(Click on a rule for more information)

For rule 2 and 3, visual 2 is taken from James R. Evans and William M. Lindsay. 2005 (6th ed.). The Management and Control of Quality. Thomson, p. 524.

It might be esaier to understand the logic of the last three rules if we keep in mind the follwoing reference points for each adjustment (remember, the following are private cases of management by results):

Rule 2 - Adjustment relative to the last (n-1) event/result
The best way to understand this case of over-adjustment is to look at an adjustment following (reaction to) a specific event, such as: reaction to rumor; drivers chasing each other (in fact, any competition); raction to a single customer's complaint.

Rule 3 - Adjustment relative to a pre-determined value/standard
Weight watching; The Red Bead Experiment; calibrating a gun/rifle.

Rule 4 - The last result becomes the new target/standard
Playing telephone; mimicking; the arrival time of the last student to today's class becomes the start time of the next class; training on the job; following the latest fad/fashion.

Here are the results of 15 drops with each set being governed by a different rule (Deming, The New Economics: 205.)

Take a look at this simple example of our class's start time

Take a look at the effect of each tampering method on a system's variance.

A good simulator is available at http://www.symphonytech.com/dfunnel.htm

Take a look at what this simulation produces.

Class Assignments:

1. Plot the 4 result columns. The target is 0.
2. Find examples (HR, education, business, family, etc.) for each of the four rules.
3. What are the (dis)advantages of a (un)stable system?
4. Is a stable system = quality output?
5. Identify instances where you are likely to tamper with a system, that is to react immediately to a single, chance event.
   

Remember:

Tampering is action on the system without action on the fundamental cause of the trouble (The New Economics: 67).

A process may be stable (i.e., produce results within predictable limits), yet turn out faulty items and mistakes. To take action on the output of a stable process in response to production of a faulty item or a mistake is to tamper with the process. Put differently, tampering means making harmful changes in reaction to chance events (i.e., common causes of variation). The result of tampering is only to increase in the future the production of faulty items and mistakes, and to increase costs -- exactly the opposite of what we wish to accomplish (ibid: 202). Thus, if anyone adjusts a stable process to try to compensate for a result that is undesirable, or a result that is extra good, the output that follows will be worse than if he had left the process alone (Out of The Crisis: 327). The reason being, tampering invariably increases variation in the results of a stable process. 

Note, specification limits are not action limits. Severe losses occur when a process is continually adjusted one way and then the other to meet specifications. Control limits must be calculated from pertinent data (The New Economics: 178-9).

Improvement of a stable system requires fundamental change in the process (ibid: 202).

More on variation

Variation is a fact of life. It is random and miscellaneous. Thus, the same process can produce two things that are not alike. In the days of hand-crafted products, this could be accounted for by "fitting" things together. In modern industry where interchangeable parts are assembled into mass-produced final products, controlling variation is critical to customer satisfaction. This is one of the most important tasks a manager faces.

Dr. Walter Shewhart identified two kinds of variation, controlled and uncontrolled.

Controlled Variation/Stable System:

1. stable
2. exhibits a consistent pattern over time
3. results of a stable process can be predicted with greater certainty
4. a stable process can be improved because outcomes of changes can be predicted

Uncontrolled Variation/Unstable System:

1. changes over time due to special causes
2. cannot predict results of process
3. process cannot be improved easily since outcomes of changes are unpredictable

Management's job is to manage (reduce and stabilize) variation in order to produce predictable results, such as quality, cost, and production schedules. Since all data contain random variation or noise, the noise must be filtered out, otherwise two kinds of mistakes could arise:

1. Interpreting stable variation (noise) as if it were a meaningful change (signal) and expending efforts to "fix" stable/natural variation. In other words, here we ascribe a variation or a mistake to a special cause when in fact the cause belongs to the system (common causes). Overadjustment is a common example of this mistake.
   
2. Interpreting uncontrolled variation (signal) as if it were noise and not recognizing when a change has taken place. Here we ascribe a variation or a mistake to the system (common causes) when in fact the cause is special. Never doing anything to try to find a special cause is a common example of this mistake.

One Last Time: Process Improvement