Saturday, July 10, 2010

Software Measurement - We Are Like Eskimos

I forget at times that Software Engineering is a young discipline. At times I am frustrated that it is not easier to measure good software and good software developers. The following gives some insightful perspective:

Measurement can progress from lower to higher scales as societies, organizations, and practices mature. An illuminating example:

We can imagine, for example, that certain Eskimos might speak of
temperature only as freezing or not freezing and, thereby, place it on a
nominal scale. Others might try to express degrees of warmer and colder,
perhaps in terms of some series of natural events, and thereby achieve an
ordinal scale. As we all know, temperature became an interval scale with the
development of thermometry, and, after thermodynamics had used the
expansion ratio of gases to extrapolate to zero, it became a ratio scale."
— S. Stevens, 1956

There is an important lesson here for software engineers and managers—do not expect that everything you will want to measure can be expressed with ratio scales today. Software engineering is a young discipline. Just as with Stevens's Eskimos, it may take us time to evolve to where our measurement practices become comparable with those of other disciplines. Be willing to start with nominal and ordinal scales, just to get measurement started. But be mindful of the limitations of the computations and interpretations that you can make with the kinds of data you collect, and look for opportunities to evolve your measurement practices toward scales that provide greater information.


The above is taken from

Goal-Driven Software Measurement —A Guidebook by Robert E. Park

No comments: