Project Managers have many responsibilities. And one of the most difficult is Project Estimation: to predict what will happen in the future. Because projects are fraught with uncertainty. One of the most widely used approaches to improving estimation is the PERT Method.
And Project Sponsors dislike uncertainty. So, they expect Project Managers to reduce it. They want you to manage project risks and estimate uncertain future outcomes like:
- How much will this project cost?
- When will this project finish?
In this article, we will look at:
- Predicting the Future
- PERT Estimation
- Monte Carlo Simulation
- But aren’t PERT Estimates Predictions?
- The ‘So what?’
For a more general introduction to project estimation, we recommend:
The PERT Method and Predicting the Future
Many project managers respond to these kinds of uncertainties by predicting the future. Project predictions are one way to estimate the future. Predictions are also called deterministic estimates. They give a single number to represent a single project uncertainty.
Think of a typical, traditional project…
The project schedule sequences all the activities that the project team must complete. The project sponsor will want to know how long the project will take to finish, and how much the project will cost?
To answer those questions, the Project Manager will estimate the duration and cost of every activity on the project schedule. These are predictions. Using them, the project manager can add all the activity costs to find the total project cost. Then, using scheduling software, they can learn the project’s finish date.
Project Managers use scheduling tools like Microsoft Project to build their project estimates. These tools can handle both costs and durations.
Unfortunately, there are several problems with predicting the future.
Problem 1: Likelihood
First, it isn’t clear to anyone who views a project prediction how likely that prediction is to occur. Is the prediction an optimistic view of the future? Or a pessimistic view? Is the prediction a most likely outcome, or the expected (average) outcome?
Most project uncertainties have more than one possible outcome. And project predictions are only one outcome from among many. For instance, let’s consider an activity called, ‘Write business requirements’. A Project Manager might predict this will take 80 hours for the business analyst to complete.
But, could this activity take less time than 80 hours? Could it take more than 80 hours? Eighty hours may be a ‘most likely’ outcome, but there is a range of possible outcomes. Some of them are both possible and probable, and some of them are possible… but improbable.Single-point #project #estimates fail to capture the likelihood of being right. via @OnlinePMCourses Click To Tweet
Problem 2: Stakeholder Expectations
The second project prediction problem happens when you do not align expectations among stakeholders. No one knows how likely a project prediction is to occur. So, each person can attach their own meaning to what they see.
Let’s go back to our 80 hour task: ‘Write business requirements’.
One person sees it and thinks, ‘That task will most likely finish in ten business days, but it might take longer.’
Another person sees the same task duration and thinks, ‘That task should take no more than ten business days to finish.’
Those are two very different perceptions! Project Managers work to align stakeholder expectations. But project predictions make that difficult, if not impossible.
Problem 3: Poor Decision-Making
Decision-making is an important topic to OnlinePMCourses. And project predictions can lead to poor business decisions by organizational leaders. If project predictions are too optimistic, projects will exceed their schedule and budgets. If project predictions are too pessimistic, projects will appear too long and too expensive.
How can you and your stakeholders know how reliable your project predictions are? It’s time to move to a method that can deliver structured project estimation.Single-point #estimates make for poor #project decision-making. via @OnlinePMCourses Click To Tweet
The United States Navy faced these same issues as you do, back in the 1950s. They were developing the nation’s first, submarine-launched, intercontinental ballistic missile. Project Managers working on the Navy’s Polaris missile program knew that task duration has many possible outcomes. Some were probable and others not.
To create a schedule for the Polaris program, the Navy developed a new method. The Program Evaluation and Review Technique (PERT Method) creates a risk-adjusted schedule. This considers both probable and improbable project outcomes for task duration.
PERT can also estimate project costs, in the same way. Sixty years later, the PERT formula is still widely used today, for project estimation.
The PERT Formula
Here is the PERT formula:
In this formula, we make three estimates of a task’s duration, to account for the uncertainty:
- Minimum is the shortest duration possible
- Most Likely is the most probable duration.
This is your best estimate of the duration. It would be your prediction, in a single-point project estimation.
- Maximum is the longest duration (reasonably) possible
You can use the same formula for costs.
Writing business requirements for a project is most likely to take 80 hours to complete. But it could be finished in as little as 60 hours, and it might take up to 120 hours to complete. What would be the PERT estimate for this task?
(60 + 4×80 + 120) / 6 = 83.33 hours
The PERT estimate is the expected value for the task ‘Write business requirements.’
In this case, it is 83 hours, 20 minutes. By ‘expected value‘, we mean an estimate of the average result. If we carried out the task many times, the average duration would be 83.3 hours.
PERT Statistics in Project Estimation
Let’s look at the project cost using a PERT statistical probability curve.
If you are interested in math, the PERT method is based on a statistical distribution called the ‘beta function’.
In a skewed bell curve like this, the PERT method average will usually be near the peak of the curve. But many other outcomes are possible between 60 and 120. The more asymmetric the curve; the further the expected value will lie from the most likely value.
There’s something that many project managers – even experienced ones – don’t realize. A PERT-created estimate is only about 50 percent reliable.
This is not a flaw in the method. It is precisely the nature of robust project estimation. The best possible estimate must be one that is equally likely to be too high or too low.Good #PERT #project #estimates are 50% reliable… by design. Equally likely to be too high, or too low! via @OnlinePMCourses Click To Tweet
So, if you use the PERT Method well when you create task duration estimates on your project:
- Each task has about a 50 per cent chance of finishing on-time (or early), and
- Each task has about a 50 per cent chance of finishing later than expected.
The same is true if you use the PERT Method for project budget estimation.
