I've made the mistake of engaging the #NoEstimates crowd in questions about the usefulness of their idea. Not a group that likes to get questions, by the way. They use the old Ron Jeffries approach (which has since passed for him) of well just try it to see if it works for you.
The big question is:
If you been asked to make a decision about something related to software development, say a make buy decision, is making an estimate of the comparative costs or comparative durations a good idea?
Turns out that is not a question that can be asked in that crowd. So I learned my lesson again. In the mean time, this diversion caused me to dig out my slightly tattered edition of Time Series Analysis: Forecasting and Control, Box and Jenkins, 1976. I'm working on a presentation for the Fall Integrated Program Management Conference for the use of time series analysis to forecast the Estimate At Completion using the CPI/SPI numbers at the Control Account Level available in the XML submissions of the Contract Performance Report.
The approach currently used to forecast the EAC wouldn't get a D in any statistical analysis class.
- Take all the past performance an accumulate it to a single number
- Take the current period as a single number
- Forecast the future performance of this number is a linear ratio
- Do not adjust this number for the underlying variances of the past
- Do not adjust this number of the assured underlying variances of the future
- Commit millions and many times billions of dollars on the forecast
Along with the Box book I use, there are of several books for applying R to time series analysis. Here's the list of resources for those interested in this topic.
- Time Series Analysis: Forecasting and Control; George E. P. Box and Gwilym M. Jenkins, Holden-Day, 1978
- Empirical Model-Building and Response Surfaces, George E. P. Box and Norman P. Draper, John Wiley, 1987.
- Bayesian Inference in Statistical Analysis, George E. P. Box and George C. Tiao, Wiley Interscience, 1992.
- Introduction to Multiple Time Series Analysis, Helmut Lutkepohl, Springer, 2005.
- Time Series Analysis and Its Applications with Examples in R, Robert H. Shumway and David S. Stoffer, Springer, 2011.
So when you're asked how much will this cost or when will you be done spending my money for the things I asked you to develop or what's the confidence you'll be ready for the "go live" date or better yet the "launch" date of this project, hopefully youranswer woun't be well you see we don't make estimates on our project work, we just develop small increments of fucntionality till the time and money runs out.