BI's Promised Land, Part IIMeasuring processes is an important step, but improving processes is critical if you want to realize full performance management benefits
by Erik Thomsen In Part I, March 20, 2003, I described some of the statistical concepts used by advocates of Six Sigma that the business intelligence (BI) tribe — in its embrace of organizational performance management (OPM) — needs to adopt if it's going to make it to the promised land. These concepts include how to think in terms of processes and how to measure and interpret current events against a backdrop of historically determined relative likelihoods. By following these steps, you can measure the performance of your organization's processes while knowing the degree to which any variability in the measurements is the result of natural variation (also called noise) or extrasystemic forces that are signaling a shift in your process model. But measurement is only part of the Six Sigma story. Although necessary, measurement alone won't get you to the promised land; improvement is required for entry. And improvement depends on learning, which in turn depends on intentional interaction. In Part II, I'll describe why you need to measure operational aspects of the processes themselves — not just process outcomes — and why simply mining your data (regardless of how clean it is) doesn't guarantee that you'll learn anything useful. From Outcomes To DriversProcess-improvement decisions need to stem from some quantitative assessment of the process's current quality. Six Sigma offers two common measures of quality (and quality improvement). The most common is the count of the number of standard deviations (called sigmas) from the mean process result that fall within acceptable boundaries (as defined by the business). The greater the number of sigmas that fall within these boundaries, the lower the process variability. Consider, for example, the histograms of process outcome measurements shown in Figure 1 and Figure 2. Although both histograms have the same mean, the process variability is substantially larger in Figure 1. In that figure, process results are only acceptable 85 percent of the time. In Figure 2, less than 0.1 percent of process outcomes are unacceptable. The next most common is the measure of the mean process result. As Figure 3 and Figure 4 show, you can improve the average quality of a process (for example, improving the average lifespan of a battery or the average number of days within which receivables are collected) without changing the degree of process variability. Therefore, you can improve the quality of a process by reducing the variability in the mean process result or by improving the mean. Imagine you're in charge of customer relations for an exclusive retail chain. Let's say the current variability in customer satisfaction resembles Figure 1, and your goal is to improve the consistency of the customer experience to the point at which it resembles Figure 2. Although you need only measure customer satisfaction to determine its quality, to improve that quality you need to learn the drivers or causes of customer satisfaction. And you need to know those drivers are under your control.
|
 |
Most Popular This Week
IE Weekly Newsletter
Subscribe to the newsletter
| |
| ||||||||||||||||||||||||||||||||
