative, not periodic, incremental data is suggested. Periodic data is too subject to anomalies caused by placing good data into the wrong time frame. Cumulative data smoothes out variations, but nevertheless retains its value as a long-term forecasting tool.
The cumulative (not monthly) CPI is a particularly reliable index to watch because it has been proven to be an accurate and reliable forecasting device. The cumulative CPI has been shown to be stable from as early as the 15–20 percent point in a project's percent completion point.
One scientific study described the value of using the cumulative CPI: “…researchers found that the cumulative CPI does not change by more than 10 percent once a contract is 20 percent complete; in most cases, the cumulative CPI only worsens as a contract proceeds to completion.2 Some people consider this cumulative CPI formula to represent the “minimum” that the project will likely do. Others consider it to be the “most likely” final estimate of project costs. Either way, the cumulative CPI has been proven to be most accurate in providing a statistical forecast of the final project costs. Such forecasts should not be ignored by any project team.
Figure 2. The “Cumulative CPI” Estimate at Completion (EAC)
Figure 3. The “To Complete (the work) Performance Index” (TCPI)
Figure 4. The “Cumulative CPI x SPI” Estimate at Completion (EAC)
One of the most effective earned value statistical formulas, which seems to immediately get senior management's attention, is the direct opposite of the cumulative CPI. This formula is called the “To Complete (the work) Performance Index” (TCPI) and is displayed in Figure 3. Its formula is also shown at the bottom of the figure.
The TCPI determines what cost performance factor will be needed to complete the remaining work, according to some management financial goal. It simply takes the “work remaining” and divides it by the “funds remaining.”
Note that the “funds remaining” can be a variable amount. The values for the funds remaining can be set for any financial goal of the project: the project budget, or the latest estimate, or even the fixed price ceiling of a contract, always less the actual costs that have been incurred to date.
The formula for the TCPI is thus:
Work Remaining [Total Budget less
Earned Value] / Funds Remaining
[(1) Budget or (2) Estimate or (3) Ceiling less Actual Costs] = TCPI
Management at all levels is quick to grasp the realities that when you are halfway through the project and have achieved a cumulative performance factor of only .8, then the project must achieve a performance factor of 1.2 for all remaining work in order to stay within its financial goals. At some point such performance challenges becomes impossible, and the realization of the cost overrun has to be acknowledged.
The last statistical formula that has wide acceptance in predicting final project costs is the one that combines both the cost (CPI) and schedule (SPI) efficiency factors. It is displayed in Figure 4, and is referred to as The “Cumulative CPI x SPI” Estimate at Completion.
There is a logical basis to support the use of a formula that incorporates both cost and schedule dimensions. Project people do not like to be in a position of establishing a performance plan, and then falling behind the authorized plans. There is a natural human tendency to want to get back on schedule, even if it means using more resources to accomplish the same amount of scheduled work. Extensive use of overtime and additional resources is often authorized, which simply causes permanent, non–recoverable damage to their cost efficiency factor, their Cost Performance Index (CPI).
Some people consider this formula to be a “worst case” formula, while others consider it to be the “most likely” model. In either case, it is a widely used and accepted formula to statistically forecast the final costs on any project.
Question: What are we attempting to accomplish with the use of earned value performance data to statistically estimate the final costs on a project? Simply put, it is to test the reasonableness of the “official” position of the project manager against a statistical forecast based on actual performance. Earned value statistical forecasts provide management with a sort of “range” of cost estimates at completion, depending on the formula selected. If a given project uses the low-end formula, per Figure 2, and additionally the high-end formula, as portrayed in Figure 4, they have in effect provided a “range” of estimates to compare against other positions.
If the project manager is forecasting the final cost performance of the project outside of the statistical range, then the basis for this position should be explainable to all interested parties. Interested parties may be the owner, senior management, shareholders, anyone who has a vested interest in the final cost results of the project.
Predicting Project Duration
How long will the project take to complete? This is another related issue that is always of concern to a project manager, to senior management, and especially to the owner.
By definition the project will be completed within its critical path. The critical path for any project determines the shortest time frame for its completion. By definition the critical path constitutes “the shortest length of the project.”3 Management of the project's critical path and near critical paths is vital to the success of any project, and its completion in the shortest possible duration.
In addition to the monitoring of the critical path, and as another way of predicting the final duration of the project, the earned value planned schedule data may be compared to the critical path, as a way of predicting when the project will likely be completed. Critical path data and earned value performance data work well in predicting the final completion date for any project.
Figure 5. Monitoring Earned Value (planned) Schedule Performance
Displayed in Figure 5 is a chart that depicts the three essential elements of an earned value performance plan: planned schedule value, earned value, and actual costs for accomplishing the earned value. All three curves are quantified in monetary values. For purposes of this discussion we will set aside the cost issue, which is that relationship between the earned value and the costs being incurred.
Instead we will focus on the schedule variances of the work we planned to accomplish, as of a point in time. As of the project status date, one year into a two-year project, our plan called for us to complete 50 percent of the planned work. But as of the status date we had only accomplished 40 percent of the effort, as reflected in the earned value line.
If we take the point of intersection of the earned value line, move it backwards to the planned schedule value line, then move it downward to the bottom of the time scale line, we can immediately see that we are approximately two and one-half months behind in accomplishing our scheduled work. This approach is displayed with the dotted lines in Figure 5.
The comparison of the earned value schedule variances with the management of the critical path puts fidelity into the prediction of how long it will likely take to complete the project. An accurate assessment of the critical path, plus the management of task float, in conjunction with the accomplishment of all planned work, will help the project manager accurately forecast when the project will likely end.
The bottom line for any project is typically How much will it cost to complete the job? and How long will it take until it is over? Two vital tools to make this determination are available to any project manager: the use of the critical path method, and the earned value performance measurement technique.
While many projects will employ the critical path method to predict and to manage the time frame for their projects, the use of earned value measurement has been somewhat limited. It is our position that the two techniques work well together, and are in fact complementary. When used in concert, by people working from the same integrated database, they can provide accurate predictions of the universal question: How long and how much is the project likely to cost us? ■
1. Studies by Gaylord E. Christle, et al., from the Office of the Under Secretary of Defense for Acquisitions, as documented in the Department of the Navy memorandum of November 28, 1990, p. 6, on the “A-12 Administrative Inquiry,” by Chester Paul Beach, Jr., inquiry officer. This study cited over “400 programs since 1977.” Updates of this same finding presented on September 10, 1991, in Boston, have increased the sample to over 700 programs since 1977 without a change in the conclusions.
2. Christensen, Major David S., Ph.D., USAF, “Using Performance Indices to Evaluate the Estimate at Completion,” Society of Cost Engineering and Analysis, Journal of Cost Analysis, Spring 1994, p. 19.
3. Project Management Institute, Project Management Body of Knowledge (PMBOK), dated September 1, 1987, in the Glossary of Terms prepared by R. Max Wideman.
Quentin W. Fleming is a senior staff consultant to Primavera Systems, Inc., with over three decades of industrial experience. His latest book is Cost/Schedule Control Systems Criteria—The Management Guide to C/SCSC.
Joel M. Koppelman,, P. E. is president, co-founder, and co-owner of Primavera Systems, Inc. He has spent more than a decade managing capital projects in the transportation industry.
PM NETWORK • January 1996