Project managers need to be able to assess product quickly with an intelligible set of quality metrics. In general, we recommend paired metrics; for example, in software development we can look at lines of code versus errors per line to keep both metrics “honest.” In the automotive world we like to see Cpk, which is a measure of both quality and centrality; however, both of us are happier when we see the Cp value as well because this paired metric will give us an idea of our theoretical capability versus what we are actually seeing and maybe even provide a hint as to what is wrong.

The same kind of single-value mentality prevails when people try to sell us company stock. The usual statement is something like “the stock market has average roughly 10% per year over the last 70 years.” Unfortunately, those of us who own stock generally don’t see the average, we see the variance (the square root of this is the standard deviation) as stock value random walk their way around the various indices. Mean and standard deviation are pretty basic measurement—we would also add the median, since if the mean and the median are significantly different, we either have substantial outlier values or we are not dealing with anything faintly resembling a normal distribution.

Sample sizes are important, too. The number “30” seems to have acquired religious significance over the decades, but the serious project manager will want to see how the sample sizes were derived to avoid distortions in information that lead to poor judgment.

We also recommend that the project manager ask for control charts whenever feasible since these are far more revealing than red-yellow-green metaphors on a dashboard program. Good quality software will mark the areas of the chart that violate control chart thresholds.