If an inspector controls the quality of your products in China, he probably checks only a portion of the whole batch.
But how does he decide how many pieces to pick for his inspection? In other words, what sampling plan does he follow?
The most commonly used is the single-stage sampling plan by attributes.
What does it mean, and what are the other options?
A “single-stage sampling plan” dictates that a certain number of pieces (n) should be drawn and inspected. That number n depends mainly on the size of the batch (and also on the inspection level). If the number of defects is under the AQL limit, the result is passed.
In a “double-stage sampling plan“, the inspector would start by taking a smaller number of sample (n1). If the number of defects in n1 is above certain limits, more samples are picked.
An “accept on zero” plan is a weird animal that is usually not a good idea in China. Some importers, who are sensitive to legal litigation by their customers, accept batches only if no defective unit is found. The only advantage is the fewer samples need to be checked. More in-depth information here.
A “sampling plan by attributes” classifies the samples as either “non defective” or “defective”. There is nothing in between.
A “sampling plan by variables” allows for a finer evaluation. For example, the length of the product is measured, and the exact findings are taken into account when a decision is made.
A “rectifying sampling plan” is applicable if the defects that are found can be corrected immediately. It is not very different from a sampling plan by attributes, but it takes into account that the batch is of higher quality after the inspection… and, in case of inspection failure, the whole batch should be inspected.
“Continuous sampling” is the best plan when products are made individually in a continuous flow. It makes no sense to pick samples inside each “batch”. It consists of several phases:
- At the beginning, each piece is checked (that’s the “screening”).
- After a certain number of pieces were found satisfactory, only certain pieces are checked randomly (that’s the “sampling”).
- If a piece is defective: back to screening.
—–
If you want to use one of these plans, you should visit the excellent SQC Online website and get the numbers you will need.
And if you want more in-depth information about these plans, you should read Practical Acceptance Sampling: A Hands-On Guide. It was pointed out to me a few months back by Etienne Charlier.
I pasted this book’s description (from the publisher) below. I should add that it is a college textbook first, before being a practitioner’s guide.
Acceptance sampling plans provide criteria and decision rules for determining whether to accept or reject a batch based on a sample. They are therefore widely used by manufacturers, suppliers, contractors and subcontractors, and service providers in a wide range of industries.
The book introduces readers to the most popular sampling plans, including Military Standards and civilian ISO and ANSI/ASQC/BS standards. It covers the design, choice and performance evaluation of different types of plans, including single- and double-stage plans, rectifying and non-rectifying plans, plans for pass/fail and continuous measurements, continuous sampling plans, and more.
What do you think?
In the North American auto industry it is common to use an acceptance number of zero nonconformities, on the basis that:
1) Process capability studies have been done in advance and prove that when the process is kept in control, the probability of nonconforming product is extremely small (e.g. parts per million.)
2) If a single nonconformity is found, the batch is segregated for a 100% rectifying
inspection.
These practices are generally incorporated into the production process, using a process similar to the “continuous sampling” plan discussed above. There will be a first piece approval for each process, where, say, the first 5 pieces of the shift will be checked. If any are out of specification, the process will be corrected and a new
first piece check done. If nonconformities are found after an hour of production, a
100% check back to the last :”in control” point will be done.
This protocol has been the expectation for many years (at least 30 in my recollection).
The justification here is that the finished automobile is the result of an assembly of thousands of individual parts. A single defective part can have serious consequences – for example, a single bad piston can result in poor performance of the engine. One client of mine makes bolts used for assembling engines; a few bad bolts (worth $.50 usd each) can result in the recall of the vehicle (cost of $2,000 – $5,000 per vehicle). So there is zero tolerance for defectives.
While no one will admit it for the record, the tolerance is greater for appearance issues – say, for example, paint quality. There is certainly an effort to control
appearance, but the subjective nature of appearance checks demands a little more latitude.
Brad
Very interesting, thanks a lot Brad!
As you point out, working on processes (in addition to doing manual inspections) is necessary to have the best results.
Renaud, do you have any other suggestions of books or online courses for people that want to learn more about QC inspection procedures and sampling?
I have learned from reading directly the ISO2859 standard, the book I mentioned above, and by talking to people in the industry. But I am sure there are other good sources of information.
can you tell me how many box to be opened for sampling of packaging material according to AQL (MIT STD 105,ISO 2859).
The standards you cite above say nothing about the number of cartons to pick. They only specify how many units of the product to pick.
In general, inspection agencies pick at least the square root of the total number of cartons (example: pick at least 10 boxes if there are 100 boxes).