Preface

“The height of sophistication is simplicity.”

Clare Boothe Luce, 1931

Overview

In this preface, we provide an introduction to our book that includes our experiences in formulation development. Guidance is also provided on what you will learn and important success factors to be aware of and applied.

At all times we are focused on simplicity: simplicity in experimental design, data analysis, interpretation and communication of results. By focusing constantly on simplicity, we have found that formulations are developed faster and their characteristics are easier to understand and communicate to others.

Many products are formulations in which various ingredients or components are blended (mixed) together and processed to produce the final product. Some examples are shown in Table 1 (adapted from Smith 2005). In understanding formulations and how they arise, it is helpful to see the various industries that create and manufacture formulations. Some examples of such industries are summarized in Table 2.

Table 1 - Products Created by Blending Two or More Ingredients or Components

Adhesives	Dyes	Lubricants	Rocket Propellants
Aluminum	Fiber Finishes	Metal Alloys	Rubber
Animal Feed	Floor Coverings	Paints	Sealants
Artificial Sweeteners	Floor Finishes	Paper Coatings	Soaps
Beverages	Foams	Personal Care Products	Steel
Biological Solutions	Food Ingredients	Pesticides	Surfactants
Cement	Froth Flotation Reagents	Petroleum Products	Synthetic Fibers
Ceramic Glazes	Gasket Materials	Pharmaceuticals	Tobacco Blends
Ceramics	Gasoline	Photoconductors	UV Curable Coatings
Cleaning Agents	Glasses	Photoresists	Water Treatment Chemicals
Cloth Fiber Blends	Hair Spray	Polymer Additives	Window Glass Coatings
Cocktails	Herbicides	Polymers	Wine
Combination Vaccines	Hydrogels	Powder Coatings
Construction Materials	Inks	Protective Coatings
Cosmetics	Insecticides	Railroad Flares

Table 2 - Industries that are Major Producers of Formulations

Biotech	Metals
Ceramics	Paint
Chemicals	Petroleum
Coatings	Pharmaceuticals
Electronics	Plastics
Food	Textiles

The authors first met at while working at the DuPont Company’s engineering center in Newark, DE. DuPont’s original product, gunpowder, is a mixture of three components: potassium nitrate (saltpeter), sulfur, and charcoal.

Figure 1 – Ingredients of Black Powder

© Jeff J. Daly, FUNDAMENTAL PHOTOGRAPHS, NEW YORK

The quality of the powder was a function of the proportions of the components in the mixture, NOT the total amount of the mixture (Figure 1). The typical formulation consisted of:

75% Saltpeter, 12.5% Charcoal and 12.5% Sulfur. Other formulations used for specific applications were:

Use	Saltpeter	Charcoal	Sulfur
General	75	12.5	12.5
Hunting	78	12	10
Military	75	15	10
Blasting	62	18	20

We note that the blasting formulation was considerably different from the other formulations. Wilkson (1966, page 23) noted that “Manufacturers would often experiment, changing their formulas after tests of a finished powder proved it was not giving the results desired.” The objective then, as it is today, was to find a desirable balance between product properties and manufacturing costs.

The DuPont Company was founded in 1802 to produce high quality black powder, as the quality of the black powder in the US was of very poor quality at that time (DuPont Company 1952). Guides at the Hagley Museum in Wilmington Delaware, the site of DuPont's original powder mill on the Brandywine River, explain that one of DuPont’s advantages was development of a device to measure the explosive charge of gunpowder in manufacturing, which enabled them to reduce variation below that of their competitors (Hoerl 1990). That is, DuPont’s product was more consistent than their competitors’ products.

Formulations are typically developed through experimentation. One quickly recognizes that experimenting with formulations is different from typical experimentation, as the response is a function of the proportions of the components in the formulation. This results in the component levels being dependent on each other, as the total amount of the formulation must add to 100%, or 1.0 when expressed as fractions of the total amount.

Our Experiences with Formulations

The first author to recognize this summation constraint was Claringbold (1955). The methodology, literature, and software has developed significantly over the years to the point that formulation scientists have a sound methodology to use, supported by software such as JMP (marketed by the SAS Institute in Cary, NC).

