Past As Predictor And Earned Value - Part 1
Some Observations by Bill Duncan William R. Duncan Primary Author of the 1996 version of "A Guide to the Project Management Body of Knowledge" and creator of the "process model" of project management.
The Earned Vallue technique was NOT developed by the USDoD in the 1950s and 1960s. It was developed by some Industrial Engineers way back in the 1930s as an alternative to simple budget-vs-actual reporting. Although there wasn't much of a formal discipline of project management in those days, there obviously were projects -- much of President Roosevelt's economic recovery program was driven by federal spending on major construction projects. EV remains superior to BvA.
We should be careful to separate the USDoD version of this technique from
the basic concept. Lumping all forms and implementations of the technique
into one bucket is dangerous
and confuses the issue. The USDoD approach was designed for MAJOR weapons
procurements. The last time I checked, the USDoD procedures exempted a variety
of projects from its EVM requirements, including what it considers smaller
projects and most R&D projects.
The USDoD maintains an extensive database of historical results. Their results
show that if a project's CPI is 0.90 or worse (i.e., 10% over budget) after
15% of the budget has been
expended (i.e., 1/6 of the way through the project), the project's CPI never
improves. This result is often misquoted as "the CPI never gets better."
In addition, a large percentage of the USDoD database includes cost-plus projects
where there is limited incentive for the contractor to improve the CPI. In
my opinion, these results are illuminating but not generalizable.
The only other detailed study that I have seen comes from an Israeli electronics contractor. Their results (covering about a dozen projects in the late 1980s and early 1990s) showed that their CPI was a reliable predictor of total project costs after 25% of the project's funds had been expended. I consider these results more generally applicable because these projects were done under a fixed price contract.
Although Critical Path Method ctivities with float can appear to distort
the numbers, I maintain that these kinds of variances still require analysis.
If a non-critical path activity is behind schedule, the experienced PM generally
wants to know why: presumably there was a reason for scheduling that activity
at that particular time. If it is not on track, the delay could propagate
throughout the
project and cause problems. There is nothing that says the PM must implement
corrective action in response to every variance, but someone should certainly
be looking at each variance to assess the need for action.
There are a variety of ways to use EV data to project the project's cost
at completion. A simple straight line projection is recommended ONLY when
the project management team feels
that the project's future activities are "similar" to its past activities.
"Similar" could mean that it is the same kind of work, was estimated
in the same way, will be performed by people with similar
skills, or has similar degrees of uncertainty. Both "A Guide to the Project
Management Body of Knowledge" and the USDoD EVM guidance documents explain
a variety of alternative projection techniques and detail when each is appropriate.
EVM is more sophisticated than simple
straight line projection. It also relies heavily upon trends rather than simple
point-in-time projections. Trend analysis improves the reliability of the
projection.
No process is perfect. No intelligent, rational person expects that EV projections or any other kind of projections will always be able to predict the future exactly and precisely. Remember that the term is ESTIMATED cost at completion. But that doesn't mean that imperfect projections aren't valuable. If I think I'm doing okay, and the numbers say I'm not, I better be able to explain why (or else I better change my mind).
Use of SPI. SPI is less reliable than CPI in predicting future performance
because SPI can almost always be improved by spending more money. However,
an SPI of less than 1.0 does mean that you are behind schedule, and an SPI
that is trending down means that you are getting
farther and farther behind schedule. That information should be useful.
I have yet to see a viable alternative that integrates cost and schedule
performance. Rolling forecast of cost-to-complete can be viable in the hands
of a skilled and experienced PM, but it can also be dangerous. I know of one
commercial software developer that drove itself into bankruptcy because its
rolling forecasts were delusional -- but believed by management. In
this case, the EV projections showed a 400% overrun.
The approach used to be called EV Analysis. Its proponents started calling
it EV Management several years ago because (in my opinion) the USDoD PMs saw
"analysis" as a staff
function. Personally, I prefer the old name because it emphasizes that the
technique only provides information; that it is still up to the PM team to
figure out what to do with the information.
There is no silver bullet. EV is just one tool among many. It can be highly effective when applied in the proper environment (larger projects with relatively stable scope) with knowledge and skill. Once again, let's not confuse the tool with the application. Let's not throw out the baby with the bath water.
About the Author
William
R. Duncan is a principal of Project Management Partners, a project management
consulting and training firm headquartered in Lexington, MA USA. Mr. Duncan
has nearly thirty years of management and consulting experience including
five years with a major international consulting firm. He was the primary
author of the 1994 and 1996 versions of A Guide to the Project Management
Body of Knowledge, the most widely used project management standard in the
world.
Mr. Duncan is a 1970 graduate of Brown University in Providence, RI and has done post-graduate work at Boston University and Northeastern.