Mixture Designs vs. Factorial Designs: Choosing the Right DOE for Formulation Work

When you are developing a new coating, beverage blend, or pharmaceutical formulation, the first question is not which statistical model to use. The first question is: what kind of factors are you actually working with? That single distinction drives everything else in your design of experiments strategy.

This article walks you through a clear selection framework for choosing between mixture designs and factorial designs in formulation work. You will learn the core decision rule, see how it applies to real formulation scenarios, and understand how tools like Scheffé polynomials and simplex-lattice designs fit into the picture without getting lost in theory.

Key Takeaways

• Mixture designs are for factors that must sum to 100%; factorial designs are for independent variables.
• Mixture designs use a simplex space and Scheffé polynomials (no intercept term).
• Simplex-lattice and simplex-centroid designs generate the blend points for mixture experiments.
• Factorial designs suit screening, interaction detection, and process optimization.
• Combined mixture-process designs optimize blends and process settings together in one experiment.

The Core Decision Rule for Mixture Designs and DOE Selection

The decision between mixture designs and factorial designs comes down to one question: are your factors constrained to sum to a fixed total? If yes, you need a mixture design. If your factors are independent settings — like temperature, pressure, or machine speed — a factorial design is the right tool for the job.

This is not just a technical preference. Using a standard factorial design when your factors are component proportions will produce a flawed experimental region. The simplex constraint means that increasing one ingredient automatically decreases another, so you cannot vary them independently the way factorial designs assume.

Consider a sports drink formulation with four ingredients: water, sugar, citric acid, and electrolyte concentrate. Each ingredient is expressed as a proportion of the total, and all four must sum to 1. You cannot set sugar to 40% and water to 70% at the same time. That is the defining characteristic of a mixture problem, and it demands a mixture design framework built around response surface methodology on a constrained simplex space.

With that framework in place, it becomes easier to understand why each design type exists and how to apply them correctly in formulation chemistry.

How Mixture Designs Work in Formulation Chemistry

Mixture designs are a specialized form of response surface methodology built specifically for situations where the factors are ingredient proportions. The design space is not a cube — it is a simplex, which is a triangular or higher-dimensional shape defined by the constraint that all components must be non-negative and sum to one. This creates a fundamentally different geometry that standard factorial methods cannot handle correctly.

The most common structural tools in mixture DOE are simplex-lattice designs and simplex-centroid designs. A simplex-lattice design for q ingredients at degree m places experimental points at all combinations where each component proportion belongs to the equally spaced set . This gives balanced support across the simplex for fitting Scheffé models of the appropriate order.

You might be wondering what makes the Scheffé polynomial different from a standard regression model. The key difference is that Scheffé polynomials drop the intercept term entirely. Because the component proportions always sum to 1, a standard intercept would be redundant and create collinearity. The model uses only the proportions themselves and their cross-product terms to capture blending effects across the formulation space.

Applying Mixture Designs to a Coating Formulation Example

Suppose a manufacturer is developing a protective industrial coating with three components: epoxy resin, hardener, and thinner. The goal is to find the blend that produces the best combination of adhesion strength and drying time. Each experimental run is a specific combination of the three proportions, and all three must sum to 100%.

A simplex-lattice design — three components at degree 2 — generates exactly six design points: the three pure-component vertices and the three binary-blend edge midpoints, calculated using the formula N = (p+m−1)!/(m!(p−1)!). If deeper coverage of the interior blend space is needed, the design can be augmented with a centroid point or axial check-blends, effectively turning it into a simplex-centroid or augmented mixture design. The resulting Scheffé quadratic model would capture both the individual component effects and the pairwise blending interactions that often drive formulation performance.

When Factorial Designs Are the Right DOE Choice

Factorial designs are the standard tool when your experimental factors are independent process variables. Temperature, mixing speed, cure time, and pH are examples of factors that do not share a summation constraint. You can set each one independently, which means the design space is a standard hypercube rather than a simplex.