We encountered formulation experimentation early in our careers. Roger Hoerl first worked on paint formulation studies while working as an intern at the DuPont Company in the early 1980’s. He went on to work on other formulation studies as part of his work at Hercules, Inc., Scott Paper Company, and at General Electric (GE) Plastics. At Hercules, his formulation work included coatings and polymer formulations, especially for wrapping films. It was during this work that he developed an approach to applying ridge analysis to mixture problems (Hoerl 1987).

At Scott Paper Company, Roger worked on formulation problems such as wood pulping chemicals and the impact of incorporating recycled fiber with various "virgin" fibers in the pulping process. This was at the beginning of the recycling movement in the paper industry, and a lot of engineers were concerned that recycled paper fiber wouldn't work. We found out that it did!

Ron Snee was introduced to mixture experiments during his PhD work at Rutgers University. Upon joining DuPont, he was thrust immediately into gasoline blending studies for DuPont’s petroleum industry customers. Other formulations followed, including lubricant blending, plastics, and hair sprays. Since the beginning of this century, he has been working on formulation development for pharmaceutical and biotech products.

Ron’s work at DuPont led to several advances in the design and analysis of formulation studies that form the basis of a considerable portion of the DuPont Strategy of Formulation Development approaches that are described in this book. Some of these advances include:

•   Formulation screening experiments: concepts and designs (Snee and Marquardt 1976)

•   Models for the analysis of mixture data (Snee 1973)

•   Computer-aided strategies for designing formulation experiments involving constraints (Snee and Marquardt 1974, Snee 1975a, Snee 1979, Snee 1981, Snee 1985)

•   Estimation of component effects: analytical and graphical techniques (Snee 1975b, Snee 2011, Snee and Piepel 2013)

•   Nonlinear models for designing and analyzing formulation experiments involving mixture and process variables (Snee et al. 2015)

Based on these advances, Ron developed the “Formulations Development Course” that was taught numerous times to DuPont formulation scientists and marketed during the 1980s outside of DuPont. This was the first publically available course on formulation development.

Focusing on simplicity in experiment design, data analysis, interpretation and communication of results includes developing a strategy for experimentation, using the Pareto Principle (Juran and Godfrey 1999) and screening experiments to identify the most important components, using graphical analyses in the exploration and analysis of data as well as in the interpretation and communication of results. Developing parsimonious models to simplify interpretation of results and assessing the practical significance of findings is also an important consideration.

How to Learn Formulation Experimentation with the Use of the Computer

In our experience, people learn best by doing. Accordingly, we have included a number of examples in the book. These examples provide the reader with evidence of the broad utility of formulations in our world and how the methods discussed can enhance the development of formulations.

Many of the examples are discussed in sufficient detail so that the reader can take the raw data provided and reproduce the results reported in the book. In the process, the reader’s confidence builds regarding the understanding and potential use of the methods provided.

All analyses reported in the book were completed using the JMP 13 software marketed by SAS Institute, Inc., located in Cary, NC. We believe that as of this writing, JMP appears to be the best available software for design and analysis of formulation experiments, because of its broad array of design and analysis tools.

Tips and Traps – Success Factors

As we have designed, analyzed and interpreted the results of formulation experiments over the years we have found the following success factors to be particularly important in doing our work:

•   Define clear objectives for the experiment.

•   Create and test theories that will help satisfy the objectives. Iterate between theory and data to confirm or deny theories and build models: Theory A ⇒ Design ⇒ Data ⇒ Analysis ⇒ Theory B ⇒ Repeat

•   Understand the components, including their role in the formulation and the region of experimentation.

•   Be bold, but not reckless. At the beginning of a development project:

◦   Study a large number of components – use screening experiments.

◦   Study the components over a wide, but realistic range.

•   Use a sequential approach with realistic experiment sizes.

•   Be patient – some problems take several experiments to solve.

•   Understand how the data will be analyzed before the experiment is run.

•   Always plot the data.

•   Look for dominant components - components with large effects – that can enhance your understanding of the formulation system and identify useful formulations.

•   Good administration of the experimentation process is critical:

◦   Be sure that the component levels are set and the data are collected as specified.

◦   Avoid missed communications.

•   Test any suspect combination of component levels first:

◦   If no problems are encountered, proceed with the rest of the design.

