Statistical modeling and business expertise, or where is the truth? Igor Mandel Advanced Marketing Models, 37 Overlook Terrace, Suite 3F New York, NY 10033, igor.mandel@AMModelsInc.com Abstract Statistical model results may strongly contradict domain knowledge or expectation, which generates many problems considered in this article. In a nutshell, who is right: a statistician stating that advertising does not work (based on the model), or a marketing officer stating the opposite (based on experience or “gut feeling”)? The answer is not as obvious as it seems. All statisticians face this dilemma almost every time they present results to a “non-statistical” audience. The combinations of subjective and objective, derived and desirable, free and forced aspects of modeling could be complicated, often posing difficult scientific and moral challenges. The problem is considered from practical and theoretical points of view. A brief review of important statistical concepts is given to position a statistician’s role in dialog with a client. Different types of statistical and non-statistical uncertainties, which are usually not holistically considered in a modeling, are systemized. Key words: statistical modeling, marketing mix modeling, regression, causality analysis, yield analysis, business decisions. Introduction Statistics is the science of mass phenomena under the condition of uncertainty (it is one out of many possible definitions, but it bears most of the important features). Business is the art of “getting things done”, presumably pursuing one’s profit or the “general interests of society” (if the term “business” is applied to governmental structures, which is often the case). Statistical analysis or modeling is performed by qualified specialists and is based on the developed, always-changing methodology. Decision makers who execute business solutions are not specially qualified; rather they gain status through right of birth, talent, hard work, or luck. Even this simple difference sets the grounds for many possible (and real) odds between two ways of thinking and acting. More and more business decisions in recent years have been made based on some statistical knowledge, just as they always have been (maybe unconsciously) throughout history. This knowledge is not necessarily “scientific”; it’s just “business intuition” based on some key parameters. As a hero in a novel (Murukami 1999), a successful owner of a restaurant chain explains: before opening a new restaurant he goes to some preliminarily picked corner in Tokyo and observes people running there day by day. He doesn’t take any notes; he just tries to imagine what will be changed at this corner if he opens a restaurant there. In couple of weeks of such observations he makes a decision. He chooses to either purchase a spot for a new restaurant or to go to another corner. Is it a “statistical” decision? Yes indeed, but not the only one. Imagine that an educated statistician always follows this businessman and collects data about people running around, their demographics and so on. At the end of the task he proposes the recommendations to his client, and it happens that they are the opposite of the boss’s intention. What is the expected outcome? Most likely, a businessman will follow his gut rather than any scientific evidence. The first question is why? And the next question is – what is the statistician’s reaction to that decision? This article is about possible answers to questions like these. Its subject is the relationship between statistician and client. A client is determined either an employing company for a statistician or an outsider’s company (further a chain of stores is considered as a client and an object of modeling). Typically a statistician believes in domain knowledge of a client, whereas the client believes in statistical techniques – otherwise a contact (and contract) would be impossible. They both are right and wrong. The statistician is right to say that domain knowledge is lying somewhere on the client side, but he may be wrong in locating it there. A businessman is right trusting statistical techniques because there is nothing else he can do (to say that “science does not work” is taking a step back and cannot be defended on an upper level). But he is wrong