Essays on marketing models and modeling environments

Radha Chandrasekharan, Purdue University

Abstract

This paper describes the design and implementation of OR/SM, a computerized modeling environment based on Structured Modeling. The uniqueness of OR/SM is in the following: (1) the use of ORACLE$\sp\circler$ Tools and Database as the delivery platform; (2) an interactive link to QS (Quantitative Systems)--a commercial software package for solving a wide range of operations management models; and (3) automatic and interactive links to SAS$\sp\circler$, a powerful and widely used commercial statistical analysis software system and optimization solver. Some other key features are: (1) automatic generation of relational database tables for model data; (2) interactive checking of model syntax and semantics; and (3) automatic generation of several reference documents. Examples from inventory control and marketing mix management are used to illustrate the capabilities of OR/SM. Disaggregating macro forecasts or market potentials (e.g., at the national and/or annual level) into micro (e.g., at the territorial and/or weekly level) forecasts or potentials is an important task frequently faced by firms in marketing, planning, control, and performance evaluation. Some key examples of marketing activities requiring the application of disaggregation techniques include territorial demand assessment, sales forecasting, salesforce allocation, and budgeting for promotion and advertising. Given its importance, it is surprising that the development of disaggregation schemes and methods has received little (if any) attention in the marketing and management science literature. This study proposes a new econometric procedure for disaggregating market potentials and forecasts on a time series and/or cross-sectional basis. In our empirical application, we use two years of A.C. Nielsen SCANTRACK weekly data on a major food category. The data set contains information by territory on sales, marketing instrument variables such as retail prices and promotions, demographics, and store count data by product line in the 50 Nielsen markets. A significant finding of our exhaustive empirical analysis that the proposed procedure clearly outperforms two existing econometric methods for disaggregation. Our results suggest a new area of potentially significant theoretical and empirical research on the application of econometric and time series models in marketing.

Degree

Ph.D.

Advisors

Wright, Purdue University.

Subject Area

Marketing

COinS