In-Process Sampling Plan for a Secondary Package Adjustment Process

[BERMUDEC]

Dear All,

I would be grateful if you could help me with a couple of questions:

1) There is a sampling plan for a secondary package adjustment process:

a. Process description:
Remove the original leaflet from the box (package), place alternative leaflet, place warning labels on the box (the quantity of these labels depends on the product (2/3/4)). Each one of these steps will be dubbed as "process stage".

There a 13 different products, different sizes, different number of stages.

b. In-process sampling plan description:
Every 2 hours a QC samples the packages according to this pattern: in a process composed of up to 2 stages - 3 packages must be sampled, in a process composed of up to 4 stages - 5 packages must be sampled, in a process composed of more than 4 stages - 6 packages must be sampled.

I fail to see the logic behind this plan - there's no mention whatsoever of the total quantity of the product in the batch. During 2 hours, different quantities of different products could be adjusted (even if some of the products require the same ammount of stages).
Another thing I don't quite get where do the relationship between the ammount of stages to the sampling sizes come from - I couldn't find a standard or any historical data that justifies these values.

2) Do you know any standards/techniques deals with similar in-process sampling?

Thank you in advance.

 

[Bev D]

First: the appropriate inspection sample sizes are not based on the lot size. (even tho the popular ANSI tables use them, they are not statistically correct)

The plan you are describing sounds it was developed to increase the units inspected as the opportunities for a defect (more process steps) increases. While teh logic for this was no doubt well intentioned it is in contradiction to the reality that the ability to detect a defect increases as the defect rate increases.

statistically based sample plans would be based ont eh defect rate that one wants to detect and reject and the probability of detectign that rate (or worse) if present. Plans can be augmented by aslo specifying the defect rate that is acceptable to ship and the probability of accepting a lot at that rate or better...

On teh hand the situation you describe sounds like a 'dock audit' and these are typically not statistically based. Just take a random small sample; you are looking for catastrophic large defect rates at this point.