Statistical PERT – Beta Edition Winter, 2016

The tagline for Statistical PERT is “Estimation Made Easy.”  Dealing with probabilities isn’t easy, but using Microsoft Excel and the simple concepts behind Statistical PERT, anyone can make probabilistic estimates for bell-shaped uncertainties.

Unlike PERT, which is based on the beta probability distribution, Statistical PERT uses the normal distribution.  As long as the uncertainty has bell-shaped properties, this is absolutely not a problem.

Below is a graphical representation of a normal distribution; it has a standard deviation of 7, a mode of 50, and 95% of the area is between 36.3 and 63.7: Below is a graphical representation of a beta distribution; it has a standard deviation of 7, a mode of 50, and 95% of the area is between 36.3 and 63.7: Can you tell a difference?

No.  There is no easily discernible difference in shape of these two precisely drawn probability curves, although the bell-shaped curve created by the beta distribution has hard-stop anchors of 0 and 100, whereas the bell-shaped curve above it has infinitesimally small probabilities that are smaller than 0 and greater than 100.

Using Statistical PERT with bell-shaped uncertainties will yield very, very accurate results even though Statistical PERT was built with Microsoft Excel’s normal distribution functions, NORM.DIST and NORM.INV.  Even bell-shaped curves that are skewed but still bell-shaped can be used with Statistical PERT to create very accurate estimates.

But what if your uncertainty does not have bell-shaped properties to it?  What if it has a uniform shape, or a triangular shape, an extremely skewed bell shape, or some other shape besides a bell shape?

That is the premise I’m using for a very exciting, new addition to Statistical PERT!

I’m “in the lab” right now working to create a different edition of Statistical PERT using the beta distribution.  This opens up using Statistical PERT with a wider variety of uncertainties, and it allows for more-accurate estimation than using Statistical PERT beyond just the normal distribution.

Using the beta distribution is inherently difficult to do, though.  Microsoft Excel’s BETA.DIST function requires two arguments called alpha and beta, and working with them isn’t intuitive.

So my work between now and this winter is to create Statistical PERT – Beta Edition.  I want to create a simple, easy way to leverage the beta distribution without the estimator having to take a statistics class, and without having to worry about the shape of a beta distribution.  The resulting new Excel templates for SPERT Beta I’ll create will still be free to download, free to use, free to modify, and free to redistribute, without having to register on the site, without adware, without worrying about malicious macros embedded in the spreadsheet file.