Tampering

A man is drunk; knows not which is north, south, east, or west; wishes to go home. He takes a few steps, stumbles, rights himself, takes a few steps, knowing not which is north, south, east, or west; takes a few steps, stumbles, and continues this performance under his handicap. His chance to be within a kilometre of home diminishes with increasing attempts. This conclusion was foretold by Lord Rayleigh in 1898.

We can make two types of mistakes when trying to improve a system. It is a mistake to make local changes to a stable process. The term "common cause variation" is used by Dr. Deming to describe this type of system. The second type of mistake is when we treat abnormal events for normal ones. This type of variation is referred to as special cause variation by Dr. Deming. Making mistakes like these is tampering. A stable system is considered to be in statistical control. Meaning all the variation in the system is between statistically derived limits. A control chart can be used to determine whether a process is under statistical control. An example of a stable system can be seen in Figure 1. Neither the upper nor lower control limits are exceeded at any control point.

Figure 1 - Control Chart for Rule 1

A simulation of fifty marble drops in the Funnel Game.

Red Bead Game

A stable system does not necessarily mean the system is in a desired state. As an example, the foreman in the Red Bead game demands zero red beads. Since 20 percent of the beads in the bin are red, zero red beads seem unlikely. An example of a first mistake here would be to fire willing workers due to high red bead counts. A second mistake is giving a bonus to a worker who has a good day, for example, pulling fewer than five red beads that day.

Funnel Game

Dr. Deming uses the Funnel Game (1) to illustrate these mistakes. The game is simple. To do this experiment, you need a marble, a funnel, a stand to hold the funnel, and a target to mark where the marble hits (see Figure 2). You drop the marble through the funnel and mark its position on the paper.

View fullsize

Figure 2 - The Impact of Flipped Learning on Cooperative and Competitive Mindsets - Jung Eon Kwon and Hyung Rok Woo 

Rule 1

In this example, the marble is dropped, and no tampering occurs—only the points where the marble hits are recorded. An example of a simulated Rule 1 of 2000 drops is shown in Figure 3. A control chart of the same 2000 marble drops is shown above in Figure 1. The 50 drops show common cause variation, as you can see. No points are above or below the control limits. This non-tampered system is what Deming calls Rule 1. All systems are subject to inherent variation, as illustrated in Rule 1. The greatest dart player will not hit the bullseye 50 times in a row.

Figure 3 - Rule 1

Rule 2

Let's say the manager is unsatisfied with the variation of those 50 drops from Rule 1. They demand that all drops hit the exact mark, for instance. To compensate (i.e., tamper with the funnel), they mark each drop and measure the difference between the first and subsequent drops. This is what's called Rule 2, according to Dr. Deming. Tampering is evident in these adjustments. A radius of 2000 drops simulated using Rule 2 is shown in Figure 4. It's about 40% worse than the manager wanted because of this Rule 2 tampering. A real-world example of Rule 2 tampering is reacting to a customer complaint, or a survey, or adjusting a work schedule before checking if the system is stable. After a breach, the CEO of one bank demanded that they hire 50 cyber experts immediately. Before they even understood why the breach occurred.

Figure 4 - Rule 2

Rule 3

Things get significantly worse under Rule 3. In this case, the funnel is repositioned based on the distance between the dropped marble and the target. The positions of 50 marble drops for Rule 3 are shown in Figure 5. Goldratt's Theory of Constraints provides the best example of this type of tampering. The theory of Constraints postulates that when alleviating a bottleneck, one must first identify the constraint by looking for the process step that takes the longest time. Once identified, it's essential to address and elevate it by increasing capacity or reducing demand. Goldratt would describe Rule 3 as a local optimization. Local optimization will reduce global optimization.

Figure 5 - Rule 3

Rule 4

The Milky Way rule is sometimes referred to as Rule 4. The more you tamper with it, the farther you will get from the target. Figure 6 shows an example. You adjust the funnel to the last position every time you drop the marble. The "telephone game" is an excellent example of Rule 4. Each time someone passes on the information it gets mutated and gets farther from the original content. Examples of Rule 4 include organizations that have inadvertently institutionalized tribal knowledge. Most knowledge in these kinds of organizations is tacit, so explicit knowledge becomes a game of training the trainer. Brent's experience in the Phoenix Project and its impact on Parts Unlimited exemplifies Rule 4.

Figure 6 - Rule 4

Learnings

As shown in the first example, Rule 1 produced a stable system. Instead of applying Rules 2, 3, and 4, we should have addressed the opportunity as a system. Deming said the system causes 94 percent of troubles, and managers are responsible for changing it. In the Funnel Game, management could have supplied a funnel stand that was more sturdy or could have been adjusted lower when dropping the marbles to reduce the variation.

Learning from Lego

Lego's commitment to systematically reducing variation is legendary. Lego prioritizes quality by empowering workers and using complex processes. Legos fit together perfectly every time. Each brick is made to be compatible with bricks made as far back as 1958. Their system achieves precision tolerances of 0.002 millimeters, which uses very accurate molds. Lego could not have achieved its quality by tampering with its system.

Takeaways

The biggest takeaway from Dr. Deming's Funnel Game is that an organization should understand a system before changing it. Deming believed using analytical statistics was an effective tool for improving quality. The first step in understanding improvement is to use Statistical Process Control and control charts.

(1) Dr. Lloyd S. Nelson is attributed with creating the original Funnel Game.