Performance Measures/Metrics: Attribute Versus Variable Data

Basically, there are two types of data to collect as a part of a problem-solving process:

• Attribute data, or go/no-go information
• Variable data, or measurement information

Because the level of sensitivity of a measurement depends on the precision of the measuring device, there are times when variable data can be treated as attribute data.

For example, suppose you are producing aluminum pins that may be smaller than 1.065 inches in diameter but not larger. Rather than measuring each pin or even a sample of pins, you can:

• Use a plate that has a hole 1.065 inches in diameter bored through it (a go/no-go gauge)
• Insert each pin to be inspected in the hole
• Classify any pin that passes through the hole as accept, treating others as rejects

Thus, treating the variable as an attribute offers an efficient way to determine if the pin will be effective.

Performance measures, also known as process metrics or key quality indicators, should be ratios. These ratios are the statistics that describe how well or how poorly a process is performing.

Sometimes the ratios have labels such as defects per unit (dpu), defects per defective unit, defects per X units, or defects per million opportunities (dpmo) or parts per million (ppm) defect rate.

However, there are ratios that do not have labels, for example, CP, the process capability index, and CPK, the mean-sensitive process capability index.

Excerpted from Jack B. ReVelle’s Essentials: A Reference Guide from A to Z, ASQ Quality Press, 2004, pages 126-127.