Two-level full factorial designs allow you to test all combinations of high and low settings for each factor, making them powerful for identifying main effects and interactions. When the number of factors is large, fractional factorial designs reduce the number of runs by assuming that higher-order interactions are negligible, which is often a reasonable assumption during screening phases.

In formulation work, factorial designs typically apply to the process side of the problem — not the ingredient proportions. For example, once a coating formula is established through a mixture design, a factorial design might then optimize the application temperature and spray pressure used to apply that coating.

Key Situations Where Factorial Designs Apply

• Process screening: Identifying which process variables most affect a response before running a detailed optimization study.
• Independent factor variation: When each factor can be set without affecting the others, a factorial grid is appropriate and efficient.
• Interaction detection: Two-level factorial designs efficiently estimate two-factor interactions across a defined factor space.
• Pre-formulation studies: Testing equipment settings, environmental conditions, or raw material sources before locking in a blend design.
• Post-formulation process optimization: After a mixture design identifies the best blend, factorial designs can tune the manufacturing process around it.

Understanding where factorial designs end and mixture designs begin is one of the more practical skills a formulation engineer or quality professional can develop. Getting this wrong does not just affect your model — it can invalidate your entire experimental program.

Combining Mixture Designs and Factorial Designs in DOE

Modern formulation DOE rarely stops at just one design type. In many real-world applications, both the ingredient proportions and the process conditions matter, and they may interact with each other. A coating that performs well at one blend ratio might behave differently depending on the cure temperature used during application.

This is where combined mixture-process designs come in. These designs cross a mixture design — covering the simplex of ingredient proportions — with a factorial design covering the process variables. The result is a crossed or split-plot structure that allows simultaneous optimization of both the formulation and the processing conditions through a unified response surface methodology framework.

A published example of this approach appears in beverage and food science research, where ingredient blends are optimized alongside process parameters like mixing temperature and homogenization pressure. The combined design captures both the blending surface and the process response surface in a single, coordinated experiment, which is far more efficient than running two separate studies sequentially.

Structure of a Combined Mixture-Process Design

• Mixture part: Covers the simplex space of ingredient proportions using a simplex-lattice or optimal Scheffé-based design.
• Process part: Covers the factorial space of independent process variables at two or more levels.
• Crossed structure: Each mixture design point is run at each combination of process factor settings, or a subset of those combinations in a fractional arrangement.
• Model form: The combined model includes Scheffé mixture terms, factorial process terms, and mixture-by-process interaction terms.
• Optimization: A single response surface or desirability function is used to find the best combination of blend and process settings simultaneously.

Air Academy Associates has helped professionals across manufacturing, aerospace, and healthcare apply exactly this kind of combined DOE thinking to real formulation and process challenges. With more than 250,000 graduates trained over 30 years, the team brings a practical, application-focused approach that moves beyond theory into results you can measure. Explore the Introduction to Mixture Designs Short Course to see how this framework is taught in a structured, hands-on format.

Build Your DOE Skills With These Air Academy Associates Courses

Choosing the right design type is only the first step. Executing it correctly — selecting design points, fitting the right model, interpreting the response surface — requires structured training with real-world application. Air Academy Associates offers a focused set of courses that directly support formulation DOE work, from foundational concepts to advanced mixture and factorial methods.

Whether you are new to design of experiments or looking to sharpen a specific skill, these courses are built around the KISS (Keep It Simple Statistically) approach, ensuring that every concept connects directly to practical application in your industry.

1. Introduction to Mixture Designs Short Course

Introduction to Mixture Designs Short Course

This course is the most direct resource for formulation professionals working with component proportions. It covers:

• The simplex design space and why standard factorial methods do not apply
• Simplex-lattice and simplex-centroid design structures
• Scheffé polynomial model fitting and interpretation
• Response surface mapping across the formulation simplex
• Practical examples from formulation chemistry and blending applications

This course gives formulation engineers the tools to design, run, and analyze mixture experiments with confidence — without requiring advanced statistical theory as a prerequisite.

