Leaderboard

It is great to see detailed answers for a question on a tool deemed as complex by many Six Sigma experts. The question sought answers to what apart from subjectivity in the rating scale is the limitation with using RPN. Most answers bring out one key point that RPN does not draw your attention towards the severity number. This is explained very well in several answers - hence making it complex to choose the best answer. Breaking the tie is the "suggested treatment" to this limitation which is provided by Vishwadeepak Choudhary, hence chosen as the best answer. I do recommend everyone reading all the answers to get a well rounded understanding.

A deep intrinsic problem with FMEA is how we calculate RPN (Risk priority number) by performing a mathematical operation on three ordinal scale data. Severity, occurrence and detection are purely ranked numbers and we never get to see the absolute difference between two ranks so any mathematical operation like addition, subtraction or multiplication don’t hold true however they can definitely throw a mathematical number. We calculate RPN in the similar fashion and then use this number to prioritize risks. Moreover, three building blocks of RPN are not on the same scale. They have different priorities in different organization. Severity should definitely be considered of top most importance. Let’s look at a scenario. We will try to calculation RPN for two earthquakes with different magnitudes. One at Richter scale of 2.0 and another at 6.0. 1. Richter scale 2.0 earthquake: Severity = 2 (as per Richter scale reading) Occurrence = 5 (assuming that this occurs very often) Detection = 4 (we would use same detection for both scenarios) RPN = 2 * 5 * 4 = 40 2. Richter scale 6.0 earthquake: Severity = 6 (as per Richter scale reading) Occurrence = 1 (very less frequent) Detection = 4 (we would use same detection for both scenarios) RPN = 6* 1 * 4 = 24 If we simply go by prioritizing risks as per RPN, then the first risk would get prioritized however practically that’s a lot safer than Risk 2. Richter scale 6.0 earthquake is rare but if it occurs for once, it’s a disaster. The RPN calculation doesn’t take care of such individuality which makes a great sense in practical scenarios. One way to overcome above problem could be to use weighted count method for calculating RPN. Severity should get highest weightage (may be 3), followed by Occurrence (may be 2) and then Detection (may be 1). Let’s redo the above earthquake scenario and we would call our metric as Weighted Ordinal RPN (WO-RPN). 1. Richter scale 2.0 earthquake: Severity = 2 (as per Richter scale reading) – we would consider it as count 2 and multiply it by weightage 3: gives the value of 6 Occurrence = 5 (assumption that this occurs very often) – Weightage 2, so count gives 2*5 = 10 Detection = 4 (we would use same detection for both scenarios) – Weightage 1, so count gives 1*4 = 4 WO-RPN = 6+10+4 = 20 2. Richter scale 6.0 earthquake: Severity = 6 (as per Richter scale reading) – Weightage 3, count = 3*6 = 18 Occurrence = 1 (very less frequent) – Weightage 2, count = 1*2 = 2 Detection = 4 (we would use same detection for both scenarios), Weightage 1, count = 4*1 = 4 WO-RPN = 18+2+4 = 24 And this weighted ordinal RPN brings second risk as top priority which is the cause of the concern. I welcome your thoughts on this subject further.