Occurrence Ranking: What is the basis for 1 in 1,000,000?



I have a customer holding me to the AIAG Manual word for word. One of there complaints is my low occurrence ranking but our supplier is only supplying 480 pieces/parts per year. So it got me wondering where does 1 in 1,000,000 come from? Or more clearly, how often does they 1,000,000 occur? There are some companies that manufacturer that quantity of parts of couple months and there are companies that do that every couple years. Or does time not come into this equation? Thanks for your help and time.

Steve Prevette

Deming Disciple
Super Moderator
I suspect it is the same reason that OSHA chose to calculate injuries per 200,000 hours. Counting defects per million gives you a number that should just be a few digits long, especially if you are anywhere near "six sigma capable".

Easier to say the defect rate is 25 per million, rather than 0.0000025 (do I have the right amount of zeroes there?) or 25,000 per trillion.
Last edited:

Ron Rompen

Trusted Information Resource
I have run into this arguement/discussion a few times with customers (and auditors) over the years. The best response that I have been able to come up with is based on statistical process capability - if you have a stable process then it is (relatively) easy to determine the statistically probable number of defects, either in percent or in parts per million. Use that to determine your occurence rate.
With that being said, the PFMEA manual (pg 92) states very clearly "A consistent occurence ranking system should be used to ensure continuity. The occurrence ranking number is a relative ranking within the scope of the FMEA and may not reflect the actual likelihood of occurrence"
You don't have to use the OCC table in the manual, and you don't have to be tied strictly to 1 ppm for an occurrence of 2 - it is only a guideline.


Well if you dont have similar products and not experienced a failure with the type of root cause given, then i like to state 2. Never choose 1. As FMEA is an continuously and ongoing process, later if you face NC?s that is based on the root cause given, you can raise the value and make sure there is some kind of history log in your FMEA.

And yes FMEA is still there to prevent failures, but for which price?

Some potential and theoretical failure modes based on their root causes can sometimes really be hard to set any value on it. The most common failure is anyway that people spending to much time with discussing the values instead of focusing on keeping it simple in the beginning and listen to feelings.

Also some customer don't understand the FMEA process and bomb you with RPN targets that do not really make sense as RPN is only a helping tool to organize your FMEA. I would rather look at how severity is placed in relationship to occurrence. Also i like to focus to eliminate human error factors first, while i often see that people rather invest $$ in detection.
Top Bottom