Once in a while you come across a book that contains a single sentence that is worth its list price. This is such a book, and its definition of measurement is such a sentence.
I have the honor and privilege of leading seminars and classes for senior software design engineers in test at Microsoft. When I ask for a definition of measurement I have yet to hear anybody offer the one that's at the center of How to Measure Anything, and I have yet to see a single person not smile when I share the definition Douglas offers in his book.
Invariably, the smiles grow to grins when I suggest that by seeing tests as sets of observations that reduce uncertainty, and by incorporating the other mathematical tool Douglas explains in the book, Bayesian inference, testers are now capable of seeing the effect a single test result has on the joint probability distribution of uncertainty for the entire test run.
From this new foundation we are able to incorporate the work of other researchers to produce a measure of the relative value of individual tests selecting those that most reduce the overall uncertainty that the product is fit for service. This ability to measure relative value, then, enables testers to add ever more effective tests to the suite so we can measure the incremental change in overall uncertainty. What are the odds that the end result of this simple definition of measurement armed with Bayesian inference would be so profound? For in a word, we can now model and measure, not merely the software itself, but the judgment of the tester, as well. Management loves it, the development team loves it, and the tester loves it.
If only Douglas had written this book ten years ago...
Get more detail about How to Measure Anything: Finding the Value of "Intangibles" in Business.
No comments:
Post a Comment