A Bayesian Nonparametric approach to Salesforce Productivity

By Sudhir Voleti, Venkatesh Shankar
Journal of Marketing

Citation

Voleti, Sudhir., Shankar, Venkatesh. A Bayesian Nonparametric approach to Salesforce Productivity Journal of Marketing .

Journal of Marketing

Abstract

Many manufacturing firms in the B2B space often use the services of independent third-party sales agencies to sell their product lines. Such firms may be interested in (i) assessing the sales performance of these agencies and (ii) grouping them into clusters of 'similar' agencies in terms of performance profile based on these assessments. Whereas traditionally, a sales response function is specified and estimated for the former, a clustering method is used ex post for the latter. We propose a Bayesian semi-parametric model based on the Dirichlet Process prior that avoids traditional parametric assumptions in the estimation of the sales response function, jointly models quantities of interest in a multivariate setting and also endogenously classifies the sales agencies into groups that share similarities in their sales performance parameters. We contend that compared to alternative methods, this setup avoids several drawbacks such as distributional assumptions, model selection problems and the lack of identification of the mixture components. We design and implement a simulation experiment to assess the magnitude of gain in parameter recovery in the proposed model over benchmark models under different data generation and sample size conditions. To illustrate a real-world application, we implement our model on a short, unbalanced panel of independent salesforce organizations, compare it with extant approaches and find that the proposed nonparametric method yields more accurate results, enables improved prediction in holdout samples, yields a reasonable grouping of similarly performing sales agencies and enables improved interpretation of analysis results compared to benchmark methods.

Sudhir Voleti is an Associate Professor of Marketing at the Indian School of Business (ISB), where he is also a distinguished faculty member in Business Analytics. A renowned researcher in the fields of marketing research and business analytics, he has previously served as Associate Dean of Faculty Alignment and the Registrar's Office (FARO) at ISB.

Professor Voleti holds a PhD in Marketing and an MS in Applied Statistics from the University of Rochester, a PGDM from Indian Institute of Management (IIM) Calcutta, and a BE from the Birla Institute of Technology, Ranchi, along with years of industry experience.

Professor Voleti is recognised as one of India's leading data science academicians. His research focuses on combining data with econometric and statistical methods to explain phenomena of marketing interest such as evolution in the equity of brands across time, valuation of brands using secondary sales data, the sales impact of geographic and abstract distances between products and markets, and the performance, productivity, and benchmarking of salesforce organisations.

Professor Voleti has published numerous research articles in leading academic journals such as Management Science, Journal of Marketing, Journal of the Royal Statistical Society, the International Journal of Research in Marketing, and the Journal of Retailing, as well as book chapters and articles in the popular media. He also serves on the editorial review boards of numerous journals. Some of his significant works include "Impact of Reference Prices on Product Positioning and Profits", "The role of big data and predictive analytics in retailing", "Why the Dynamics of Competition Matter for Category Profitability", "A Bayesian non-parametric model of residual brand equity in hierarchical branding structures", and "An Approach to Improve the Predictive power of Choice - Based Conjoint Analysis".

More publications by author

Sudhir Voleti