All Estimates Are Wrong, All Estimates Are Useful #NoEstimates

There's a thread on #NoEstimates that goes like this "can you point me to literature that shows all the problems with estimates." One response was "industry is full of examples," bit no actual reference of course. There are many places to look. Start with "root cause analysis" and append RAND or IDA or MITRE. You see papers and reports on the many ways estimates and estimating is troubled.

But the famous phrase "All models are wrong, some models are useful," is serious misquoted. It comes from George Box's paper "Science and Statistics," But that paper comes from Box's two seminal works "Time Series Analysis Forecasting and Control, George E. P. Box and Gwilym M. Jenkins and Empirical Model-Building and Response Curves, George E. P. Box and Norman R. Draper. The time series book is the source of the Box Jenkins algorithm that is the basis of ARIMA (Autoregression Integrated Moving Average) used to forecast performance on past performance.

The real point here is anyone used the "all models are wrong..." phrase needs to have all three of these documents on their desk. With that they'll see the context of the phrase. See §2.3 of the "Science and Statistics" paper.

This brings me back to the core issue. How can decision be made in the presence of uncertainty without estimating the impacts of that decision? The "exploring" process provides ZERO answers. Those originally suggesting the #NoEstimates idea seem to have intentionally ignored making any suggestions on how to actually make those decisions. This appears to me to be a complete coop out - here's a provocative idea, but I haven't really thought about how to fix the problem - "I'm just making an unsubstantiated claim" - talk among yourselves. It's a Saturday Night Live skit.

If you listen to Ryan and Zach's Agile Podcast, there is more details about estimating in the presence of uncertainty, approaches, problems, and so possible solutions to the misuse and abuse of estimates in low maturity management paradigms.

Reply · Report Post