A reader asked me an interesting question:

Every article I read on your blog assumes machined manufacturing and I wondered how would quality control be different with handmade products where human factor play the most important role?

The short answer is: no I don’t assume machined production, and yes quality control should take this information into account (but only when production is under way).

And here is my long answer.

The notion that “quality is uniform since the products are machine-made” is very dangerous.

Sometimes it is true. Here are a few examples:

• All the pieces go through one machine that is precise but biased (poorly calibrated). All the pieces come out too thin.
• The thread used for the bartack stitching operation is in the wrong color.
• A wood furniture manufacturer selects an out-of-date program for the automatic cutting machine. All the products follow an old design the importer doesn’t want.

However, it is often wrong. Here are some counter-examples:

• Half the components are out of specifications but they are used in production anyway because incoming QC procedures are not appropriate and/or not effective.
• Four similar machines were used simultaneously to produce a batch. In one of them, a critical component is severely damaged (inducing much variation) but has not be noticed yet by the maintenance technicians. (Note that maintenance is very poorly done in most Chinese factories.)
• One of the process steps is to harden the products in a kiln. The kiln worked fine for 90% of the batch, but a power shortage means 10% of the batch was not processed correctly. The factory naturally assumes they can ship those goods anyway — after all they did go through the kiln and they did get a heat treatment.
• A testing station doesn’t detect a serious flaw in an electronic product.
• The factory ran out of components after making 95% of the batch because of unforeseen scrap. The quantity to replenish is too low, and the lead time to get it is too long. The factory decides to ship that scrap to its customer. This is unfortunately frequent in China.

Now, what are the implications?

1. From an end-of-production-quality-control perspective I would follow the same inspection procedure for all types of products. The inspector needs to do a proper sampling and takes pieces from different cartons, since 5% or 10% of the order might be defective.

The right approach for final random inspection is “make no assumptions”. Skipping some checkpoints or reducing the sample size is quite dangerous if the buyer is not fully aware of what can go wrong in production. From experience, I believe the ISO 2859 (AQL) standard is appropriate whatever the production process. It has the advantage of being simple and of being widely known by Chinese suppliers.

2. From a process control perspective there are differences.

2.1 For a highly automated process, the manufacturer needs to pay a lot of attention to these elements:

• Management of setup programs
• Preventive maintenance of automated equipment
• Calibration of gauges
• In certain industries, statistical process control

And in-process quality control should be regular along the production cycle. Not only at the very beginning.

2.2 For a manual production process, critical success factors are listed below:

• Training the staff and writing good work instructions
• Having operators fully understand that bad work should never be passed to the next station
• Coming up with good tool, dies, or fixtures to make it easier to do a good job
• Mistake proofing operations as much as possible

3. From a “it is not acceptable and you need to do re-work” perspective it is also different.

In most cases, a manual sorting is necessary after some issues were found. The importer asks the manufacturer to check every piece, to sort out the “bad” ones, and to re-work/reproduce/discard them. A Chinese factory that focuses on labor-intensive work will have a lot of people to throw at this job (but they might be unable to find the defects that caused the rejection), while the more automated supplier might be unwilling to do so.

Let’s assume the sorting was done properly. The second step is, in most cases, to rework the faulty pieces. If processes are manual it is often possible (but the question is, will do it right the second time if they messed up the first time?). In machining or fabrication workshops, it is often impossible — they might have to scrap those parts and re-produce them, which is often long and costly.

—

To recap, in a final inspection the buyer should not assume “this comes out of a machine so I can check one piece and all the others will be exactly similar”. But I strongly encourage buyers to pay attention to the factory’s process controls. We have found nearly every time that improving process controls results in better quality and lower costs.

What do you think?