How many samples to inspect?

It depends…
by RENAUD ANJORAN on 1 AUGUST 2011

Let’s say you want to follow the general procedure for quality inspections. How many samples should you draw for inspection?
Let’s also say you want to follow the ISO 2859-1 standard, under normal severity, and in level II. (If you are not familiar with these
concepts, you can read Quality Control Basic Concepts).
It is fairly easy to read the AQL tables and to see what the number of samples to inspect is…

… Except in certain situations, in particular when your AQL limits are low and your sample size is low.You might have to select different
numbers of samples for each category of defect.
Let’s walk through two examples to illustrate this.

Case 1:
Order quantity is 20 pieces. Level II, normal severity. AQL is 1.5 for major, and 2.5 for minor.

The letter is C:

The inspector draws 8 samples at random, and the batch is rejected if he finds at least 1 defect in any of the 2 categories. It means there
are 3 samples that are checked for major defects but that are not checked for minor defects.

Case 2:
Order quantity is 30 pieces. Level II, normal severity. AQL is 2.5 for major, and 4.0 for minor.

The letter is D:

The inspector draws 13 samples at random. For major defects, only 5 samples are checked, and the batch is rejected if 1 or more defects
are found. For minor defects, 13 samples are checked, and the batch is rejected if 2 or more defects are found.
This is one of the details that many people in the industry tend to overlook. Yet it is indispensable if one is to apply these statistical rules
correctly.

On a similar subject, you can read Statistical inspection levels: beware of the confusion.

The four types of quality inspection services

by RENAUD ANJORAN on 26 FEBRUARY 2010
Four types of quality inspection services are usually distinguished. Each one corresponds to a particular step in the production process.
They are all part of the toolbox of every importer, when it comes to buying in China and other low-cost Asian countries.
I tried to summarize the options available to buyers, in a visual manner.

The four types of quality inspection services:

All quality inspection services are not adapted to the same situation:

1. A pre-production inspection tells the buyer which kind of raw materials (or components) will be used. Factories are often suspected
of lowering their costs by purchasing substandard materials, and this can be disastrous for the buyer (e.g. the wrong kind of chip in an
electronic device).
The pre-production inspection can also focus on the processes followed as production starts. Sometimes this can also be critical, as
Chinese factories very often cut corners and do not respect the buyer’s blueprints (e.g. patterns for cutting fabric are received from the
buyer, and they are modified to make the process easier and faster).

2. A during production inspection (often called “DUPRO” in the industry) allows the buyer to have an idea of average product quality,
early in the production cycle. It is the most useful and the most under-rated tool at the disposal of importers, who often only rely on final
inspections.
It usually takes place once some finished products have come out of the lines. If quality issues are found, what is already produced might
be re-workable, and corrective actions can be taken for the rest of the job. It gives buyers the time to plan ahead, and even to avoid
delays (repairs and re-inspections take much more time when problems are noticed after all production is finished).
3. The final random inspection (also called “pre-shipment inspection”) is by far the most common type of QC check. It takes place
once 100% of shipment quantity is finished and at least 80% is packed, so it can be a real random inspection (this is not exactly the case
if quality is checked earlier earlier) and suppliers cannot play games.
It puts pressure on suppliers and gives power to buyers. Its objective is really to confirm a shipment’s quality, rather than catching issues
early. Therefore I usually advise my clients to complement final inspections with a DUPRO, to avoid finding disasters at the last minute.

4. The container loading inspection, like the pre-production inspection, it is seldom used. But it can be a worthwhile option in some
specific cases.
It can be useful if the buyer has a precise loading plan and needs it to be respected very precisely (e.g. some cartons are too fragile to be
placed at the bottom), or if the packaging is not conventional (e.g. some garments hung on bars, with no carton protection).

It can also ensure that the right kind of products are shipped out in the right quantity, when the importer places no trust in his supplier or
when several suppliers bring their products for consolidation.

Only the most sensitive projects require all four types of inspection. Generally, only one or two of these tools are used, depending on the
risks identified by the buyer.

These quality inspection services are used mostly for consumer goods involving little customization. Different approaches are often
chosen for ensuring that industrial products are up to specs (much more attention is spent during development and early production).

