Switching from Acceptance Sampling to Process Control

Most importers who buy products in China resort to acceptance sampling — in particular, QC inspections based on AQL limits.

The main reason is, they can’t trust their suppliers. Actually, they often don’t even know where certain orders is made! This situation is not ideal (they’d rather avoid paying for inspections) but is manageable for certain buyers. Many importers are happy so long as they don’t receive a shipment with more than 4% of defective products.

Acceptance sampling is not suitable for certain importers

However, some importers have higher requirements. For example, we work for several companies that purchase precision machined parts that need to be assembled in another plant, or empty shampoo bottles that will be fed to a filling machine in the importing country.

In these cases, having 1% of products out of specifications results in high costs. So the question becomes, how to adopt a more mature and more pro-active quality assurance strategy?

3 solutions for ensuring better quality

Quality management system (QMS)

There are many QMS standards (ISO 9001 for most industries, ISO/TS 16949 in the automotive industry, etc.). A manufacturer that follows one of these standards seriously will become more reliable.

(Notice that I didn’t write “a manufacturer that got certified is more reliable”. I don’t give much importance to company certifications.)

The danger with a manufacturer that “wings it” and doesn’t put a system in place is that products are checked only when employees have time for that activity. Entire batches are commonly shipped out without any measurement or testing! No matter QC inspections are so frequent in China.

Once a supplier pretends to respect a standard, the main question is, “is it effective”? Most Chinese manufacturers are very cynical about management systems and see this as paperwork to prepare before the auditor comes. A good auditor will uncover that.

Statistical process control (SPC)

These days, in the US and Europe, SPC is seldom used as intended when Shewart developed control charts in the 1920s. The reason is, all these calculations are often embedded in modern machines, especially when it comes to mature processes. So why care about SPC?

Well, in China there is a habit of buying cheap machines and throwing them away after a few years. There is a lot of variation due to the machining & fabrication equipment, and good old SPC tools can help keep it under control.

Certain industries still use SPC data (for example, suppliers of car parts are often required to have a Cp of at least 1.33). Several standards explain how to correct and analyze SPC data. It only requires one person who understands how process control statistics work for maybe 1 day, to set a basic system up.

The analysis of the variations, for example in a few critical dimensions of a machined product, is the starting point of process improvements. See this video about SPC for a few pointers.

Proper calibration

Garbage in, garbage out… You need to make sure decisions are made based on properly collected data.

In theory, an ISO 9001-certified manufacturer makes sure calibration is done right. But I have yet to see an auditor go beyond checking if a sticker is on a few instruments and comparing that to a log. That is insufficient!

Very few Chinese factories evaluate the R&R (Repeatability and Reproducibility) of their gauges. If measurement/testing precision is very important, it is worth the buyer’s time to collect some data in the factory.

Here are two important things to check:

  • If the P/T (Precision to Tolerance) ratio is lower than 0.1, all is good. But most of the time if will be between 0.1 and 0.5!
  • If the primary standard is at least 10 times more precise than the measurement instruments, all is good. If it is the same precision, some variation due to the measurement method will be unavoidable.

Can you think of other solutions? What do you think?

Best Practices for Managing Chinese Suppliers [Upcoming Conference]

The European Chamber is organizing the second edition of the “Profitable Sourcing” series of conferences. The previous session, in February, was quite a success. It attracted close to 70 participants and was extremely interesting.

I hope to see you on Nov. 11 (afternoon only) in Guangzhou.

Who will the speakers be?

  • The head of Manufacturing Excellence for Footwear at Adidas
  • The president of the FAW-Volkswagen Foshan branch
  • A supply chain executive from Bureau Veritas (to discuss the role of factory audits and similar services)
  • David Collins, one of my parters at China Manufacturing Consultants, will go over two case studies and will explain how his team turned two factories around — with plenty of details about the challenges encountered, the tools used, and the results obtained.
  • And myself, mostly to conclude the conference.

We might add one more speaker — the details are not 100% confirmed yet.

The overal topic this time is how to manage suppliers to reduce risks and attain superior performance.

For the details of the conference, you can click here.

The 3 Types of Hong Kong Trading Companies

Sourcing New SuppliersAfter working myself in a Hong Kong trading company a few years ago, and then observing my clients’ suppliers, I have noticed a lot of similarities among these companies.

I think we can distinguish between three types of trading companies:

1. Good salesmen

They are good at getting new clients… but not necessarily at keeping them.

