I'm 30% sure that 80% of the work will be completed within 5% variance of the estimated completion date. What? Just finished reading an article about using probability calculations (Monte Carlo) to determine when tasks will be completed. This provides the percentage of certainty of when tasks will be complete. The example provided consisted of 2 tasks each with an estimate of 8 hours and it was determined that there is a 50% chance of them being completed within the 8 hour estimate (each) or potentially 80% chance of being complete with in 16 hours (each) - aka

'it took you twice as long to complete??? why??'. Where I think this type of discussion interesting - I find the approach useless. Realizing that a small project could consist of 30-50 base tasks, many relying on prior tasks completions/output to begin, etc. - the ability to use probability to determine the resulting bell curve to determine percentages of completion based on time, etc.....is slim based on typical system noise (chaos theory 101). Maybe if I went to MIT for Probability type classes (math) - I might have more faith in utilizing it for estimating and realizing the value (

http://ocw.mit.edu/OcwWeb/Mathematics/index.htm?gclid=COzStYCO4IoCFQk_gQodMxTAxw).

In addition to dependency on related tasks one also needs to consider:

- quality of delivered product (poor delivery results in additional future work, etc.)
- quality of resources
- definition of complete
- external factors (chaos)
- overwork (over delivery - fish getting as big as the tank)
- communication ('oh, by the way task x was complete a week ago - sorry about that')
- vacation, sick time, to much beer, general distractions (planned and otherwise)

How probable is the probability usage accurate

enough for use? If at best you can get within +/- 100% (example from article) - I think dart throwing (currently the most used method in estimation) is just as good. In other words, what is the cost of using probability to provide estimates where the current process is relatively the same. If you want better estimates and better results take the $ and spend it on better resources.

## No comments:

## Post a Comment