Santa Clara University

Money Well Spent?

A New Model for Determining Marketing Effectiveness

Ling-Jing Kao, assistant professor of marketing at SCU’s Leavey School of Business, is concentrating her research in the area of Bayesian statistics in marketing applications. Named after the British mathematician and Presbyterian minister Thomas Bayes, Bayesian statistics provides a formula for accounting for uncertainty. For example, one aspect of her research has to do with improving marketing decisions by understanding consumer behavior and decision-making.

“The focus is on how consumers make purchase decisions among so many competing products in the marketplace,” she says. “Modeling consumer behavior is challenging because consumers are all different and do not always think in a linear fashion when they make purchase decisions.”

That type of research is suited to the Bayesian approach, which lends itself to analyzing marketing data involving heterogeneous units, such as households, survey respondents or decision units within a business. In this instance, by studying consumer demand and sensitivity to marketing stimuli, retailers can benefit by designing a more effective marketing campaign or targeting more profitable and responsive consumers.

In addition to consumer demand, her research looks at two other areas: Developing models to improve current market-research methods and evaluating the effectiveness of marketing decisions within a firm.

An example of the latter is her paper, “Evaluating the Effectiveness of Marketing Expenditures,” co-authored with Thomas Otter, Chih-Chou Chiu, Timothy J. Gilbride and Greg M. Allenby. It is currently under review at Quantitative Marketing and Economics.

The paper had its genesis in 2004, when Kao was doing her doctoral work at the Ohio State University’s Fisher School of Business and Chiu was a visiting scholar from the National Taipei University of Technology.

Kao and her colleagues were able to get a rich vein of data in the form of information on the marketing expenditure allocations of a financial services firm that maintains multiple branch offices in 21 regions and makes yearly promotional allocations that are spent over the course of the year.

Over the next three years, Kao and her colleagues used the Bayesian methodology to address the question of whether the firm’s marketing expenditure allocation rule was optimal, given the performance of each geographic region. Also addressed was the question of how the firm could effectively allocate expenditures given managers’ assumed knowledge of each branch’s performance.

They found that large promotional expenditures have the same effect as small ones, within the expense range investigated. When promotional expenditures generated positive results, the effect depended on the number of branch outlets in the region. However, branch outlets themselves do not generate new customers without promotional help. Therefore, the model showed that the firm could increase new customers twofold by running less expensive promotions more frequently so the promotions occurred in every measurable time period.

Part of the value of the model developed by Kao and her colleagues is that it relies on a direct estimation strategy that allows a calculation to be done without exact financial information, which is often confidential and unobtainable. The method used in the paper covered that challenge, as well as several others: the mingling of capital and regular expense items, the timing of expenses, and the short time frame (14 months) covered by the data. “It is doubtful that analysts in marketing ever have a complete set of variables for analysis,” Kao writes in the paper, and this approach provides a more flexible basis for assessing marketing efficiencies.

“In the end, this is about how to make better decisions based on a marketing manager’s knowledge,” Kao said. “Marketing decisions interact with each other, and studying the interaction among those decisions is how you improve the overall performance of the firm.”

 
 
Printer-friendly format