Path of Ascent! is the ultimate guide to Design of Experiments. In an amusing and narrative format, it guides the reader through the techniques of the art of designed experimentation, a powerful statistical technique that is finding applications in vast fields from marketing to production processes. Designed experimentation allows a complex process to be characterized, and thus optimized, in a cost-efficient manner. Eugenio Ballarin has over ten years experience working as a process engineer in the semiconductor and chemical industries, and this work is the synthesis of his cumulative experience using designed experimentation. Having been vastly impressed by the results obtained using the DOE methodology, it was his wish to popularize it for all production engineers and managers. Eugenio is currently employed as a process engineer in the semiconductor industry, working in Italy. 1)Screening experiment. A screening experiment is a sophisticated matrix which allows the filtration of relevant factors from irrelevant ones. Below are some examples of what questions can be answered by a screening experiment and how many experiments are needed to determine the important factors with relative certainty: Which of 4 factors in your process are relevant to achieving a high yield? (only 8 experiments needed) Which of 8 elements are relevant to reach high productivity in your process? (only 16 experiments needed) Which of 11 elements are relevant in determining the appeal of your product? (only 24 experiments needed) Consider that in the third case above, with 11 elements at three variable settings, there are over 177,000 possible combinations to choose from. If one were to approach this universe of combinations without the use of designed experimentation it is nearly impossible to make any sense of it. Being able to understand which of the 11 factors are relevant with only 24 experiments is simply remarkable! Marketing research can thus be carried out in an incredibly effective manner with these techniques. 2) Response Experiment. A response surface experiment attempts to identify the optimal setting of the factors determined to be relevant in the screening experiment. In a range of settings for the given factors a Response Surface Experiment determines what is the optimal setting for each. Traditionally, experimentation for characterizing a production line or any other process involved changing one variable at a time. One variable would be varied over various settings, and then the next variable would be adjusted and so on. Unfortunately, when the ideal settings for each of these variables are combined the results are often mediocre. This is because of what is called the interaction between variables, which means that concurrently increasing two variables which give improved results when increased independently may give negative results when both are increased at the same time. This is why the one-variable-at-a-time approach is inherently flawed, and usually gives misleading results. The alternative is the Response Surface methodology, a science (although many would call it an art) which is based on extensive statistical techniques and years of development. This methodology varies several variables at once in a structured manner that allows extremely efficient statistical analysis. The number of experiments needed to conclude a response surface experiment depends on several factors including the number of variables, the type of variables (continuous or discreet) and the level of accuracy needed. Consider that with five continuous factors, representing an infinite number of combinations, the process can be characterized with around 30 experiments! 3)Path of Ascent Experiment. Selecting the limits for a response surface experiment is not always easy. Often the interactions of all the variables are not understood. It is not uncommon that the best process is found at the edge of the experimental space. In this case, it is necessary to determine if continuing in the same direction will give better results. To do this, a Path of Ascent experiment is needed. This type of experiment requires a limited number of runs, but can greatly increase your bottom line!