Engineering Reference Library
Menu
Eng Calculators

Reference

Estimating Proportion And Nonconformance Rate

Estimate a proportion, defect rate, or nonconformance rate from binary observations and understand the assumptions behind the result.

Referenceengineering & statisticsPublishedLast reviewed: 2026-05-16

Use proportion methods when each observation has two outcomes, such as conforming or nonconforming, pass or fail, leak or no leak, defective or not defective.

What This Means

The observed proportion is the count of outcomes of interest divided by the total number inspected or tested. If 4 of 80 units are nonconforming, the observed nonconformance rate is 4 / 80 = 0.05, or 5%.

The observed rate is a point estimate. A confidence interval communicates uncertainty around that estimate, and a sample-size calculation plans how many binary observations are needed for a target margin of error.

Key Formula

phat = x / n
  • phat is the observed proportion.
  • x is the count of nonconforming, defective, or selected outcomes.
  • n is the total number of binary observations.

Use This When

  • Estimating a defect rate from inspection results.
  • Estimating a pass/fail proportion from validation testing.
  • Planning a binary test around a target margin of error.
  • Reporting uncertainty around a nonconformance rate.

Assumptions

  • Each observation has a binary outcome.
  • The count of selected outcomes is clear and consistently classified.
  • Observations are independent enough for a binomial model.
  • The sample represents the process, lot, supplier, design, or operating condition being discussed.

Limitations

  • Rare-event rates can need large samples for useful precision.
  • The simple observed proportion can be a poor uncertainty summary near 0 or 1.
  • Acceptance sampling plans, reliability demonstration, and standards-based quality decisions can require different methods.
  • Changing inspection definitions during data collection invalidates simple interpretation.

Common Mistakes

  • Reporting only x / n without sample size.
  • Treating zero observed defects as proof that the true defect rate is zero.
  • Mixing data from different processes or lots without noting the change.
  • Using continuous-measurement sample-size formulas for pass/fail data.
  • Confusing percentage points with percent relative change.

Sources

This reference is based on the NIST/SEMATECH Engineering Statistics Handbook for proportion defective and binomial confidence-interval context, with Penn State statistics material for proportion sample-size planning and Agresti-Coull for interval-method cautions around simple proportion intervals.