How high should Cpk be? Why 1,3 - 1,67 is the sweet spot and 2 is too high.

There is a school of thought that suggest you should have a Cpk of 2.0 - that is, in fact, very much in line with Six Sigma: having 6 standard deviations of process variance between the process mean and a specification limit will yield a Cpk of 2. But I would argue that this is too high for most companies and industries.
In this video, we'll cover some rules of thumb for how high your Cmk, Cpk and Ppk should be; what the difference between those three is, to begin with; and how you can decide on a proper target for your organisation.
In short, I suggest a Ppk of more than 1.0, a Cpk of at least 1.33 and a Cmk of at least 1.67 - check out the whole video for the reasoning behind this.
#continuousimprovement #cpk #sixsigma
• How high should Cpk be...
00:00 Understanding Cpk Basics
02:56 Process Performance vs. Process Capability
06:46 Buffer Capability for late Detection
07:36 Machine Capability (Cmk) Explained
11:08 Debating the 1.67 Rule
14:04 Why Not Aim Higher Than 1.67?
19:37 Conclusion and Resources

[@paulallen5321]

Hi Tom - I would say look at the taguchi loss function, which says that the better the Cpk the lower the loss to society. It's another point of view on how high should your CPK be? The Batavia gear box plant is a great case study to look at as well....

[@TomMentink]

Hi Paul, thanks for you reply, those are very valid points indeed. For anyone interested in this great video by Paul: ruclips.net/video/vH-1deoUIZw/видео.htmlsi=f3ZiBfe7v36B2IBp)
I agree that continuous reduction of process variability will lower the loss to society (and is almost always beneficial for the producing company too).
The issue I have with Process Capability (variation divided by specs) is that this ratio does not inherently benefit society, because the specification limits are more or less arbitrary. For a quality characteristic that applies to the design or direct function of the part (like the size of a nut, the positioning of a label, etc.) it's best to be in the centre, yes. I would still argue that you should consider transferring this decreased process variability downstream by promising tighter specifications (especially if the customer is willing to pay a premium for it - something Ford's competitors did, by the way, and that drove market for Mazda).
When we're talking about a quality characteristic that is more like unilateral tolerance (like weight, break strain, etc. that usually have a hard minimum and a maximum that's more to have product categories), I would say it's often wasteful to stay more than 4-5 SD's away from that limit. In fact, especially in filling weights/volumes, you'll often see companies choose to stay much closer to that minimum, reducing the average when they successfully reduce process variation.

[@paulallen5321]

Good point about the specs - The loss function ignores the tolerances, as they are as you say arbitary often, when they should include an element of the voice of the customer. Good stuff as always Tom...

[@TomMentink]

@@paulallen5321 and thank you for the challenge - it’s always good to add these extra nuances and the different angle of approach/thought 👍🙏