Fractional Factorial Design

What Is Fractional Factorial Design?

Fractional factorial design is an experimental structure that tests a carefully selected subset — typically 1/2, 1/4, or 1/8 — of all possible combinations in a factorial test. It trades the ability to detect higher-order interactions (3-way, 4-way) for dramatically reduced sample size requirements, making multivariate testing feasible on sites that couldn't support a full factorial design.

Also Known As

• Marketing teams rarely use the term — "efficient MVT" or "partial factorial."
• Growth teams say fractional factorial, partial factorial, or screening design.
• Product teams use fractional factorial.
• Engineering teams refer to Taguchi design (a related methodology) or fractional factorial.
• Statisticians call it fractional factorial, Plackett-Burman, or screening design.

How It Works

Testing six factors with two levels each would require 2^6 = 64 combinations in a full factorial. A 2^(6-3) fractional factorial uses only 8 combinations — a 1/8 fraction — chosen to keep main effects independent of each other and of two-way interactions. You get clean estimates of the six main effects and many two-way interactions, at the cost of being unable to separate three-way and higher interactions (which are almost always negligible anyway). Traffic requirement drops from 64 cells to 8 cells — an 8× reduction.

Best Practices

• Consult a statistician or use platform tools for design — choice of confounding structure matters.
• Use fractional factorial for screening — identifying which factors matter most.
• Follow up a screening design with a focused full factorial on the 2–3 factors that emerged as important.
• Pre-register the design and confounding structure before launch.
• Document assumptions about which higher-order interactions are negligible.

Common Mistakes

• Rolling your own fractional factorial design without understanding the confounding structure.
• Interpreting a confounded estimate as "the main effect" when it's actually "main effect + interaction."
• Using fractional factorial when a simple A/B test would answer the real question.

Industry Context

• SaaS/B2B: Rarely practical — traffic and factor count typically make simple A/B tests the better choice.
• Ecommerce/DTC: High-traffic sites use it for screening many merchandising factors quickly.
• Lead gen: Niche — usually more factors than traffic can support, and screening designs help.

The Behavioral Science Connection

Fractional factorial designs embody a pragmatic epistemology: we can't know everything, but we can know enough. By sacrificing ability to detect rare higher-order effects, we gain the ability to detect common main and two-way effects at a fraction of the cost. This is Herbert Simon's satisficing applied to experimental design.

Key Takeaway

Use fractional factorial for efficient screening when you have many factors but limited traffic — and follow up with focused tests on the factors that matter.