They are often creative. In the garment/fashion industry, they often have at least one designer who proposes ideas. It is very easy for overseas buyers to develop new products with them.

I have observed a few such companies. They are often managed by a European or an American who regularly wins new accounts from his own country. But very little attention is given to work flows, structures, formats of critical working documents, staff training, IT systems, and the like. The management seldom gets out of Hong Kong to visit factories, except when customers ask for a visit.

The downside is often that the production follow up is inconsistent, and they have a hard time keeping their good clients for more than a few years.

2. Small brokers

These traders find small factories that are interested in small quantities but don’t have the capability of dealing with export customers or processing export documentation.

The obvious advantage is that MOQs (Minimum Order Quantities) tend to be low. And prices tend to be on the low side, too.

What importers should question is the type of production follow up that is taking place, and the grade of the workshops used (generally quite poor). And this is NOT the type of supplier that will help you develop your complex, custom-made cool new product.

3. Real partnership with a good factory

A few trading companies are closely associated with one factory. They might place small orders, or (especially) products not made by that “primary” factory, in other factories. But most of their orders go to that one factory. One can expect the factory owner to consider the trading company as an extension of his organization — maybe as a sales office.

The enormous advantage they get is control over what the manufacturer does. The lack of such control is what frustrates all other traders to no end. (Note that this is not always the case, though.)

It means these trading companies usually offer reasonable quality, can communicate well and be responsive, and generally speaking they “walk their talk”.

What they might lack is creativity, understanding of export markets, breadth of product offering, and… low prices (though not necessarily).

Overall, this category is severly under-represented. I wished there were more of them.

Do you see a 4th and maybe a 5th category?

The Different Sampling Plans Contained in the ISO 2859 Series

The vast majority of inspectors follow the ISO 2859-1 standard. But about ISO 2859-2, -3, and so on?

The relationship between these parts is explained in ISO 2859-10:2006 (introduction to the ISO 2859 series of standards for sampling for inspection by attributes):

⎯ Part 1: Sampling schemes indexed by acceptance quality limit (AQL) for lot-by-lot inspection

⎯ Part 2: Sampling plans indexed by limiting quality (LQ) for isolated lot inspection

⎯ Part 3: Skip-lot sampling procedures

⎯ Part 4: Procedures for assessment of declared quality levels

⎯ Part 5: System of sequential sampling plans indexed by acceptance quality limit (AQL) for lot-by-lot inspection

⎯ Part 10: Introduction to the ISO 2859 series of attribute sampling standards

As I explained in this video, part 1 of ISO 2859 (based on the “AQL tables”) is not to be used systematically. It should be reserved for business with regular suppliers. The standard explains it this way:

ISO 2859-1 was developed primarily for the inspection of a continuing series of lots all originating from the same production or servicing process.

For one-shot orders, or for new suppliers, part 2 is to be used. The reason is, the switching rules can’t be used since there is not a regular flow of batches to inspect:

ISO 2859-2 provides sampling plans applicable for use when individual or isolated lots are to be sampled. These sampling plans are in many instances identical to those in ISO 2859-1.

Is it a big problem if part 1 (based on the AQL tables) is used in circumstances where part 2 is actually applicable? Not really.

Sampling plans in ISO 2859-1 may also be used for the inspection of lots in isolation, but in this case, the user is strongly advised to consult the operating characteristic curves to find a plan that will yield the desired protection.

When to use part 3, and check only random batches rather than all batches? For very good suppliers only:

ISO 2859-3 provides skip-lot procedures for use when the process quality is markedly superior to the AQL for a defined long period of delivery or observation.

And what about part 4?

ISO 2859-4 provides a procedure that may be used to verify a quality level that has been declared for some entity.

The procedures in ISO 2859-4 have been developed in response to the need for sampling procedures suitable for formal, systematic inspections such as reviews or audits.

And part 5, which is particularly complex?

ISO 2859-5 provides a method of establishing sequential sampling plans of discriminatory power essentially equivalent to that of corresponding plans of ISO 2859-1.

In conclusion, QC practitioners tend to make their job easy and simplify the rules to follow. I think it is, to a large extent, not a problem. The main issue is that very few people follow the switching rules (which influences the number of samples and/or the limits for nonconformities, based on a score earned by each manufacturer).

By the way, are you wondering what the difference is between “inspection by attributes” and “inspection by variable”? See The different types of sampling plans for QC inspections.