As several Guesstimate users noted, it’s very difficult to talk for negative minutes. A Guesstimate example video featured the multiplication of two normal distributions (talk time and viewer count) to find the total viewer time, which was shown to possibly be negative.
Talks of negative time are what happen when you use the wrong distribution type. Normal distributions are intuitive to understand, but are not limited to positive values. They are also symmetric, so it’s equally likely a value could be far under the mean as that it will be far over it.
Real life normally isn’t like that. There’s typically one big tail. Blog posts can’t go anti-viral and many businesses can only go bankrupt.
This could be one reason to take big risks! And it could be a reason to use lognormal distributions in Guesstimate, which are newly available and work like a charm.
Here’s how they work. You enter a 90% confidence interval, say you estimate that a blog post will take 2–8 hours to write. Select the lognormal icon. Then, there will be a very small chance it could take over 14 hours, but zero chance it will take -4 hours.
Even More Distributions
Guesstimate now also supports several other distribution types, including beta distributions, cauchy distributions, exponential distributions, gamma distributions, and several other kinds of distributions. These go in functions, which means that their inputs could be other kinds of distributions.
Want to model an exponential function but not sure of the exact shape? Don’t choose an exact parameter, choose another distribution with a lower and upper bounds. Put distributions in your distributions.
All this gives you a lot of flexibility and possibly reason to learn about each kind of distribution. It also allows for a change of meaning in guesstimates. While the original normal distribution stood for uncertainty ranges on the behalf of a modeler, some of these instead could indicate a real distribution of a given population.