A Personal Story of DOE

Bill Kappele

Design Of Experiments, or DOE, has done more for my career as a chemist than anything else I have learned. It is the most powerful tool for experimenters that I have ever seen.

I first encountered Design of Experiments at Hewlett Packard after receiving my chemistry degree. My boss asked me to help chemist at the Corvallis plant with an experiment. I was to prepare inks to her specification and make the measurements for her. When I saw her experiments, I was dismayed. She was varying all the ink ingredients at once. Every scientist knows that you vary only one ingredient at a time so you can understand which one is having an effect. I felt that she was wasting her time and mine with her poorly planned experiments.

I talked with my mentor, an excellent chemist, and ask him how to avoid wasting time and do the experiments right. I explained the situation and arrogantly stated that this was the worst set of experiments I had ever seen. My mentor calmly told me that, although he didn’t understand how it worked, he knew that their was a powerful technique called DOE that actually varied all the factors in an experiment at the same time. He suggested that I perform her experiments for her as she had specified. He also told me to ask her about DOE.

Feeling the sting of humility, I called my Corvallis co-worker and asked her about DOE. She told me that HP offered an internal class in DOE and that I should sign up for it. I signed up for the class immediately. In the class I learned how to use published tables of experiment designs to vary all the factors at the same time in a logical way to gain the maximum amount of experimental information with the smallest number of experiments. I also learned about interactions. When you bake, the time and the temperature interact. If you increase the temperature, you must decrease the time, etc. Temperature and time work together – they interact. Interactions are extremely common in nature, but I had left college with the mistaken impression that it was best to look at factors one at a time, ignoring the fact that they often interact. I had the basics, and I was ready to apply them to my work.

Back in the lab, I tackled ink with DOE. I was able to perform a small number of experiments and learn about interactions among the ingredients. I could see which ingredients appeared to be the most important, which ingredients interacted, and which interactions were most important. This really was a powerful technique. It wasn’t as good as I had hoped, though. I learned a lot, yes, but my real goal was to know how to combine the ingredients to make the best ink. It seemed like I had the information I needed, but I couldn’t see how to use it to full advantage.

One day good fortune struck. I received a postcard from a company called Edgework. They were offering a class in DOE and something called "Response Surface Methodology." In the upper left-hand corner of the card was a little picture. The picture showed a plot of Hours vs. Temperature for some unknown process. But the plot had contour lines on it – lines of equal product performance. It looked like I could predict the combination of factors that would give me the best product. I called the number on the card and talked with Dave Doehlert. He assured me that he could teach me to predict the best product – what he called the "Sweet Spot." I registered for the class.

I discussed the course fee with my boss and told him that the software used in the class was offered at a discount price to students. I said I thought we could wait and see if I liked the software before purchasing it, although we would not get the discount. My boss was much wiser than I was. He told me to purchase the software at the discount price with the class. He reasoned that if I learned the techniques with the software, I should have the software when I returned to work so I could begin using what I had learned immediately. I called Dave Doehlert back and told him I wanted the software too.

The class was eye opening. Dave explained everything in plain English. I didn’t have to learn a new vocabulary – I could concentrate on learning the method. I could not believe how easy it was to determine the best combination of factors to get the best product. I had never seen anything like this in college.

When I returned to work I put the method to work right away. Hewlett Packard had introduced the Paintjet printer a few months earlier and its performance on overhead transparencies was poor. Many people had tried to fix this problem with no success. My boss asked me to try DOE on this problem. I did – and I found the answer very quickly. The resulting inks were so successful that HP awarded me stock options for my work.

This success alone would have made my time in class worthwhile, but I continued to have success with the method in different jobs. I used it to optimize plating processes, perform chemical interference studies, optimize chemical analyses, and optimize chemical stripping operations to name a few.

When I returned to ink-jet ink formulation several years later at Lexmark international I experienced a little déjà vu. When I was hired, Lexmark chemists had been struggling to develop a black pigmented ink-jet ink for a few years. They were in a quandary – nothing was working. I was able to use DOE to solve the problem. Lexmark had the black ink on the market nine months after I was hired. They awarded me stock options for my work.

I can’t guaranty you stock options. I can guaranty that if you use Design of Experiments properly and consistently in your work, you will be a top performer.