Multi-response Fractional Factorial Design — Simultaneous Optimization of Multiple Outputs
Multi-response Fractional Factorial Design of Experiments · Also known as: MRFFD, multi-response FFD, multi-objective fractional factorial design, simultaneous multi-response fractional factorial
Multi-response fractional factorial design (MRFFD) applies a resolution-efficient fractional factorial experiment to study multiple response variables simultaneously. By running only a carefully chosen fraction of the full factorial treatment combinations, the experimenter gathers enough information to fit individual response models for each output and then optimize all responses jointly — typically via a composite desirability function — while keeping the number of experimental runs tractable.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use multi-response fractional factorial design when you have two or more response variables that must be controlled or optimized simultaneously, four or more controllable factors, and experimental costs or time that make a full factorial impractical. It is particularly suited to early-stage product or process development where the goal is to identify the dominant factor effects across all responses efficiently. Do not use it when strong curvature (quadratic effects) is expected over the factor range — in that case, move directly to a central composite or Box-Behnken design. Also avoid it when response correlations are so strong that a single combined index (e.g., a physical constraint) more naturally captures the trade-off, or when any individual response model has poor fit, since the desirability aggregation will then be meaningless.
Strengths & limitations
- Drastically reduces run count compared to a full factorial while retaining the ability to estimate main effects and selected interactions for each response.
- The desirability-function framework provides a transparent, tunable way to encode different priorities among competing responses.
- Compatible with standard DoE software — design selection, model fitting, and desirability optimization are all supported in Minitab, JMP, and the R 'desirability' package.
- Scales readily as the number of responses grows: adding another response variable requires no additional experimental runs, only an additional model-fitting and desirability-mapping step.
- Produces interpretable factor-effect estimates for each response separately, so trade-off mechanisms can be understood, not just optimized blindly.
- Fractional factorial designs cannot estimate all two-factor interactions at lower resolutions; aliased effects that happen to be non-negligible will bias the response models and distort the optimization.
- The desirability approach compresses information: a D value close to 1 may mask one response that is barely acceptable while others are excellent.
- The polynomial response models assume linearity or mild curvature; if true response surfaces are strongly curved, the fitted models and the resulting optimum will be unreliable.
- Choosing individual desirability function shapes (target values, weight parameters) requires subject-matter judgment that can introduce subjectivity into what appears to be a quantitative procedure.
Frequently asked
How do I choose the resolution for a multi-response fractional factorial design?
For multi-response optimization, resolution IV is generally the practical minimum: it ensures that main effects are not aliased with other main effects or two-factor interactions, though some two-factor interactions are confounded with each other. Resolution V is preferable when resources allow, as it permits clean estimation of all two-factor interactions — information that is often needed to understand how factors interact in driving each response toward its target.
What if the responses conflict — can I still find an optimum?
Conflicting responses are the rule rather than the exception. The desirability approach handles this by letting you assign weights and target ranges to each response, producing a trade-off surface. The global desirability maximum represents the best achievable compromise given those priorities. You can also plot Pareto fronts across pairs of responses to visualise trade-offs explicitly and then choose a factor setting that reflects the true engineering priority.
When should I switch to a response surface design instead?
Switch to a central composite or Box-Behnken design when you have already identified the key factors through screening (often via a resolution III or IV fractional factorial) and now need to model curvature — quadratic effects — to accurately locate an optimum in a narrower region. Multi-response fractional factorial design is best for the screening and early characterisation stage; response surface methods are best for fine-tuning near the optimum.
Can I add more responses after the experiment is run?
Yes, as long as all response values were measured (or can be derived) from the original experimental runs, fitting additional response models and including them in the desirability aggregation requires no new experiments. This is one practical strength of the framework: a response the team initially considered secondary can be incorporated into the optimization at the analysis stage.
Is software required, or can this be done by hand?
Design selection and model fitting for simple cases (e.g., 2^(4-1) with two responses) can be done manually using standard tables and multiple regression. However, desirability optimization over the fitted surfaces is tedious by hand when more than two factors are involved. Software such as Minitab, JMP, or R (packages 'FrF2' and 'desirability') is strongly recommended for reliable, reproducible results.
Sources
- Derringer, G., & Suich, R. (1980). Simultaneous optimization of several response variables. Journal of Quality Technology, 12(4), 214–219. DOI: 10.1080/00224065.1980.11980968 ↗
- Montgomery, D. C. (2017). Design and Analysis of Experiments (9th ed.). Wiley. ISBN: 978-1119492443
How to cite this page
ScholarGate. (2026, June 3). Multi-response Fractional Factorial Design of Experiments. ScholarGate. https://scholargate.app/en/experimental-design/multi-response-fractional-factorial-design
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Design of experimentsExperimental design↔ compare
- Multi-response Response Surface MethodologyExperimental design↔ compare
- Multi-response Taguchi methodExperimental design↔ compare
- Response Surface MethodologyExperimental design↔ compare