PPM behaves badly when the base is small. With 4,000 parts shipped, one defective part is 250 PPM and zero defective parts is 0 PPM. The process did not change between those two outcomes; you simply saw one event or you did not. Treating a 250 PPM month as a crisis and a 0 PPM month as proof of a fix is reacting to noise.
A practical rule of thumb: a PPM figure is only stable enough to judge against a target when the base is large enough that you would expect at least five to ten defects at the target rate. At a 50 PPM target that means at least 100,000 to 200,000 parts. Below that, report the raw counts ("1 defect in 4,000 shipped") rather than the PPM, and roll several months together before drawing conclusions.
The same caution applies to a single large return. If a customer rejects an entire lot of 600 pieces because of one mixed-up part, your month shows 600 defects. Separate lot-based rejections from piece-based ones in your records so the trend chart does not get distorted by counting rules.
Standard Procedure
- 1Before reporting a PPM value, check how many parts are in the base.
- 2Work out how many defects you would expect at the target rate (target PPM × base ÷ 1,000,000).
- 3If that expected number is below about five, report raw counts and accumulate more months.
- 4Record lot-based rejections separately from piece-based rejections.
- 5Still investigate every individual defect; small samples excuse the statistics, not the problem.
Lesson Summary
- ✓One defect in a small shipment produces a scary PPM that means very little.
- ✓Accumulate volume before judging against the target, but investigate each defect regardless.