In most production environments, machine operators are responsible for taking samples to monitor the process and empowered to make process adjustments to keep the line running and product flowing. In many cases, they can be reprimanded if they neglect to react to variation in the process. What if the better strategy is to ignore this data and keep chugging along, with no adjustments to the process?

First, let’s discuss our expectations when we sample from a process. Deming outlined his red bead experiment over 40 years ago in his seminal text Out of the Crisis (Massachusetts Institute of Technology, 1982). Imagine a large bucket of beads, mostly white beads with a scattering of red beads well mixed in the bucket. Deming often did this experiment in his seminars, and he would explain that The White Bead Company sells white beads to their customers. He would ask audience members to collect a sample of beads by dipping a paddle with 50 pre-drilled holes, each the right size to capture a bead, into the bucket. An example of possible results is shown in Figure 1, which plots the percentage of red beads collected in each sample (by a given participant). Note first that the number of red beads varied with each participant’s sample. Deming chastised those with a relatively large number of red beads in their sample (“Customers don’t pay for red beads; you’ll be docked pay!”) and congratulated those with a smaller number (“This employee is management material!”). It was all good fun, yet ludicrous, as each participant had zero effect on the number of red beads in their sample. And that was one of Deming’s points: It is ridiculous to grade employees for actions that they have no control over, such as the process design, the equipment, the raw materials, and so on.