In yesterday’s blog post, I explained why the SPERT-7 Rule equates near certainty in the most likely outcome of a bell-shaped uncertainty with a 7% Ratio Scale Multiplier. In today’s post, I’ll explain why under a condition of great uncertainty surrounding the most likely outcome, the Ratio Scale Multiplier is 42% in the SPERT-7 Rule.
Some people question how you can have a 3-point estimate that has a lot of uncertainty surrounding the most likely outcome, though, so let’s deal with that.
If you have expert knowledge, historical data, benchmarks, research and the like, you’ll probably have pretty good confidence in what the most likely outcome is, and how likely that outcome is relative to any other outcome. But what if you don’t have expert knowledge, historical data, benchmarks or research? What if all you have is a vague guess?
Even with a vague guess, you can still guess what the most likely outcome will be, and what the minimum and maximum point-estimates are, too. You’ll probably be wrong; the actual outcome will probably not be what you expected. But that doesn’t mean you can’t estimate under a condition of great uncertainty about the most likely outcome!
The SPERT-7 Rule says that a Guesstimate occurs when you distribute 100 hypothetical trials over a 3-point estimate using a 33-34-33 split, where 34 trials are equal to the most likely outcome, and both the minimum and maximum point-estimate have 33 trials.
Suppose we have a 3-point estimate of 60-120-180. If we place those 100 trials in Excel, in cells A1 to A100, then use the STDEV.P function, we get a standard deviation equal to 48.74. To calculate the ratio scale multiplier, divide the standard deviation (48.74) by the range (180-60); the result is 40.6% If we model a skewed normal curve, where the 3-point estimate is 60-120-240, though, the standard deviation using STDEV.P is 74.48, and the Statistical PERT ratio scale is then 41.3%, which is a little closer to the SPERT-7 Rule’s 42% for a condition of great uncertainty over the most likely outcome.
Instead of using a non-evenly-divisible-by-7 value (and breaking the continuity of the rule), the SPERT-7 Rule uses 42%, which is close enough to be useful in creating Statistical PERT estimates.