This thread recently appeared in myASQ:
Question by Aju: Hello Team, We have a situation where we did double sampling on Incoming parts from a Supplier(Machine shop). The lot was accepted and released to Inventory. These parts then later got assigned to WO and during assembly, the Manufacturing team found one non-conforming part (missing slots). manufacturing initiated NCR and Investigation found out the root cause to be supplier issue. and the probability of finding the defective during sampling was less than 20%. A week later 2 more NCR opened for a similar issue (same supplier). The supplier was notified and requested Root cause and corrective action from their end. What’s the best course of action here to avoid non-conforming parts ending up in manufacturing. Tightened sampling/100% Inspection?
Reply Jake: The general practice for us is to revert to 100% inspection for the next three deliveries. If there are no issues, then we revert to sampling plan, if there are issues, we maintain the 100% until we get 3 lots with no issues. Once we get three lots with no issues, we double the sample size for the next three lots, increase to Level III, then normal sample after that.
Here is my reply to Aju and Jake. Acceptance sampling was once widely used in quality control. In fact, it defined quality control. The “other quality control” was process control. Mostly statistical process control (SPC) but also other tools that helped identify and correct root causes of problems. Acceptance sampling focuses on output, looking at product after it’s been produced. I.e., on after effects. Process control monitors the inputs to the production process and through the use of statistical methods it determines the best course of action to prevent future non-conformances. If there are statistical signals then an immediate investigation into the cause of a change ensues and corrective action is taken on the spot. If no signals exist but non-conformances still occur, action is focused on redesigning the process itself.
In fact, it can be proven mathematically that acceptance sampling is futile if a process is in statistical control!
What are your thoughts? I’m truly interested. Please comment below.