Those of us who are accustomed to work with standards like ISO/TS 16949 are also accustomed to “hear voices”: the voice of the customer, the voice of the stakeholders, and so on. The only voice we are not accustomed to hear is the voice of quality. Perhaps this should be called the “sounds of silence,” since quality managers are almost invariably the last to be listened to, but the first to be accused during any quality trial.
The point I want to raise concerns measurements, the devices used to produce measurement data, and how these devices are used. I don’t intend to accuse the entire device industry; I only want to give some indications on how this process could be made more effective, based on my field experience.
First, let’s acknowledge that quality assurance is still a legend—quality control is the reality. Therefore we must take measurements, and therefore we need measurement devices. Statistics are used to evaluate but not to predict process performances. Consider the automotive industry, which frequently requires Cpks higher than 1.67 and 100-percent inspection. While developing new products or processes, or transferring production from one site to another, manufacturers focus not on understanding the process and the products that result. Instead, they focus on measuring the product, therefore demonstrating how little confidence there is in the production process—or their knowledge of it. It reminds me of when I worked at developing structural adhesives: “It works, but let’s add some fasteners” (and a belt and braces).
What does this approach cost manufacturers? And is it a quality or a nonquality cost? If a process is stable, and I get the proof via the statistical studies, why should I intensely measure its output? That takes time, which contributes to higher production costs, and the process requires measurement devices, which increases assets costs. And that’s not factoring in the cost of controlling these devices.
I’m not saying that we shouldn’t control the products, but we should control them more effectively. If a measurement system that uses one CMM turns out to be more effective, and more cost-effective, than a system made of 50 or more calipers, well, let’s go with the CMM. If I want to prevent using the wrong sort of steel to make my fasteners, why would I test them after production? Wouldn’t it make more sense to keep the production process under control? Shouldn’t I test the incoming bars with a spectrometer and forget about the after-production quality control?
The Automotive Industry Action Group’s (AIAG) Advanced Product and Quality Planning (APQP) manual is a quality bible for many of us. Its introduction clearly implies that measurements—both of the process and the product characteristics and parameters—need to be identified. But instead of putting into practice the “do what’s necessary,” principle from the now-defunct QS 9000, the more common approach to this mandate is, “Let’s measure them all.” To make matters worse, this “belt and braces” approach drives many managers to ensure their measurement devices are calibrated by independent, ISO/IEC 17025 or equivalent, accredited laboratories.
Again, I’m not totally against this approach, yet common experience has shown me that:
• Most companies do not define and document the acceptance criteria of calibration results. They simply file the calibration lab certificate. Often they don’t even read it.
• Accredited laboratory certificates don’t document an evaluation of the device.
• Quality managers rarely show even the basic skills (such as knowledge and competence) of metrology. Therefore they neither know nor understand the risks of the measuring process. I often wonder whether they even think of subjecting the process to an FMEA-like risk assessment. Their approach is like how we use a scale to keep our own weight under control, or a thermometer to monitor fever: We assume the device works correctly, and it doesn’t occur to us to take it to an accredited lab for calibration.
A well-known ball-bearing manufacturer has a process by which 100 percent of the bearings are inspected before packaging. Each inspection device costs about $80,000, but the devices inspect only two-thirds of the bearings’ surfaces. As for the process controls, well, we’d better off predicting the outcome of a horse race.
Accredited laboratories could do a more effective job by educating manufacturers about the risk implied in controlling devices used for process and product measurement. Measurement device manufacturers should work more closely with their customers to develop more effective and meaningful measures, and devices.
Among the many slogans about quality, we should include the principle, “Correctly measure what should be measured when it’s correct to measure it.” I wish both the measurement device industry and accredited laboratories would educate their manufacturing customers to a more sophisticated measurement culture. Then maybe we’ll start hearing the voice of quality assurance.