I'm very much enjoying reading How to Measure Anything by Douglas Hubbard. The book
- is practical in its approach to measurement, showing that one has to define a thing in order to measure it and that many of the things that people think are hard to measure are actually not well defined.
- defines measurement as a reduction in uncertainty - thus measurement happens when we increase our knowledge about the subject (by either increasing our confidence in the bounds or reducing the bounds at a certain confidence level).
- describes how to calibrate individuals so that they become accurate measurement tools themselves by learning the skill to estimate with specific confidence levels and bounds.
- provides recipes for simple statistical methods for interpreting sample data.
Measurement is a key aspect of development, especially when one is developing solutions for the acquisition and publication of data.
- How accurate is the data that one intends to harvest?
- What impact will this new data have in the application area?
- How much effort should we put in to measuring the quality of the data?
One of the drawbacks of the book is that while it gives some very interesting mechanisms for working with samples, it doesn't actually provide the statistical underpinnings explaining them, resorting instead to instructions involving looking up tables and reading distribution graphs.
Overall, the message is: you can measure (reduce the uncertainty about) any quantity you can define with a lot less effort than you think.