Measurement Failures

Another reason measurement programs fail is often they point out painfully obvious things. When I started WeightWatchers, I had to admit to myself I was actually obese, and this was hard for me to admit; it is still hard for me to admit. I searched around height and weight charts looking for some data to say, “David you are fit.” I had to come to the realization I was obese and I needed to lose weight. I am not unique in this belief either. Of adults who are considered obese by the National Institute of Health eight five percent do not think they are obese. Kimberly P. Truesdale of the University of North Carolina says, “If obese people don’t identify with being obese, they are going to ignore messages warning of the health risks.” Like obese individuals most software organizations have mistaken belief regarding their levels of performance. They think they are doing fine when in actuality they are not. I write more about this problem in the chapter Unskilled and Unaware.

I am not unique in this category either; eight-five percent of those individuals who are obese do not consider themselves even overweight. The first step to improving the software development process is an admission that there is a problem. The major input to developing software is from humans, and it is only natural that organizational and individual behaviors would be a major stumbling block for any initiative.

Another reason for measurement program failure is many think the solution is to just measure. Measurements are gathered up only to realize nothing changes. I was working with an executive who pointed to a chart. It looked like a horizon line or flat line, and he said, “This tells me nothing.” In return I asked, “Has the organization done anything to change?” His answers was, “No.” This would be similar to getting on a scale and diligently weighing yourself expecting that by just weighing yourself, you will see positive results.

Another problem is interpretation of results. Often the measurement results are all over the place. When the results are plotted they appear to be random such as a scatter diagram; tey have no pattern whatsoever. When the software organization develops software a variety of different ways, a large variation in results should be expected.