Don't Be Misled by Specification Limits


Matt J. Savage, VP, Product Development, PQ Systems, Centerville, OH, USA


Cpk, Charts, Specifications





A specification limit is the universally accepted ‘litmus test’ that we all agree to abide by. Specification limits seem quite simple to use and apply. After all, if the product or service being delivered is measured and found to be in-between the Upper and Lower specification, it is accepted as good. It is so clear, so black and white, how could anything be gray about applying this concept? In fact, if this was the only rule, it would be hard to argue that something is “In” or “Out” of spec. In tennis, the Hawk Eye line-calling system is used to determine if the ball was inside the line, hit the line (but still considered in), or outside the line. In the quality improvement world we live in, anything measured outside the specification limits is unacceptable. Given this operational definition of acceptance, there should be no ambiguity. Unfortunately, there are problems in this naive world of acceptability. Customers do not simply want a product produced or service delivered anywhere between the specification spread. They want two things: 1) the measured value to be at the desired target and 2) minimal variation.

Practical Takeaways

  • Understand and explain why specification limits should never be placed on control charts for subgrouped data.
  • Understand the pitfalls of adding specification limits to X-mR control charts.
  • Understand the relationship of capability analysis to control charts.
  • How a statistical dashboard can be used to quickly and efficiently increase the bandwidth to evaluate many processes for process stability and capability analysis.