ISO 22514-5 for process capability-Attributes

[ntmhx]

Hi everyone,

When dealing with attributive data, I’ve used tools like the P-chart or calculated and the defect rate over time. While this gives a good view of process performance, it doesn't provide a capability metric like Cpk or Ppk that we’re used to for continuous data.Recently, I’ve been reviewing ISO 22514-5, which provides a structured way to quantify capability for attributive characteristics, using probability-based indices that mirror the intent of Cpk/Ppk but adapted for binary outcomes. But I’m quite confused if these attributive indices give better insight into process capability, or are they mostly…I don’t know academic. Additionally

• Are you or your organization actively applying ISO 22514-5 for attributive capability analysis?
• Do you find it practical in real-world settings

 

[Bev D]

So I am known for my adamant opposition to capability indices, although I do understand that some aer under their Customer’s misguided thumb regarding the reporting of these indices. If you don’t have any Customer requirement for capability indices - don’t use them. They offer no real insight into your process as they are essentially a single number that is used to describe variation (that is a contradiction in terms). You are much better off plotting the data in a time series plot against the spec limits to understand the variation. Having done this for over 40 years in many industries, I can confidently say there is nothing better.

On to your actual question regarding translation of an attribute number to a capability index there is a formula. Basically continuous data indices a re converted to defect rates using the “Z table”. All you need to do is reverse the formula to get the capability index for a defect rate. In EXCEL, Ppk = Normsinv(Yield)/3. Do not use the percentage or ppm, use the actual proportion (a number < 1). But again this a lot of work for no real insight into your process.

 

[ntmhx]

So I am known for my adamant opposition to capability indices, although I do understand that some aer under their Customer’s misguided thumb regarding the reporting of these indices. If you don’t have any Customer requirement for capability indices - don’t use them. They offer no real insight into your process as they are essentially a single number that is used to describe variation (that is a contradiction in terms). You are much better off plotting the data in a time series plot against the spec limits to understand the variation. Having done this for over 40 years in many industries, I can confidently say there is nothing better.

On to your actual question regarding translation of an attribute number to a capability index there is a formula. Basically continuous data indices a re converted to defect rates using the “Z table”. All you need to do is reverse the formula to get the capability index for a defect rate. In EXCEL, Ppk = Normsinv(Yield)/3. Do not use the percentage or ppm, use the actual proportion (a number < 1). But again this a lot of work for no real insight into your process.

I totally agree with your point. However, I’ve noticed that there is an internal instruction referencing this ISO standard within the company. As a new recruit, I understand that I may need to justify the omission of its use. In this case, the most reasonable justification would be that applying ISO 22514-5 would not enhance decision-making, I guess it wont be difficult since there is no customer requirement.

Thanks for answering

 

[Bev D]

You can also check out my paper on statistical alchemy. In it I address many of the drawbacks of capability indices. There are also references for further study; many are free…

 

[Miner]

I would also consider which value provides more useful information on which to make a decision. Which gives you a more intuitive understanding of the magnitude of your problem: 1) a percent nonconforming = 0.1%, or 2) a Cpk = 0.77? The first is information that can be interpreted directly. The second is information that is encoded in an index that is not easily interpreted. Its major value is when you are not producing ANY nonconforming and gives you a feel for how much safety margin you have. If you are producing nonconforming, you already know you have used up any safety margin and really need to know the magnitude of the problem.