Inferring salesperson capability using stochastic frontier analysis

This paper proposes an easy-to-implement econometric method for inferring salesperson capability from archival panel data, namely stochastic frontier (SF) analysis. We demonstrate this method with a sample of salespersons provided by a life insurance company. Using the proposed SF model, we are able to estimate each salesperson’s capability. Furthermore, we examine the relationship between the estimated salesperson capability and three future outcomes (i.e. future sales performance, future customer attrition, and future salesperson turnover) under different time horizons. We find that, in general, the estimated salesperson capability has a stronger explanatory power for the near than for the more distant future. Since an individual salesperson’s capability cannot be directly observed by researchers (and thus is typically omitted), traditional analyses of sales performance suffer from an omitted-variable problem that can lead to biased estimates of focal variables. The SF model can significantly mitigate this omitted-variable problem. Statistical tests indicate that our sales performance model with estimated salesperson capability results in a statistically significant improvement in model fit. Of note, our model differs methodologically from SF models previously used in the marketing literature in that it is based on a three-component model that disentangles unobserved individual heterogeneity, efficiency, and random shocks.