InsightFab
Knowledge Base/PFMEA Severity, Occurrence, Detection Scoring Standards
Quality Assurance6 min read

PFMEA Severity, Occurrence, Detection Scoring Standards

Drawing from a personal experience where a CPK report of 1.08 caused significant concern, this article highlights common mistakes in PFMEA, particularly the intuitive scoring of Severity, Occurrence, and Detection. It emphasizes that effective PFMEA demands thorough risk assessment beyond mere form-filling to prevent major issues from seemingly minor defects.

The day the CPK report came out, the room fell silent for three seconds, and the boss's face turned green.

I remember once, we switched to a new supplier for consumables on our machine. We thought it was just a small screw, so there shouldn't be any major problems. What happened next? Within two days, the line started reporting errors, and a batch of goods was completely ruined. In the meeting room, Quality Assurance tossed out the CPK report: 1.08. Do you know that feeling? The entire room instantly fell silent, not even a breath could be heard. The boss's face went from red to green. Everyone looked at each other, and there was just one question: How was this PFMEA originally conducted? How could such a critical failure mode have been missed?

Where did the problem lie? It's not that you didn't do it; it's that you didn't 'think it through'.

To be frank, everyone knows how to conduct PFMEA (Process Failure Mode and Effects Analysis). SOPs are thick, and the forms are filled out completely. But do you know? Many times, we're just 'filling out forms' instead of genuinely 'thinking about risks.' Especially for the three scores: Severity, Occurrence, and Detection, many people simply assign scores based on intuition and impression. The result is that what should be high isn't, and what should be low isn't, leading to truly problematic areas being glossed over. That screw with a CPK of 1.08 that day is a vivid example; during the initial assessment, everyone thought, 'What big deal could a small thing like a screw cause?'

How to actually do it? Numbers speak for themselves.

Honestly, these three scores really cannot be evaluated with 'approximately' or 'seemingly.' I'll give you a few practical scoring principles:

  1. Severity: How bad will it be if this happens?

* This point is most easily underestimated. Don't just look at 'direct' impacts; consider 'indirect' impacts.

* For example, a small burr might only directly affect appearance. But what if it jams the mechanism, causing the machine to stop for two hours? What's the downtime cost? Or even worse, the customer directly returns the goods, or even claims compensation?

* When scoring Severity, don't be afraid to give high scores. If it leads to customer returns, complaints, or production line stoppage for more than an hour, directly assign a 9 or 10, without hesitation. These factors directly impact company revenue and reputation.

  1. Occurrence: How often does this happen?

* This absolutely requires data, not intuition. Do you track DPMO (Defects Per Million Opportunities) in your factory? Do you have process yield data?

* For example, if a defect's DPMO is 6210 (roughly equivalent to a yield of 99.379%), this means there are 6210 defects per 1 million products. Can this occurrence rate be considered low? Absolutely not!

* I would suggest directly mapping DPMO or defect rate to the Occurrence scoring standard. For instance, DPMO > 5000 would be a 9 or 10, and DPMO 500-5000 would be a 7 or 8. Use concrete data for evaluation to avoid distortion.

  1. Detection: If it happens, what's my chance of catching it?

* This point is often overly optimistic. Many people will say, 'We have AOI!' or 'We have QA sampling inspection!'

* But the question is, can your AOI truly catch 100%? What's its false alarm rate? How frequent is your QA sampling inspection? Can it only detect large batches of defects?

* To put it plainly, if detection relies on 'human eyes,' the detection rate won't be high. Humans get fatigued and overlook things. If there's no automation, poka-yoke (mistake-proofing), 100% inspection, or if instrument calibration frequency isn't high enough, then the detection score should be high (meaning difficult to detect).

* In other words, the harder it is for you to find the problem, the higher the detection score. If you only have post-production QA sampling inspection, the score should definitely be 7 or above.

The most common pitfall: Treating 'possible' as 'impossible'.

The most absurd thing I've encountered is when people conducting PFMEA treat risks that are 'possible but low probability' as 'impossible.' Take the screw case: during the initial assessment, someone said, 'The screw defect rate is very low, and we have QC sampling inspection, so it should be fine.' What was the result? If they had considered the risk of a supplier change, combined with potential latent defects inherent in the screw itself, and elevated both the Occurrence and Severity scores by one level, then the R.P.N. (Risk Priority Number) would not have been a number low enough to be ignored.

One thing you can do today.

Re-examine the PFMEA you have on hand, especially those three scores, and ask yourself: 'Is this truly what the data says?'

Want to try it yourself?

Every tool mentioned in this article is available on InsightFab — just upload a CSV to analyze.

Go to Tools

More in this category

The Significance of First Article Inspection (FAI) and Last Piece Confirmation

This article illustrates a common real-world scenario in semiconductor manufacturing where a failed First Article Inspection (FAI) can cause an entire production line to halt, leading to significant disruptions and financial losses. It highlights FAI's critical role as a product's 'identity card,' ensuring initial production samples meet all design specifications and preventing costly issues in subsequent processes.

The Difference Between IATF 16949 and ISO 9001: Additional Requirements for the Automotive Industry

This article provides practical insights for those interested in quality management and the rigorous demands of the automotive sector. The author recounts an eye-opening experience of product rejection due to a CPK of 1.08, underscoring why ISO 9001 is insufficient and why the industry has transitioned to the more stringent IATF 16949, offering a deeper understanding of its strict requirements.

Customer Complaint Handling SOP: A Standard Process from Reception to Closure

This practical article opens with the author's personal 'Cpk 1.08' disaster, underscoring that customer complaints always arrive suddenly and often when least expected. Likening complaint handling to a 'firefighting SOP,' it emphasizes immediate resolution over theoretical analysis, providing readers with a calm, structured approach to avoid common pitfalls.