This time we look at using data large and small to counter the very human fallacies that prevent us from solving problems.
An example: While talking to the manager of the facility that was having a production problem, they stated “Every time I go to the facility, the ___ machine is down and the line’s down too.” After which a very long discussion followed about how to fix that machine or circumvent its use. At no time was it quantitatively determined that that device was what was actually causing the line to be less productive, or if the line stoppage was directly caused by the machine, something upstream of it, operator failure, or any number of other causes.
What’s going on here? It’s an “anecdotal fallacy”—what’s observed at a particular moment is taken to indicate the permanent condition of something—and it’s often misleading. Did we know that this was the “real” problem, one of many, or just observational bias? Had we used any sort of analytical methodology, we might have found several or several dozen causes of the problem, then winnowed them out properly. If we can use complex data systems to correlate what our market systems will do globally, we can use simpler tools—many available free for use on our laptops, tablets, and phones—to support decision making at a more practical level. Although correlation isn’t causation, it is a primary step to tracking down causation.
Production machinery of all types continuously produces useful data, but it is often ignored except to prove that quotas have been met or that a line is functioning. It is often thought to be too complex or too time-consuming to break down causal factors into likely failure scenarios, but to ignore analysis for an “obvious” answer is to throw away the quality revolution that has crept into every corner of manufacturing. We cannot admire international industries that have gone from literal rubble to icons of quality and productivity in a few decades and ignore the tools that can give us the same benefit.
Where can you start? Most academics will point at long tracts by Deming or Shewhart but my favorites are “The Thinker’s Toolkit” by Morgan Jones, particularly for causal analysis, and a pocket book “The Memory Jogger” produced by Goal/QPCi. Neither is expensive, and used editions can be bought online for a few dollars each. You may find your own favorite, but whatever is chosen, the first response to “the guys say it’s always…” should not be “Let’s fix that?” but rather “What does the data show?”
Scott A. Morris (firstname.lastname@example.org) is the director of the packaging program at U. of I. Urbana-Champaign.