If Cp>1, the six sigma process spread fits inside the tolerance or specification limits. Then how the defects can exist? For example, if Cp=1.67 (i.e. 5 sigma level), the DPMO is 233. Does six sigma consider defect as per control limits (whereas process capability considers specification limits), rather than specification limits?
Parag Sen - This is a very good question. I would first humbly suggest a preference to use Cpk rather than Cp as a measure of process capability as it reports the minimum of either Cpu or Cpl and hence accounts for a process that may not be centered within its specification limits. With a stable process as you propose the likelihood of defects are quite rare - but not impossible. Defects should indeed be defined by the specification limits - NOT the control limits. With a stable process at a high level of capability rare excursions of the control limits are possible (e.g. 0.3% of the time for a 3 sigma process, Cpk=1.0), very rare for a 5 sigma process, and an even more remote possibility at the specification limits. If you are producing widgets this may be quite good, for launching rockets perhaps an even higher level of confidence is necessary.
Best wishes and good luck!
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics