Statistical modeling and business expertise, or where is the truth?

Article type: Research Article

Authors: Mandel, Igor

Affiliations: Advanced Marketing Models, 37 Overlook Terrace, Suite 3F New York, NY 10033, USA. E-mail: [email protected]

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 a dialog with a client. Different types of statistical and non-statistical uncertainties, which are usually not holistically considered in a modeling, are systemized.

Keywords: Statistical modeling, marketing mix modeling, regression, causal analysis, yield analysis, business decisions

DOI: 10.3233/MAS-2008-3102

Journal: Model Assisted Statistics and Applications, vol. 3, no. 1, pp. 3-20, 2008