Improvement Actions for Unqualified MSA Results: 4 Strategies

The Day the CPK Report Came Out, the Entire Room Fell Silent for Three Seconds

I remember several years ago, a new machine was about to go live in our factory, and everyone was incredibly busy. Production lines, processes, equipment – everything was in motion. After much effort, the machine was finally tuned to a point where we could see light at the end of the tunnel. With CPK already at 1.08, Ah-Fa from the process team patted his chest and declared, "It's stable!" However, once the MSA was run and the data came out, the meeting room instantly fell silent, with only the hum of the air conditioner. The feeling was like thinking you've become a martial arts master, only to be knocked back to square one by a child's punch. At that moment, I knew we had a big problem.

Where the Problem Lies: Simply Put, 'Inaccurate Measurement'

You might think, isn't it good enough if CPK is sufficient? Why would there be a problem if MSA fails? Simply put, CPK talks about your process capability, meaning whether the output from your machine is "good enough." But MSA (Measurement System Analysis) talks about whether your "measurement system" is accurate enough. Imagine your home weighing scale is broken, adding 5 kg every time you weigh yourself. Even if you've lost weight, the scale still tells you you're fat. This is the problem with an inaccurate measurement system. You simply don't know if your product is genuinely good, or if it's just the measurement system's error making you think it's good. For values like DPMO 6210, if the measurement system is inaccurate, you're essentially treating trash as treasure.

So the key point is, if your measurement system itself is inaccurate, then no matter how much effort you put into improving the process, it's like building a house on sand – without a stable foundation, everything is in vain.

What to Do in Practice? Four Strategies for Your Reference

If MSA fails, the problem usually revolves around "repeatability" and "reproducibility." Repeatability refers to whether the same person using the same instrument to measure the same item gets consistent results. Reproducibility refers to whether different people using the same instrument to measure the same item get consistent results. The following four strategies are the most effective and common ones I've encountered over the years:

1. Standardized Measurement SOPs: The most basic and often overlooked. Many people assume they know how to measure, but each person's posture, force, or even placement might differ. Write an extremely detailed SOP, preferably with illustrations and text, clearly outlining every step, then rigorously train operators. We once saw repeatability drop from 80% to 60% due to a minor probe angle, which was only restored after retraining.
2. Instrument Calibration and Maintenance: This is the equipment engineer's responsibility, but the process team must also follow up. Instruments age and drift; regular calibration is extremely important. Sometimes erratic measurement values aren't product issues at all; it's just that your instrument needs to be sent for calibration. We once discovered a regular deviation in measurement values and, upon investigation, found that the calibration cycle was too long, leading to a decrease in precision.
3. Fixture and Clamping Method Improvement: Often, human error arises from the lack of good fixture assistance. If your product needs to be handheld during measurement, reproducibility can easily become an issue. Design a fixture that can securely hold the product, reducing human intervention and significantly improving stability. Previously, for a wafer thickness measurement, without a dedicated clamp, the results fluctuated every time. After creating a custom clamp, reproducibility immediately improved.
4. Operator Retraining and Qualification Confirmation: This is a reality; some operators are less patient or have poorer comprehension skills. Regular on-the-job training, and even re-evaluation of skills, is necessary. Don't worry about hurting feelings; the ultimate goal is to improve quality. Sometimes, simply changing the operator resolves the issue.

So the key is to think about which link in the "Man, Machine, Material, Method" (4M) aspects has gone wrong.

The Most Common Pitfall: Believing a New Machine Will Fix Everything

The most absurd situation I've encountered was when MSA failed, and a senior manager immediately said, "Is the instrument too old? Why not just get a new one?" As a result, they replaced it with a new, multi-million dollar measurement machine, but the MSA results were still just as bad. Do you know why? Because the problem wasn't with the machine at all, but with "Man" and "Method." Operators weren't properly trained on SOPs, and measurement methods weren't standardized at all. You could replace it with an even better instrument, and it would be the same. It's like driving a Ferrari to climb a mountain and then complaining the car is terrible. Simply put, you didn't understand the root cause of the problem at all.

Frankly, often, human "expediency" is the MSA killer. Everyone wants to be fast and save trouble, but it's these small details that turn your data into a pile of waste paper.

One Thing You Can Do Today

Go back and check your measurement SOP. Is every single step clearly written?