- In this video, I want to answer the question, what is a PERT chart? PERT stands for program, evaluation, and review technique. And as a methodology, it dates to the late 1950s and was invented by the US Navy and consultants Booz, Allen, Hamilton. And it's based in a form of network chart. A PERT chart is just a proprietary brand of network chart.
PERT charts are less frequently used than their more commonly seen sibling, the critical path method. The critical path method uses activities on nodes. And those activities are connected up by lines. The PERT technique flips it over, and it shows the activities on the connecting lines. And then the nodes represent the branching structure of the network chart.
In many ways, the two are pretty much the same. It's just a stylistic approach to how they're represented. So if PERT charts have activities on arrows, is there a substantive difference between the PERT method and the critical path method? And the answer is yes, because the second difference is of much more substance.
The critical path method uses a much simpler single point estimate for the duration of an activity. You estimate how long the activity is going to take, and that's the number that goes into the model.
In PERT, we actually use a three point estimate for every duration. So we estimate what we consider to be the most likely duration, which is the estimate that would have gone into the critical path method. But we also make a pessimistic estimate and an optimistic estimate. The pessimistic estimate is the longest we think, plausibly, this task is likely to take. The optimistic, the fastest we think we are likely to be able to do the task.
And typically, these estimates are not symmetric, in the sense that the optimistic estimate is not the same distance from the most likely estimate as the pessimistic estimate. So this gives us a kind of triangular distribution.
As a result of some simple mathematics, we can therefore estimate an average duration for the task, based on our three point estimate. And the way that we do that is, we average the three estimates-- the optimistic, the pessimistic, and the most likely.
But we don't give these equal weighting, because there's a reason why the most likely is the most likely. So we give it a higher weighting. Typically, the method suggests that we multiply both the optimistic and the pessimistic by 1, and we multiply the most likely by 4. We add them all together and divide by 6.
That's not the only way we could do it. We could take a highly pessimistic view and multiply the optimistic by 1, the most likely by 4, and the pessimistic by 2. And divide by 7. Or we can increase the weighting on the most likely.
We can also calculate the standard deviation, get some sort of idea of the spread of likelihoods around our average. And the way that we calculate that is, we subtract the optimistic estimate from the pessimistic estimate. And then we divide the answer by 6. And that gives us a plus or minus on our average.
So the PERT methodology is a way of estimating the duration of projects, firstly by building a network where the convention is that we use the arrows to define the tasks, and then the nodes show how the arrows relate to one another. And secondly, the lengths of those arrows, the durations that are recorded on those arrows, are based on three point estimates from which we can calculate average durations and standard deviations.
And of course, with today's modern software, it's very straightforward to put in all of your estimates for all the tasks, and for the software to calculate not only an average duration for the whole project, but to composite all of the standard deviations up and figure out the likely spread of from the most optimistic to the most pessimistic estimate for the whole project.
So the PERT method should be used more than it is. But let's not be seduced into thinking that just because we're using three point estimates, that any one of those three points is any better an estimate than the one that goes into our critical path method.
Good PERT analysis requires that do good estimating. But that is a topic for another time.