Monday, March 2, 2015

No Dice!

It seems that every PM starts with dice or coins to understand both probability and statistics -- mean, variance, sigma, etc

Fair enough; it's not a bad starting point. Everyone has thrown dice or tossed a coin at one time or another.

Here's the things to know if you try to port your experience with games of chance, like dice and coin tosses, into the project domain to address risk and uncertainty.

First, with dice or coins:
  • The long term statistics and probability distributions are known, and not amenable to management intervention to change or challenge the outcomes.
  • The only uncertainty is the outcome of the next throw; all other statistics are known.
  • A long string of outcomes that don't appear random, like HTTTTT for a coin toss of heads or tails, does not suggest a bias
Second, none of the above applies to projects. In the project domain:
  • The long term statistics and probability distributions of risk and uncertainty are NOT usually known, but if they were they ARE amenable to management intervention to change or challenge the outcomes
  • The are many uncertainties that could affect the outcome of the risk event; almost all statistics are unknown.
  • A long string of outcomes that don't appear random does suggest a bias
And so, is it pointless to study probability and statistics, or to get familiar with the concepts?  NO, NO! (And strong message follows)

Here's why you should know a few things:
  • The Central Limit Theorem that predicts a central value among random outcomes can work for you to simplify estimating, and to validate estimates and observations
  • Bayes Theorem is a powerful idea about using observations to improve predictions
  • The Law of Large Numbers could save you a ton of money and time by showing you that it's not necessary to measure everything
  • Expected value is a tool for avoiding many of the flaws of averages while also reducing a lot of data to something meaningful to carry forward to sponsors and non-quantitative managers
  • Confidence limits give you some wiggle room to establish credibility but avoid the pitfall of a single point estimate that is almost always wrong.
I could go on, but you get the point. Study statistics! Keep learning stuff
And, if you want to follow-up, start with the videos at the Khan Academy, probably the best material around for the beginner.

Read in the library at Square Peg Consulting about these books I've written
Buy them at any online book retailer!
Read my contribution to the Flashblog