LimitLine

Learn · updated 2026-10-28

Design of Experiments: A 2-Level Factorial, Step by Step

By Nicholas Williford, Lean Six Sigma Black Belt · About LimitLine

A designed experiment (DOE) changes several process factors together in a planned pattern, so each factor's effect and their interactions can be estimated from a small number of runs.

Changing one factor at a time takes more runs and can't see two factors working together. A factorial design changes them together in a planned pattern, so every run tells you about every factor at once, and about their interaction.

Runs2 to the power k for k factors at two levels, times replicatesMain effectmean at the high level − mean at the low levelInteractiondoes one factor's effect change with the other's level?Run orderrandomized, so drift can't pose as an effect

Worked Example: Temperature and Time on Yield

Two factors at two levels, each combination run twice in random order: 8 runs (illustrative data).

TemperatureTimeYield, run 1Yield, run 2Mean
LowShort606261.0
HighShort727473.0
LowLong545253.0
HighLong687069.0
TermEffectpVerdict
Temperature14.00.0002Real
Time-6.00.0039Real
Temperature×Time2.00.1161Not shown
506070Time Short, Temperature Low: 61Time Short, Temperature High: 73Time shortTime Long, Temperature Low: 53Time Long, Temperature High: 69Time longTemperature lowhigh
Raising the temperature adds 14 points of yield on average (12 at the short time, 16 at the long time). The lines are nearly parallel: the interaction (2) is small and not significant.

Reading it: run hot (+14) and short (the long time costs 6). The best cell, high temperature and short time, averaged 73. Confirm it with a few runs at that setting before changing the process.

Five Rules for a First DOE

  1. Pick levels far enough apart to see an effect, but inside what is safe to run.
  2. Randomize the run order and record it.
  3. Replicate (or add center points) so there is an error estimate for the p-values.
  4. Check the gauge first: a noisy measurement hides real effects (Gauge R&R).
  5. With 5 or more factors, screen with a fractional factorial first, then study the few that matter.
Design and analyze a DOE in LimitLineMakes a randomized 2-level design, then analyzes it: effects, p-values and plots (a Pro tool; Free users can try it on the example).

Sources

  1. NIST/SEMATECH e-Handbook, §5.3.3.3 Full factorial designs
  2. G. E. P. Box, J. S. Hunter and W. G. Hunter (2005), Statistics for Experimenters, 2nd edition, Wiley

The LimitLine Newsletter

About one email a month: new guides, tool updates and one practical SPC tip. Unsubscribe in one click.

We'll email you once to confirm before anything else is sent. Privacy