Design of experiments (DOE)

Design of experiments is a systematic method for planning experiments that vary multiple factors, inputs believed to affect an outcome, deliberately and simultaneously, rather than changing one variable at a time, so each factor's individual and combined effect can be estimated with a known level of statistical confidence.

A design specifies which combinations of factor levels to test and in what order, often using structured patterns such as factorial or fractional factorial layouts, that let analysts separate a factor's true effect from noise using far fewer runs than testing every possible combination would require. This differs from a simple A/B testing setup, which usually varies one factor between two groups, DOE is built to handle several factors and their interactions at once, in a single, efficient experiment.

DOE is widely used in manufacturing, chemistry, and product R&D to optimize a process or formulation, and its logic of randomization and controlled comparison underlies proper randomized controlled trial design in other fields. Pitfalls include skipping randomization or blocking, which lets other differences masquerade as a factor's effect, and misreading an interaction effect, where two factors matter together in ways neither does alone, a pattern a poorly chosen sample size may not have the power to detect reliably.