We all talk about the invention of the airplane, the automobile, or other important products, but we rarely talk about the fascinating history of Quality, so here is a snapshot.

For those who believe the Bible, I would say that God was the first Quality Inspector, since we read in the first chapter of Genesis that as He created the world, He frequently "inspected" His creation, and "saw it was good" (Note: I was not there. I am old, but not that old).

For most of antiquity, the artisan who created a product determined if it was acceptable, so Quality was highly dependent on who made the product.

When we got to Medieval times, the Quality movement started to take a new shape, as in Europe the craftsmen began organizing in guilds, which were responsible for creating strict rules for product and service quality. Inspection marks served as proof of Quality.

The Industrial Revolution brought new challenges and practices in the factories, with a reduction on the individual responsibility for quality and the creation of dedicated quality inspectors.

During World War proper inspection of military equipment was crucial, but 100% inspection quickly became unviable. Sampling tables started to be created, and the first versions of statistical control were implemented (Shewhart’s Statistical Quality Control). The process focus had started.

After World War II we had the golden era of the Quality Gurus:

• Deming (1900-1993) – Quality Management Principles / Statistical Quality Control
• Juran (1904-2008) – Quality as Fitness for Use / Quality Planning-Control-Improvement Trilogy
• Ishikawa (1915-1989) – Fault Tree Analysis
• Taguchi (1924-2012) – Design of Experiments
• Shainin (1914-2000) – Red X
• Shingo (1909-1990) – Poka Yoke
• Feigenbaum (1922-2014) – Total Quality

Total Quality Management was born. The responsibility for Quality started to go back to the manufacturing personnel, instead of a separate Quality group.

Towards the end of last century we saw the standardization of quality systems becoming the backbone of the quality initiatives. Audits and certifications took the center stage.

The last decades saw a sharp increase in automated 100% inspection methods being incorporated to the processes: Robots, Cobots and Humanoids are becoming commonplace in our manufacturing plants.

Looking at all these steps, I would venture to answer the question this way:

Who invented Quality?

We all did, and continue to do it, on a daily basis.

Thank you, Mr. Deming. Now Goodbye.

Overlapping the timeline above with my own career in Quality allows me some observations.

When I started working in the automotive industry, the quality department had an army of people, mostly tasked to inspect products all day long.

Inspections were done in large samples or even 100% and finding more than 1% bad parts was normal.

Those days are gone.

For a production employee to make bad parts that were not detected by the inspector was like cheating in his tax forms and not being caught by the IRS. He would brag about it at the cafeteria.

Those days are gone.

On the other side of the globe Mr. Deming was demonstrating one could be confident that (almost) only good parts were being produced, just by inspecting a well-defined sample of the production volume, if only some statistical rules were observed.

• Thank you, Mr. Deming, for teaching me to use statistics in the real world.
• I truly enjoyed working under your mentorship for several of my years in Quality.

This side of the globe someone noticed that the machine operators had some free time while they watched the chips flying that they could use to measure their own parts.

The combination of those brought us "Self-Inspection" and "Statistical Process Control."

That is when people started talking about 25 ppm being the target.

Those days are also gone.

Don’t get me wrong: you still need capable processes, so you don’t produce bad parts, but that is no longer enough.

Have you heard of the "Ghost of Special Causes"?

Have you had to sort thousands of suspect parts because the customer found one bad unit?

Now our world expects zero defects.

Should we go back to an army of inspectors?

Not a chance (didn’t we say that those days were gone?).

Fortunately, while those eras were coming and going, the cost of automated inspection went down and became more reliable (cameras, lasers, LVDTs, functional testing, AI, etc.).

Most new processes are already developed with 100% automatic verification built in.

Those days are here.

For the old processes that do not have such a luxurious feature, it becomes our job, as quality professionals, to look for opportunities to implement them.

Guilty Until Proven Innocent

The United Nations in its Universal Declaration of Human Rights states:

"Everyone charged with a penal offense has the right to be presumed innocent until proved guilty according to law in a public trial at which he has had all the guarantees necessary for his defense."

It is time to change that.

Instead of "innocent until proven guilty," we should have "guilty until proven innocent."

No, this is not a proposal for a totalitarian government that does not care about human rights.

This is a proposal on how we should look at quality verification.

As the customers expect zero defects, we need to have assurance (not assumption) that every part we make is correct.

So, every time we set up a die, a furnace, a machine, or any kind of equipment, we must make a thorough analysis to prove to ourselves the part is good.

In many cases our company has invested a lot of money to implement automatic verification of important characteristics. This is the situation where the "guilty until proven innocent" concept works the best.

Using a counterexample to demonstrate this:

• Years ago, we had a customer complaint about the run out being out of specification in some parts.
• We were very surprised because we had 100% automatic verification of run out in the process.
• As we investigated it, we found out the cable had disconnected from the probe, so even if a part was bad there would be no signal to reject it, so the automation would not remove it from the flow.

The real problem was the logic in the system was:

"Every part is okay, unless there is a signal from the probe indicating it is bad."

It should have been:

"Every part is bad unless there is a signal from the probe indicating it is good."

Do you get it?

If the second logic was used, when the cable disconnected 100% of the parts would have been rejected, and somebody would stop production, investigate and solve the issue, without any bad part going to the customer.

Again: "guilty until proven innocent."

From that experience we determined that the logic in automatic checking equipment should always expect a good signal before approving the part.

The Red Rabbit

While over the last decades, as we progressively implemented more stations verifying products automatically, many of us got burned when bad parts were still shipped to the customer.

Since most automated checks have some electronics and software logic behind it (both for the inspection itself, and for the automatic removal of the rejected parts), it may be difficult to verify all the possible scenarios during the equipment run off.

Besides that, things may change over time and get us in trouble (sensors may go bad, connections may get loose, etc.).

Although still imperfect in some cases, the best solution is to have a solid plausibility verification done with negative masters (or red rabbits) to confirm the system is functional.

Some guidelines we developed over the years are the following (please take these seriously):

• Red Rabbits should be required by the Control Plan (I hate when I hear someone saying: "I didn’t know I needed to do it")
  • It is necessary to clearly determine the frequency of the checks and the expected reaction plan (if the Red Rabbit is not rejected, something needs to be done immediately)
• A Red Rabbit should be a dedicated part
  • Don’t make a bad part each time to test the system
• Red Rabbits should be clearly identified (preferably red) and numbered
  • I could tell you how embarrassing it is when the customer calls to report we shipped the negative master (believe me, I have been there)
• Red Rabbits should be just marginally out of specification
  • A sample that is grossly out of spec may not guarantee a marginal part would be picked
• Red Rabbits should be used at least every shift and every part number change
  • Possibly more than once a shift
• Red Rabbits checks should be recorded (this way, the operator will have to do the check or lie about it – no "I forgot to do it" excuse)
  • If the operator is required to perform the check, he or she should be required to record that they did it (preferably recording an actual reading, if possible)
• Red Rabbits need to be in the calibration system
  • They need to be verified periodically to make sure they are still adequate samples (even if they are the attribute type)

You may be thinking of a very specific situation where a negative master is not possible or practical, but in most cases it is, so let’s do it right.

Conclusion

Quality has evolved over the years and we, as Quality Professionals, need to evolve with it, or even better, drive the evolution.

What is next in your industry?

What is next in your plant?

Let’s be inspired by this history and write the next chapters ourselves.