2. 2-Level Designs: Full Factorial Short Course

2-Level Designs: Full Factorial Short Course

When your formulation work involves independent process variables alongside ingredient proportions, full factorial designs become essential. This course covers two-level full factorial structures, main effects, interaction detection, and how to interpret factorial results in a process optimization context. It is especially useful for teams running combined mixture-process studies who need a solid grasp of the factorial component before integrating it with a mixture design framework.

3. 2-Level Designs: Fractional Factorial Screening

2-Level Designs: Fractional Factorial Screening

In early-stage formulation development, screening out non-influential process variables saves significant time and resources. This course teaches fractional factorial screening designs, which reduce experimental runs while still identifying the most important factors. For formulation teams working with many candidate process variables, fractional factorial screening is often the logical step before committing to a full response surface or combined mixture-process study.

4. Introduction to Design of Experiments

Introduction to Design of Experiments

For professionals who are new to DOE or need a solid foundational refresher, this course builds the conceptual and practical base needed to work with both factorial and mixture designs. It covers:

• Core DOE principles and terminology
• How to plan and structure an experiment
• Selecting the right design type for your situation
• Analyzing results and drawing actionable conclusions

This is the ideal starting point before moving into specialized courses on mixture designs or advanced response surface methodology.

Conclusion

Selecting between mixture designs and factorial designs is a foundational decision that shapes every step of your formulation DOE work. When components must sum to a fixed total, mixture designs on a simplex space — supported by Scheffé polynomial models — are the correct approach. When factors are independent process settings, factorial designs apply. Air Academy Associates offers structured, practical training to help you make this distinction confidently and apply the right design type to your specific formulation challenge — contact our team or explore the courses above to get started.

Air Academy Associates offers expert-led Design of Experiments training to help you master formulation challenges with confidence. Our Master Black Belt instructors bring decades of hands-on DOE experience across industries. Get started with the right course for your team today.

FAQs

What Is a Mixture Design?

A mixture design is a Design of Experiments (DOE) approach used when the inputs are proportions of ingredients in a formulation that must sum to 100%. It helps you understand how changing the blend (not the absolute amounts) affects performance, and is a core tool we teach and apply in formulation-focused DOE.

When Should You Use a Mixture Design?

Use a mixture design when your factors are recipe components (e.g., resin, solvent, filler) and increasing one ingredient necessarily decreases one or more others. It's ideal for optimizing formulations, meeting constraints (cost, viscosity, regulatory limits), and finding robust blends without wasting runs on impossible combinations.

What Are the Types of Mixture Designs?

Common types include simplex-lattice and simplex-centroid designs for exploring blend space, extreme vertices designs for constrained formulations, and mixture-process (combined) designs when both ingredient proportions and process settings (e.g., temperature, mixing time) matter.

How Do You Analyze Mixture Design Experiments?

Mixture DOE is typically analyzed with mixture-specific regression models (e.g., Scheffé polynomials), followed by ANOVA, diagnostic checks, and response optimization using contour and ternary plots. In our training and consulting, we emphasize practical interpretation—linking model terms to ingredient interactions and confirming the optimum with validation runs.

What Is the Difference Between Mixture Design and Factorial Design?

Factorial designs vary independent factors (e.g., temperature, pressure, time) and estimate main effects and interactions across levels. Mixture designs vary ingredient proportions that are dependent because they must sum to 100%, so the modeling and design space are fundamentally different; choosing the right one prevents misleading conclusions in formulation work.

Related Articles:

• Dirty Data in the Measure Phase: Cleaning Datasets Before Analysis
• When Not to Use Six Sigma: Recognizing Problems That Need Intuition
• Stacking Certifications: Six Sigma Plus PMP, Agile, or Data Analytics — Worth It?