◦   Consider redesigning the experiments if problems are found.

•   Measure several responses (process outputs or y’s) in addition to the responses of primary interest. The additional cost to do this is usually small.

•   Randomize the runs in the experiment when you can, but don’t let problems with randomization slow down your experimentation and improvement efforts.

•   Conduct confirmation runs after the analysis to verify the model.

As you read through and study the numerous examples in this book, we suggest that you periodically review these success factors and identify how these factors were or could have been used in the different studies.

Acknowledgments

Writing, editing and publishing a book is a process operated by a team. It is a pleasure to acknowledge the contributions of the following members of the SAS Press organization that helped make this book a reality:

Brenna Leath, Developmental Editor
Mark Bailey, Technical Review
Caroline Brickley, Copyeditor
Robert Harris, Graphic Designer
Laura Lancaster, Technical Review
Monica McClain, Production Specialist
Malcolm Moore, Technical Review
Dan Obermiller, Technical Review
Cindy Puryear, Marketing

Our sincere appreciation also goes to our spouses, Marjorie and Senecca, whose support and understanding went well beyond what was reasonable to expect.

Ronald D. Snee
Newark, Delaware

Roger W. Hoerl
Niskayuna, New York

References

Claringbold, P. J. (1955) “Use of the Simplex Design in the Study of Joint Action of Related Hormones.” Biometrics, 11 (2), 174-185.

DuPont Company. (1952) Du Pont:  the Autobiography of an American Enterprise, E. I. du Pont de Nemours and Company, Wilmington, DE.

Hoerl, R. W. (1987) “The Application of Ridge Techniques to Mixture Data: Ridge Analysis.” Technometrics, 29 (2), 161-172.

Hoerl, R. W. (1990) Personal Communication.

Juran, J. M and A. B Godfrey. (1999) Juran’s Quality Handbook, 5^th Edition, McGraw-Hill, New York, NY.

Luce, Clare Boothe. (1931) Stuffed Shirts by Clare Boothe Brokaw (Clare Boothe Luce), Chapter 17: “Snobs, New Style”, Quote Page 239, Published by Horace Liveright, New York.

Smith, W. F. (2005) Experimental Design for Formulation, Society for Industrial and Applied Mathematics, Philadelphia, PA.

Snee, R. D. (1973) “Techniques for the Analysis of Mixture Data.” Technometrics, 15 (3), 517-528.

Snee, R. D. and D. W. Marquardt. (1974) “Extreme Vertices Designs for Linear Mixture Models.”  Technometrics, 16 (3), 399-408.

Snee, R. D. (1975a) “Experimental Designs for Quadratic Models in Constrained Mixture Spaces.” Technometrics, 17 (2), 149-159.

Snee, R. D. (1975b) Discussion of: “The Use of Gradients in the Interpretation of Mixture Response Surfaces.” Technometrics, 17 (4), 425-430.

Snee, R. D. and D. W. Marquardt. (1976) “Screening Concepts and Designs for Experiments with Mixtures.” Technometrics, 18 (1), 19-29.

Snee, R. D. (1979) “Experimental Designs for Mixture Systems with Multicomponent Constraints.” Communications in Statistics – Theory and Methods, 8 (4), 303-326.

Snee, R. D. (1985) “Computer Aided Design of Experiments – Some Practical Experiences.” Journal of Quality Technology, 17 (4), 222-236.

Snee, R. D. (1981) “Developing Blending Models for Gasoline and Other Mixtures.” Technometrics, 23 (2), 119-130.

Snee, R. D. (2011) “Understanding Formulation Systems – A Six Sigma Approach.” Quality Engineering, 23 (3), July-September 2011, 278-286.

Snee, R. D. and G. Piepel. (2013) “Assessing Component Effects in Formulation Systems.” Quality Engineering, 25 (1), January 2013, 46-53.

Snee, R. D., R. W. Hoerl and G. Bucci. (2016) “A Statistical Engineering Strategy for Mixture Problems with Process Variables.” Quality Engineering, 28 (3), 263-279.

Wilkinson, N. B. (1966) Explosives in History: the Story of Black Powder. The Hagley Museum, Wilmington, DE.