You’re probably thinking, ‘I wouldn’t want to estimate my project tasks so they’re only 50% reliable!’
Here’s the rationale behind PERT estimation. In theory, for every task that finishes later than expected, there will be another task that should finish earlier than expected. These off-set one another.
In statistics, it’s called the ‘Central Limit Theorem’. Late-finishing tasks offset early-finishing tasks, so the project ought to finish right on time.
As a theory, that sounds great. But in the real world of project management, that never happens with PERT estimates.
Tasks very rarely finish earlier than expected. But they often finish later than expected. Oops.Project tasks very rarely finish earlier than expected – via @ONlinePMCourses Click To Tweet
Contingency in Project Estimation
So, if you use the PERT Method to estimate task durations, you also need to create a schedule contingency. This is a buffer of time that you add to the project schedule. It acts to protect the project schedule. So, even if some tasks take longer than their PERT estimates, the project will still finish on time.
One excellent approach to this is the Critical Chain method…
In project estimation, allocate more contingency to the tasks that have the most uncertainty. Schedule contingencies usually add between 5% and 15% extra time to a project’s total duration. The same is true of budget contingencies.
Part of your project estimation challenge is to add enough contingency but not too much. The contingency amounts you add need to reflect your assessment of the uncertainty of your project.
So, greater uncertainty means more contingency. But, at the same time, you need to avoid the charge from stakeholders of ‘padding your project’.
Monte Carlo Simulation: Replacing the PERT Method?
The U.S. Navy doesn’t use the PERT Method any longer. And it’s not alone.
Instead, the Navy employs more sophisticated ways of creating project schedules. This includes using Monte Carlo simulation.
Monte Carlo simulation is a way to simulate the execution of a project thousands of times. This, of course, means using a computer. The name comes from the gambling tables that Mont Carlo is famous for. Each simulation assumes a new random set of outcomes for each duration or cost. The calculations to create the random outcomes follow a distribution of likelihoods. The curve in the image above shows a typical distribution.
So, computers can simulate a project running thousands of times. This gives Project Managers a forecast for their project’s duration and cost. It also gives ranges and likelihoods for each point in the range.
But even this kind of project estimation could still be ‘wrong.’
In fact, it probably will. Think of weather forecasting. A weather forecaster may say there’s a 90% chance of rain tomorrow. But what if it doesn’t rain tomorrow? Does it mean the weather forecaster was wrong?
No. By saying there was a 90 percent chance of rain tomorrow, the forecaster implied that there was a 10% chance that it will not rain tomorrow. Project forecasting acts in the same way.
The Advantages and Disadvantages of Monte Carlo Simulations for Project Estimation
Monte Carlo simulation is a more accurate way to estimate your project schedule and budget. But it’s also more time-consuming and expensive. And it takes special software and the skill to use the software properly Plus, of course: ‘GIGO: Garbage in: Garbage out’. The quality of your final estimates is determined by the quality of each of the many assumptions you feed into the software!
For these reasons, most Project Managers today only use Monte Carlo analysis in limited contexts. More often, they prefer the simplicity of PERT to create risk-adjusted task duration and cost estimates for their projects. Then, they add a schedule contingency to protect the project schedule from tasks that run late. And they add a cost contingency to protect their budget.
But aren’t PERT Estimates Predictions?
Hey, wait a minute! Aren’t PERT estimates single-value, deterministic estimates?
Yes, they are! PERT estimates are predictions. But they have several advantages that make their use attractive to project managers.
- First, a PERT estimate is an average result, also called the expected value of an uncertainty. Project Managers should ensure their stakeholders understand this.
- Second, average results are about 50% reliable. So, there is a good chance that a task may take longer and cost more than its PERT estimate. Therefore, project managers need to add a schedule and cost contingency to their projects. Again, you need to help your stakeholders understand this.
- Third, PERT estimates account for many possible outcomes. The PERT formula starts from a three-point estimate (minimum, most likely, maximum). But beneath this, the PERT Method implicitly uses an asymmetric bell-shaped probability curve to model uncertainty. So, the PERT formula generates a value that is an average result for the uncertainty.
- Fourth, PERT is easy to calculate and easy to understand. You can create a PERT estimate using pen and paper.
- Finally, the PERT Method is flexible. You can weight the terms to give greater or lesser prominence to the three parts of your base estimate.
In this general PERT equation:
- If you are highly optimistic, you can make the multiplier (L) for the minimum estimate higher
- If you are highly confident of your Most Likely Estimate, you can increase that multiplier (M)
- Or you can reduce the value of M if your confidence is low
- And, if you want to present a cautious estimate, you can increase the multiplier (N) on your maximum estimate
The commonest formula, which we saw above, takes:
L = 1, M = 4, N = 1, L+M+N = 6
For a more pessimistic formulation of PERTT, you might use:
L = 1, M = 3, N = 2, L+M+N = 6
The ‘So what?’
Project estimation is a challenging exercise for even the most experienced project manager. So, it’s helpful to be familiar with an assortment of project estimation techniques and tools.
Project predictions are deterministic (single-value) estimates. They don’t convey a sense of risk or how reliable they are.
The PERT method is an easy way to create an expected value that takes account of project uncertainty. But you do need to remember that, even so, PERT estimates are still only about 50 percent reliable.
So, when you use PERT for project estimation, be sure to include contingency as well. This creates an extra buffer in your project to protect its schedule and budget.When you use the #PERT Method for #project #estimation, always include a contingency. via @OnlinePMCourses Click To Tweet
For more about project estimation, we recommend:
What is your experience with project estimation in general, and the PERT Method in particular? Please share your thoughts in the comments below. As always, we’ll respond to every comment.