Quality control basic concepts

In this article I explain three fundamental concepts that every buyer should be familiar with when it comes to quality inspections:

1. Inspection levels
2. The AQL
3. When to inspect?
After 10 minutes, you will be able to (1) understand the reliability of an inspector’s findings and (2) take more informed decisions based
on an inspection report.

If you have not started doing professional quality control, you will need to understand these 3 concepts to make sure the inspection plan
meets your needs.

1. Inspection levels
Why use random sampling?
Shipments often represent thousands of products. Checking 100% of the quantity would be long and expensive. A solution is to select
samples at random and inspect them, instead of checking the whole lot.
But how many samples to select? On the one hand, checking only a few pieces might prevent the inspector from noticing quality issues;
on the other hand, the objective is to keep the inspection short by reducing the number of samples to check.

The relevant standards propose a standard severity, called “normal level”, which is designed to balance these two imperatives in the
most efficient manner.

Within this normal severity, there are three general levels: I, II, and III. Level II is used for more than 90% of inspections. For example, for
an order of 8,000 products, only 200 samples are checked.

Militari Standard 105 was created by the US Department of Defense to control their procurements more efficiently. In 1994 they decided
to rely on non-governmental organizations to maintain this type of standard. The ANSI, ISO, and other institutes all created their own
standard, but in essence they are similar. The major third-party QC firms use the same standards and the same statistical tables.

When to adopt a different level

Suppose you source a product from a factory that often ships substandard quality. You know that the risk is higher than average. How to
increase the discriminating power of the inspection? You can opt for level III, and more samples will be checked.
Similarly, if a supplier has consistently delivered acceptable products in the past and keeps its organization unchanged, you can choose
level I. As fewer samples have to be checked, the inspection might take less time and be cheaper.

The relevant standards give no indication about when to switch inspection levels, so most importers rely on their “gut feeling”.
The “special levels”
Inspectors frequently have to perform some special tests on the products they are checking. In some cases the tests can only be
performed on very few samples, for two reasons:

1. They might take a long time (e.g. doing a full function test as per claims on the retail box).
2. They end up in product destruction. (e.g. unstitching a jacket to check the lining fabric).
For these situations only, the inspector can choose a “special level”.

So we have three “general” inspection levels, and four “special levels”. For a given order quantity, each level gives a different number of
samples to check. Let’s see how it plays in two examples.

Example 1: you order 40,000 products

The number of samples to draw varies from 8 to 800. Depending on the level you choose, the inspection might take only one inspector for
one day, or up to 4 or 5 days of work.

Example 2: you order 3,000 products

Note: do not mix up “inspection levels” (I, II, III, S-1, S-2, S-3, S-4) and “inspection
severity” (reduced, normal, tightened). They are different concepts. The above
examples are based on a “normal” severity.

2. The AQL (Acceptance Quality Limit)

In part 1, we explained the different inspection levels that can be used. Another basic concept rings familiar to many importers, but is
often not clearly understood: the AQL (Acceptance Quality Limit).

There is no such thing as zero defect

First, as a buyer, you have to know what proportion of defects is tolerated on your market. If you are in the aviation business, any
defective part might cause a disaster, so your tolerance will be very, very low. But you will have to accept a higher percentage of defects
if you source consumer products that are assembled by hand in China or in India.

An objective limit is necessary

So, how many defects are too many? It is up to you, as a buyer, to make this decision. There are two reasons why you should not leave
this to the inspector’s judgment:

1. When it comes to giving instructions to an inspector, you should never leave gray areas—as they might open the door to corruption.
2. Your supplier should have clear criteria for acceptability, or they will see rejections as unfair.
The AQL is the proportion of defects allowed by the buyer. It should be communicated to the supplier in advance.

The three categories of defects

Some defects are much worse than others. Three categories are typically distinguished:

 Critical defects might harm a user or cause a whole shipment to be blocked by the customs.
 Major defects are not accepted by most consumers, who decide not to buy the product.
 Minor defects also represent a departure from specifications, but some consumers would still buy the product.
For most consumer products, critical defects are not allowed, and the AQL for major defects and minor defects are 2.5% and 4.0%
respectively.

