From: Stan Hilliard
Date: 14 Aug 2003
Time: 10:09:12
YOU SAID: "...today's qualty levels achived (typically measured in hundreds of PPM or less and Cpk>1.33)"
MY REPLY: Those extremely good quality levels are seldom demonstrated with attribute data to have been achieved for individual lots. Having a C=0 plan does not support that claim of having attained those levels.
To verify extremely low levels requires more inspection than is usually possible. A sampling plan with a consumers risk of 5% and RQL=1 part per million requires inspecting 2,968,434 items, accept on zero. If you reduce the consumers risk to one chance in a million, the sample size becomes 15,130,863. Thus, confirming such levels with data for each lot is impractical.
YOU SAID: "For any given sample size and RQL, C=0 gives the smaller consumer risk."
MY REPLY: I agree.
YOU SAID: "For any given RQL and consumer risk, C=0 gives the smaller sample size."
MY REPLY: I agree.
YOU SAID: "Further more, C=0 plans can be allways treated as "sequential". Stop sampling when you find the first defect (and reject the lot) or when you reach n (and accept the lot)."
MY REPLY: I agree.
YOU SAID: "Now, If I want zero defects (or about that) from my customer, which sampling plan would I use? None. I would expect him to have capable and reliable processes." and "...our quality is based on process control and not on these inspections"
MY REPLY: Since the employees influence the process and the sampling plan influences the employees, perhaps the plan should be thought of as a feedback component of the manufacturing system (equipment, employees, procedures) -- as opposed to a way to confirm the output. It is a production tool to to help to make product right rather than a dispositioning tool to verify that lots are conforming. You don't necessarily have to have low RQLs for process feedback.