Defective parts per million is the number most likely to appear in a contract without anyone having checked whether the test plan can actually deliver it.
It is worth agreeing early, because the cost of the last few hundred DPPM is much higher than the cost of the first few thousand.
What DPPM Actually Measures
DPPM counts defective parts that escape to the customer, per million shipped. It is not the same as yield, and it is not the same as fallout — a part that fails in test and is scrapped never becomes a DPPM.
This is why a program can have poor yield and excellent DPPM, or good yield and terrible DPPM. Yield is about how much you scrap; DPPM is about how much you miss.
Typical Ranges by Market
Targets vary by consequence of failure. Safety-critical and automotive programs commonly work in single-digit to low-tens of DPPM. Commercial and industrial programs are often in the tens to low hundreds. Consumer programs frequently accept on the order of one to two hundred DPPM.
These are starting points for discussion, not standards. The binding number is whatever the customer contract says, and it should be agreed before the test program is written.
- Automotive and safety: single-digit to low-tens of DPPM is common
- Commercial and industrial: tens to low hundreds
- Consumer: on the order of one to two hundred DPPM is often accepted
Why Tighter Is Expensive
Each increment of escape reduction costs test time, additional insertions, tighter guard bands, or extra screens — and tighter guard bands also reject good parts, which costs yield.
There is a crossover point where the cost of the additional screen exceeds the cost of the failures it prevents. Finding it is the actual work.
Negotiating the Target
The useful questions are: what happens if a part fails in the field, who bears that cost, and is the failure detectable at goods-in or only after assembly? A part that fails visibly at board test is a very different proposition from one that fails in a customer's product.
Agree the measurement method too. DPPM measured at goods-in differs from DPPM measured after customer assembly, and a contract that does not say which one it means will be argued about later.
How a Target Becomes Test Time
A DPPM number is not bought directly — it is bought through the things that produce it: fault coverage on the pattern set, guard-band width, outlier limits, and any extra insertion added to catch what ATE misses. Each one has a price in test seconds, in scrapped good parts, or in both.
The chain is worth writing down before the target is agreed, because it is the only way to answer the question a finance team will ask: what does the last 20 DPPM cost per part?
- Fault coverage: stuck-at and transition coverage on the production pattern set, and the fault model the coverage number was computed against.
- Guard bands: tighter limits catch marginal parts and also reject good ones, so the yield loss has to be counted alongside the escape reduction.
- Outlier and part-average testing: statistical limits catch parts that pass every datasheet limit but sit far from the population — the usual source of early-life returns.
- Extra insertions: burn-in and system-level test catch failure modes ATE does not see, at their own cost per part.
- Sample versus 100%: some screens work as lot-level sampling; escapes then become a statistical statement, not a per-part guarantee.
Writing the Target Into the Test Plan
A DPPM target that lives only in the contract is unenforceable. It has to appear in the test plan as a measurable, reviewable item with an owner — otherwise the first argument about it happens after the first field failure.
- State which insertion the number is measured at — goods-in, after assembly, or in the field.
- State the coverage metric and the fault model, so the number can be re-derived when patterns change.
- Fix the review cadence: bin pareto at ramp, then periodically once volume is stable.
- Close the loop with RMAs — field returns are the only measurement of escapes that does not depend on your own test.
- Say what happens on excursion: containment, re-screen, or ship-under-concession, and who decides.
