Updated July 2026 · 7 min read · Adente Vision Engineering Team
What are PPM and DPMO in inspection?
PPM, parts per million, counts how many defective parts reach the customer for every million shipped. DPMO, defects per million opportunities, counts how many individual defects occur for every million chances to make one, where each part contributes as many opportunities as it has features to get wrong. Both are rate metrics scaled to a million so that small fractions become whole, comparable numbers.
The difference matters when a part carries many inspection points. A wiring harness with 40 crimp points has 40 defect opportunities per part, so a plain per-part PPM understates how the process is really performing across all those points. DPMO normalises for complexity, which is why electronics and harness suppliers often report it. For a simple single-check part, one opportunity per part, PPM and DPMO converge. Six Sigma methodology, which formalises DPMO and the sigma scale, is described in the ISO 13053 standard.
Both are the language of a quality agreement. A supplier does not promise a customer "99% accuracy," it commits to a defect rate expressed in PPM, and inspection exists to hold that number.
Why do OEMs set PPM targets instead of accuracy?
OEMs set PPM targets, not accuracy targets, because accuracy hides the thing they care about. Accuracy is dominated by the huge majority of good parts, so a line that is 99% accurate can still be a poor inspector when defects are rare. What an OEM buys is a ceiling on how many bad parts arrive, and that ceiling is a PPM figure written into the supply contract.
Automotive supply in particular runs on single-digit or low-double-digit PPM expectations for critical characteristics, and the automotive quality-management framework that governs these agreements, IATF 16949, is built around defect-rate control rather than a vague accuracy claim. When your customer speaks in PPM, reporting your inspector's "accuracy" answers a question they did not ask. The useful translation is from your inspection metric into their escape PPM.
How do you convert a false-negative rate into an escape PPM?
A false-negative rate converts to an escape PPM by a direct restatement of the fraction, then a scaling by how many defective parts actually arrive. The false-negative rate is the share of truly defective parts that the inspector passes: on a live cap-inspection line the delivered rate was 0.69%, which restated is 6,900 escapes per million defective parts, because 0.69% is 6,900 per million.
That 6,900 figure is per million defective parts, not per million parts shipped, and the distinction is the whole point. To get the line-level escaped PPM your customer feels, multiply the false-negative rate by the incoming defect fraction. If 1% of parts arrive defective and 0.69% of those escape, the shipped escape rate is 0.0069 multiplied by 0.01, which is 0.0069%, or 69 escaped parts per million shipped. Halve the incoming defect rate and you halve the escaped PPM, which is why an honest escape number always names the defect rate it assumes.
How do sigma levels map to a defect rate?
Sigma levels are a shorthand for a defect rate, so a single number carries the whole quality story. The table below is the standard Six Sigma conversion, using the conventional 1.5 sigma long-term shift, and it is an industry reference, not an Adente measurement. Read it to place any PPM figure on a scale your customer already uses.
| Sigma level | DPMO (long-term, 1.5 sigma shift) | Process yield |
|---|---|---|
| 3 sigma | 66,807 | 93.32% |
| 4 sigma | 6,210 | 99.379% |
| 5 sigma | 233 | 99.977% |
| 6 sigma | 3.4 | 99.99966% |
A useful anchor: the 6,900-per-million escape restatement of a 0.69% false-negative rate sits close to the 4 sigma band at 6,210 DPMO. That comparison is approximate, because a false-negative rate is measured against defective parts while process sigma is measured against total opportunities, but it gives a quality manager an immediate feel for where a 0.69% escape rate lands on the customer's scale.
Why does the base of measurement change the number?
The base of measurement, what you divide by, changes the metric even when the inspector's behaviour is identical. A false-negative rate divides escapes by defective parts. A shipped escape PPM divides escapes by all parts shipped. A process DPMO divides all defects by all opportunities. Same line, three different denominators, three different numbers, and quoting one as if it were another is the most common reporting error in supplier quality.
Get the base right before you report. State whether a PPM is per part shipped or per opportunity, and state whether an escape rate is against defective parts or against total parts. A number without its denominator is not checkable, and an OEM quality engineer will ask for the denominator first. The F1-score and false-negative side of this argument, why a single honest metric beats accuracy, is covered in the sibling post on false-negative rate as an inspection metric.
How should you report inspection performance in the customer's units?
Report in the unit the customer's agreement is written in, then attach the raw inspection metric underneath for traceability. If the contract is in PPM, lead with the escaped PPM and the defect rate it assumes; if the customer tracks DPMO, report per opportunity and state the opportunity count per part. Underneath, keep the inspector's own numbers so an auditor can retrace the conversion.
Adente Vision is an edge-AI visual inspection unit built by ADENTE Advanced Engineering Technologies, part of the Aden Group, sold through automation system integrators, and on a live cap-inspection line it delivered a 99.65% F1-score with a 0.69% false-negative rate, which is the honest basis a supplier converts into a customer-facing PPM. The unit inspects every part rather than a sample, and logs each decision on-device, so the reported PPM rests on 100% inspection with an auditable record rather than an inference from a sample. For the broader method behind the inspection, see the pillar guide on AI visual inspection; for where these checks run in production, see the applications overview.