Some important remarks:

 A professional inspector will notice defects and evaluate their category by himself. But it is better if the buyer himself describes the
most frequent defects and assigns categories to each one.
 Defects can be on the product itself, on the labeling or on the packaging.
 If one sample presents several defects, only the most severe one is counted.
How to read the AQL tables
The master tables included in the relevant standards are commonly called AQL tables. Let’s take an example.

You buy 8,000 widgets from a factory, and you choose inspection level II. In the table below (which is only valid for single sampling
plans), you see that the corresponding letter is L.

AQL table 1 (sample code letters):

Now let’s turn to the next table (which is only appropriate for normal-severity inspections). The letter L gives you the number of samples
to draw at random: 200 pcs.

And what about the AQL? Let’s say you follow the usual practice of tolerating 0% of critical defects, 2.5% of major defects, and 4.0% of
minor defects. The maximum acceptable number of defects is 10 major and 14 minor. In other words, the inspection is failed if you find
at least 1 critical defect and/or at least 11 major defects and/or at least 15 minor defects.

AQL table 2:
Notes:
 These 2 tables are only useful for inspections under normal severity. See all the tables for single sampling on AQL tables.
 More information about the AQL here.
Additional notes, for accuracy
 The number of defects is not the only cause for acceptance of refusal. The products can be refused because they are not conform to
the buyer’s specifications, even though their workmanship is very good.
 If you have two different products (made with different processes or in different factories), you should do two separate inspections.
If you inspect them together, one product might be accepted even though it presents too many defects. Why? Because the better
workmanship of the other product might “compensate” for its poor quality.

3. When to inspect?
The first two parts focused on the different inspection levels and on the AQL tables. So you know how to set the number of samples to
check and how many defects have to be accepted. With these settings and your detailed product specifications, a QC inspector can check
your products and reach a conclusion (passed or failed).

But importers face one more question: when should the products be inspected? This is an extremely important issue for buyers willing to
secure their supply chain. Spending a few hundreds of dollars to check and fix issues early can be an excellent investment; if might save
you weeks of delay, shipments by air, and/or lower quality products that you have to accept and deliver to your own customers.

Four types of inspections

Let’s picture the simplified model where one factory turns raw materials into finished products. (If you also have to manage the quality of
sub-suppliers’ products, the same model can be applied to them)

Pre-production inspection
This type of inspection is necessary if you want to check the raw materials or components that will be used in production. Buying cheaper
materials can increase a factory’s margin considerably, so you should keep an eye on this risk. It can also be used to monitor the
processes followed by the operators.

During production inspection

This inspection allows you to get a good idea of average product quality, and to ask for corrections if problems are found. It can take
place as soon as the first finished products get off the line, but these samples might not be representative of the whole order. So usually
an inspection during production is done after 10-30% of the products are finished.
Final (pre-shipment) inspection
Inspecting the goods after they are made and packed is the standard QC solution of most importers. The inspector can really check every
detail, including counting the total quantity and confirming the packaging. Final inspections are usually performed in a hurry, just before
shipment. To avoid creating delays, inspectors can usually start after all products are finished and 80%+ of the shipment quantity is
packed.
Loading supervision
In some cases, a buyer wants to make sure the factory ships the right products, in the right quantity, and with the right loading plan.

Further reading about the different types of inspections: quality inspection services.

4. Putting it all together

Example 1: 5,000 widgets from a new supplier
You have no information about the factory, so you should adopt level II.

The order quantity is comprised between 3,201 and 10,000pcs. In the first AQL table (higher in this document), you see the code letter is
L.

And in the second AQL table you see that 200 samples have to be checked by the inspector.

If you opt for the standard AQL limits (0 C. / 2.5% M. / 4.0% m.), the inspection is failed if at least one of these conditions comes true:

 One or more critical defects are found

 Eleven or more major defects are found
 Fifteen or more minor defects are found
And let’s say you want the inspector to check all the product functions on a few samples. This test takes some time. You can choose
special level S-2. You see this test will be done on 8 samples.

