Beyond Zero: Jump-Starting Sales With a Recommender System for Missing-By-Choice Data

Abstract

New salespeople often face a costly productivity hurdle. We develop a marketing engineering solution: a recommender system that helps new salespeople recognize suitable customers based on the historical sales records of experienced salespeople. One challenge is how to learn from experienced salespeople’s own failures to make a sale, which are prevalent but often do not show up in sales records. We specify a basic economic model in which sales records are missing if a salesperson’s gain from the sale does not justify the cost of selling. We incorporate the model into a deep learning framework to form a “Missing-By-Choice (MBC) recommender system” that handles this type of missing data. We develop our solution using sales force transaction data from Minsheng Life Insurance, one of the largest insurance companies in China. The MBC recommender system outperforms common benchmarks and predicts a reduction of failures before the first sale by 29%, while being simple, explainable, and adaptable. Minsheng has now deployed the MBC recommender system.

Keywords: sales force management, marketing engineering, recommender system, deep learning