Are you striving to be 75%, 85% or 95% confident in your task estimates?
How do you know when you have that level of confidence?

Recently, I surveyed fellow IT project managers in my organization.  I asked them what confidence level, expressed as a percentage, do they estimate their tasks?  In other words, what is the probability that their estimates will be equal to or greater than the actual time it takes to finish the task?  So if someone said, 80% confidence, they’re implicitly saying that there is a 20% chance that the estimate will be exceeded by the task’s execution phase.

The responses I got were varied and a little surprising.  One person wrote they estimated for 20-50% confidence, whereas another wrote the estimated to get 100% confidence (wow!).  Some favored 80% confidence, others favored 95% confidence.  As I reviewed the responses, I thought, “I wonder what the IT management team expects?”  Or the project sponsor?

Then I thought about this some more.  How would you even know if you HAD 85% or 95% confidence in an estimate?

One thing led to another, and over a weekend I developed a spreadsheet that I call “Statistical PERT” or SPERT.  What is SPERT?  It is a statistics-based, risk-influenced method of task estimation.  The only thing SPERT shares with the Program Estimation and Review Technique, or PERT, is (besides the name) PERT’s use of a 3-point estimate (optimistic, most likely, and pessimistic).  SPERT replaces PERT’s simple estimation formula with the use of the normal distribution curve, which is similar in shape to the PERT distribution (a special form of the beta distribution used with PERT).

With SPERT, a task estimator gives a 3-point estimate, just like with PERT, but does one thing more.  The estimator provides a subjective judgment on HOW likely the “most likely” point-estimate really is.  For a familiar task, you’d expect the estimator to say something like, “I have high confidence in the most likely outcome.”  Conversely, for a task involving new technology or an unproven, 3rd party vendor, you’d expect something more like “I’ve got a rather low confidence level in my most likely outcome.”

For SPERT, those subjective judgments are translated into the use of ratio scales that create a standard deviation that modifies the normal distribution curve leading to useful probabilities about the likelihood that a task estimate will not be exceeded during actual task execution.  Whew!

SPERT is easy to use.  Plug in your 3-point estimate and pick a level of confidence for the “most likely” outcome.  That’s it.  The SPERT spreadsheet develops probabilities based upon that information.

Email me to learn more!

Note:  This blog post, originally published in October, 2014, marks the birth of Statistical PERT.  Statistical PERT has grown up a lot since then, having been developed into a robust Excel spreadsheet which you can download from the Statistical PERT website:  https://www.statisticalpert.com.  Statistical PERT has been presented at international conferences, like the 2016 PMI Global Congress in North America.  Statistical PERT is a freely licensed, probabilistic estimation technique that can be used to estimate any uncertainty with bell-shaped properties.  Want to learn more?  Visit my active blog devoted to Statistical PERT and probabilistic estimation in general, or just download a free Statistical PERT template to estimate task duration, expenses, revenue, agile sprints, project portfolios, or event attendance.