Let’s say the 200 samples can be checked by one person in one day, so a third-party QC firm would quote you one man-day.

When to inspect? If this is a standard product and you have flexibility with timing, a final (pre-shipment) inspection should be enough.

Example 2: 30,000 watches from unknown factory(ies)

You suspect that your supplier gave orders to several workshops to produce your goods, but he won’t tell you. And these items are rather
valuable. A level-III inspection is probably the most appropriate.

If we read the AQL tables as we did above, we have to inspect 500 samples, and the maximum number of defects is: 0 critical, 21 major,
21 minor.

If a visual check on all samples and a function/accuracy check on a few pieces is enough, it probably takes two or three man-days.

When to inspect? If the supplier refuses to disclose the factory information, you have to go for a final (pre-shipment) inspection.

Example 3: four different styles of garments from a good factory

You know that this factory’s workmanship is quite good. But you want an inspector to check all the conformity elements, and in particular
the measurements. Level I should be enough.

You sell these products in boutiques at a high price, so you can only accept 1.5% major defects (M) and 4.0% minor defects (m). Note: for
garments, there are generally no critical defects.
The fitting is quite important, especially for the brassiere and the brief, so 3 to 5 samples should be measured in each size.

There are 4 different types of products, so there has to be 4 inspections:

The total number of samples to check visually is 317 pieces. And 63 of these samples also have to be measured. So it will probably take 2
inspectors (maybe 3, depending on the products complexity) for one day.

When to inspect? The best is clearly during production, after enough finished products have come off the lines. As the 4 products might
not all be processed at exactly the same time, it might be preferable to send an inspector at different times. And if an inspection is
failed, a re-inspection can probably be performed at the supplier’s charge.

——-

Tired of reading long articles? See this illustration about Random Quality Inspections.

Statistical inspection levels: beware of the confusion

by RENAUD ANJORAN on 1 APRIL 2011
When I started doing QC inspections, my first client’s quality manager was interchangeably mentioning “level II” and “normal level”. And
same thing with “level I” and “reduced level”.

I purchased the relevant standard and I read it sideways (it is quite boring). I never picked on the difference between these concepts. But
this week someone at work told me I was wrong about it. I searched the internet and I found this comment from Tim Folkerts on
the Elsmar Cove forum:
The nomenclature seems to confuse people. Level II is considered standard, but it is not the same as “normal inspection”. Level I reduces
the sample size, but it is not the same as “reduced inspection.” Level II Reduced Inspection is roughly (but certainly not exactly) the
same as Level I Normal Inspection. And so on….

That opened my eyes on the difference between the “inspection level” (which is very often used by QC firms and their clients to adjust
the number of samples to check) and the “inspection severity” (which is supposed, in theory, to be set according to precise switching
rules).

What the standard says about inspection level:

The inspection level designates the relative amount of inspection. Three inspection levels, I, II and III, are given in Table 1 for general
use. Unless otherwise specified, level II shall be used. Level I may be used when less discrimination is needed or level III when greater
discrimination is required.

Four additional special levels, S-1, S-2, S-3 and S-4 are also given in Table 1 and may be used where relatively small sample sizes are
necessary and larger sampling risks can be tolerated.

What the standard says about inspection severity:

Normal inspection: use of a sampling plan with an acceptance criterion that has been devised to secure the producer a high probability
of acceptance when the process average of the lot is better than the acceptance quality limit.
NOTE Normal inspection is used when there is no reason to suspect that the process average differs from an acceptable level.
Tightened inspection: use of a sampling plan with an acceptance criterion that is tighter than that for the corresponding plan for normal
inspection.
NOTE Tightened inspection is invoked when the inspection results of a predetermined number of consecutive lots indicate that the
process average might be poorer than the AQL.

Reduced inspection: use of a sampling plan with a sample size that is smaller than that for the corresponding plan for normal inspection
and with an acceptance criterion that is comparable to that for the corresponding plan for normal inspection
NOTE 1 The discriminatory ability under reduced inspection is less than under normal inspection.

NOTE 2 Reduced inspection may be invoked when the inspection results of a predetermined number of consecutive lots

Now you know… Semantic differences are important!