That day the CPK report came out, and the entire room fell silent for three seconds
I still remember many years ago, we purchased a new measurement machine, preparing for acceptance and deployment. As per usual, the equipment vendor would run an MSA report to ensure the machine's measurement capability was sound. However, once the report was out, the CPK was only 1.08, and the DPMO was as high as 6210. The entire conference room instantly fell silent, and everyone exchanged bewildered glances. The boss's face turned green, and he directly asked, "What the hell is this? Can this machine even be used?" The equipment vendor was also embarrassed, stammering an explanation: "Well... the R&R passed, but this NDC..." I muttered to myself, "Damn it, passing R&R doesn't mean it's usable! This NDC is the absolute key!"
Where's the Problem?
To put it plainly, NDC (Number of Distinct Categories) essentially refers to whether your gage is "sensitive enough." Imagine you have a ruler with only centimeter markings. If you need to measure millimeter differences, this ruler would be useless, right? That's precisely what NDC indicates. It tells you how many "distinct" product categories your measurement system can differentiate. If your gage cannot even distinguish the variation within the products themselves, then repeated measurements are futile; it's simply a waste of effort.
Therefore, the higher the NDC value, the better. It is generally accepted in the industry that an NDC of at least 5 is considered acceptable. If your NDC is less than 2, your measurement system has virtually no discriminatory power and is essentially measuring randomly. In such cases, even if your R&R data appears to pass, it's misleading.
How to Practically Apply This?
Determining NDC is actually quite straightforward. It will be directly displayed in your MSA report. When you run a Gage R&R, the report results typically include this value.
Let me give you an example. Suppose you have a batch of products with a specification tolerance of 100±5um. You use a measurement machine to measure them, and the measured data range is only 2um. However, your measurement machine's error for each measurement is 1um. In this scenario, your measurement machine cannot distinguish the 2um difference between products because its own error accounts for half of that. In this case, your NDC value will be very low.
In other words, if your gage variation is too large, or the product's inherent variation is too small, both will lead to insufficient NDC. Your gage must be able to "see" the differences between products, and this difference must be greater than the "noise" of the gage itself for NDC to be high.
The Most Common Pitfall
The most common pitfall I've encountered is focusing solely on the %R&R value from a Gage R&R report. Many people, upon seeing %R&R less than 10% or 30%, assume everything is fine. Frankly, this is a huge misconception! The CPK 1.08 case I just mentioned is a perfect example of this. That machine's %R&R was actually 15%, which seemed acceptable, but its NDC was only 1!
What does this signify? It indicates that while the measurement's "repeatability" and "reproducibility" appeared satisfactory, the system fundamentally could not distinguish good products from bad ones. Even if you measured a defective product, the data it produced might be similar to that of a good product because its discriminatory power was too poor. The result is that you spent a lot of money on equipment that merely creates an illusion of "precision" but cannot genuinely help you improve your process.
One Thing You Can Do Today
Go back and check your latest MSA report for the NDC value!