Optimization-Assisted Quality Function Deployment
Also known as: Optimization-integrated QFD, QFD with optimization, Mathematical programming QFD, OA-QFD
Optimization-assisted QFD extends the classic House of Quality framework by embedding mathematical optimization — linear programming, multi-objective optimization, or metaheuristics — directly into the QFD process. This allows engineers to simultaneously maximize customer satisfaction and minimize cost or resource constraints when setting target values for engineering characteristics, going beyond the largely subjective priority rankings of traditional QFD.
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 optimization-assisted QFD when you have a product or service design problem with multiple competing customer requirements and limited resources (cost, materials, time) that must be explicitly respected. It is most valuable when the number of engineering characteristics is large enough that intuitive target-setting is unreliable, or when design trade-offs need to be documented and auditable. It requires quantified customer importance weights, a completed House of Quality relationship matrix, and at least rough functional relationships between engineering characteristics and customer satisfaction. Do NOT use it as a substitute for gathering genuine voice-of-the-customer data — garbage importance weights produce garbage optimal targets. It is also inappropriate when design constraints are so tight that no feasible solution exists; in that case redesign the product concept before optimizing.
Strengths & limitations
- Converts the largely qualitative prioritization in traditional QFD into an explicit, auditable optimization decision.
- Simultaneously respects customer priorities and engineering or cost constraints, surfacing true trade-offs.
- Easily extended to multi-objective settings where cost and customer satisfaction are both optimized rather than weighted into a single score.
- Provides a sensitivity analysis capability: designers can see which constraints are binding and what relaxing them would be worth.
- Integrates naturally with downstream methods such as response surface methodology or robust design.
- Requires quantified relationships between engineering characteristics and customer satisfaction — these are often difficult to estimate accurately.
- The quality of the optimization result is only as good as the House of Quality relationship matrix, which relies on expert judgment and can be subjective.
- Setting up and solving the optimization model demands mathematical expertise not typical in all engineering teams.
- Linear programming formulations require linear relationships between inputs and outputs, which may not hold in practice.
Frequently asked
How is this different from standard QFD?
Standard QFD uses the House of Quality to rank engineering characteristics by their contribution to customer requirements, then relies on engineering judgment to set target values. Optimization-assisted QFD uses the same HOQ data but feeds the relationship matrix and importance weights into a mathematical optimization model that computes optimal target values subject to explicit constraints — replacing expert guesswork with a rigorous, auditable calculation.
What optimization solver should I use?
If the relationship between customer satisfaction and engineering characteristics is linear, a linear programming solver (e.g., simplex method in Excel Solver, GLPK, or CPLEX) is sufficient. For non-linear relationships use nonlinear programming (e.g., NLPQL, fmincon in MATLAB). When multiple objectives must be traded off simultaneously, multi-objective evolutionary algorithms such as NSGA-II are commonly applied and have extensive published precedent in QFD optimization studies.
Do I need a full House of Quality before applying optimization?
Yes. The HOQ relationship matrix (customer requirements versus engineering characteristics, with importance weights) is the direct input to the optimization model. Without completed HOQ data — including relationship scores and customer importance weights — there is no objective function or constraint structure to optimize. Shortcutting the HOQ step undermines the entire approach.
Can I apply this method when relationships are uncertain or fuzzy?
Yes. A well-developed variant replaces crisp relationship scores with fuzzy numbers (triangular or trapezoidal membership functions) and solves the resulting fuzzy optimization model. This is especially useful in early design stages where expert estimates of relationship strengths carry meaningful uncertainty.
Is multi-objective optimization always better than a weighted single objective?
Not necessarily. A weighted-sum single-objective model is simpler to solve and communicate, and works well when stakeholder preferences over trade-offs are clear and stable. Multi-objective optimization generates a Pareto front of non-dominated solutions, which is more informative when decision-makers are uncertain about trade-off preferences — but it requires additional effort to select a final design from the Pareto set.
Sources
- Akao, Y. (1990). Quality Function Deployment: Integrating Customer Requirements into Product Design. Productivity Press, Cambridge, MA. ISBN: 978-0915299416
- Lai, X., Xie, M., & Tan, K. C. (2004). Optimizing product design using the Kano model and QFD. Engineering Management Conference, 2004 IEEE International, 1085-1089. link ↗
How to cite this page
ScholarGate. (2026, June 3). Optimization-Assisted Quality Function Deployment. ScholarGate. https://scholargate.app/en/experimental-design/optimization-assisted-quality-function-deployment
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
- Quality Function DeploymentExperimental design↔ compare
- Response Surface MethodologyExperimental design↔ compare
- Robust Quality Function DeploymentExperimental design↔ compare