The paper addresses the important topic of third-party sales agencies, which numerous firms use to sell their products, in particular, in Business-to-Business (B2B) markets. These firms are interested in identifying clusters of agencies with similar productivity and in making retention, reward, training, and sales forecasting decisions associated with the agencies. Because these firms can neither observe sales agencies’ efforts nor directly control them, it challenging to model or measure sales force productivity. Existing approaches or models primarily measure the productivity of own sales force, not of sales agencies. Even so, these models make parametric assumptions in productivity estimation and in clustering of sales force, potentially misspecifying the sales response model and misestimating sales productivity. We propose a flexible nonparametric method--Multivariate Dirichlet Process (MVDP) model--that overcomes these problems and simultaneously classifies the sales agencies into groups with similar sales productivity parameters. Using simulation, we first show substantial gains in sales productivity parameter recovery for our model over alternative models. We then estimate our model on a B2B product category as well as on a Business-to-Consumer (B2C) category using two different datasets from two major Fortune 500 firms. We compare our model with alternative approaches and show that it yields more accurate descriptive results, offers a better grouping of similarly performing sales agencies, leads to more profitable training decisions, and facilitates improved interpretation and forecasts in holdout samples than benchmark models. For the B2B firm, training decisions based on our model can yield an incremental profit of $4 million and improved forecasts from our model can boost firm sales in the holdout sample to the tune of $60 million over alternative models.

Our paper offers important contributions to the sales force literature. First, it addresses the unexplored yet crucial issue of productivity of third party sales agencies. Second, it offers a new and improved model of assessing sales force productivity. Third, it provides critical substantive managerial guidelines and insights for retention, reward, training, and forecasting decisions associated with sales agencies. A key insight is that it is more profitable to offer specialized training to fewer sales agencies than standard training to many agencies