Engineering Reference Library
Menu
Eng Calculators

Reference

Practical Sample Size Planning

Choose between mean and proportion sample-size methods for engineering measurements, inspections, and test planning.

Referenceengineering & statisticsPublishedLast reviewed: 2026-05-16

Use sample-size planning before collecting data so the test has a defensible connection between effort, confidence, variability, and desired precision.

What This Means

The first decision is the response type. Continuous measurements, such as diameter, torque, pressure, or time, usually call for a mean-based planning method. Binary outcomes, such as pass/fail or conforming/nonconforming, call for a proportion method.

After choosing the response type, define the confidence level and the margin of error in engineering terms. Then choose variability assumptions from process history, prior tests, pilot data, or a conservative planning choice.

Key Relationships

mean planning: n >= (z sigma / E)^2
proportion planning: n >= z^2 p(1 - p) / E^2
  • Use the mean form for continuous responses.
  • Use the proportion form for binary outcomes.
  • Round the final sample size up.

Use This When

  • Planning a measurement study around an average value.
  • Planning an inspection study around a defect or nonconformance rate.
  • Comparing the cost of tighter margin of error against more samples.
  • Translating a vague request for "enough samples" into explicit assumptions.

Assumptions

  • Observations are independent enough for the selected method.
  • The planned sample represents the process or population of interest.
  • The margin of error is chosen in practical engineering units or percentage points.
  • Prior standard deviation or planning proportion is documented.

Limitations

  • These methods estimate a mean or proportion within a target half-width.
  • Detecting a specified shift with statistical power needs a power-based planning method.
  • Acceptance sampling, reliability demonstration, capability studies, and destructive test plans may need specialized methods.
  • Sampling cannot fix biased collection, unstable processes, or unclear acceptance criteria.

Common Mistakes

  • Using the continuous-mean formula for pass/fail inspection data.
  • Choosing a sample size before defining margin of error.
  • Guessing standard deviation without checking prior data or running a pilot study.
  • Treating p = 0.5 as an estimate instead of a conservative planning assumption.
  • Ignoring independence when repeated observations come from the same part, batch, operator, or fixture.

Sources

This reference is based on the NIST/SEMATECH Engineering Statistics Handbook for engineering sample-size context, with Penn State statistics references used as corroboration for the mean and proportion margin-of-error planning forms.