Embedding Marketing Environment with Customer Heterogeneity for Store-Level Scanner Data

We propose a marketing model by extending product embedding approach in the way of incorporating marketing variables and customer heterogeneity for store-level scanner data. Forecasting precision of recommendation has been improved by involving the marketing environment and customer heterogeneity in our empirical study, and customer’s demographics have a significant influence on their behavior through the hierarchical model. It also demonstrates that our model clearly distinguishes the distinct types of customers, i.e., (1) interest-oriented, (2) item-oriented, and (3) marketing-oriented customers, which are highly informative for personalized marketing. We also show by simulation study that managers can make efficient marketing strategies by targeting marketing-oriented customers through discovering the promotion effective customers and items by comparing the difference of recommendations between on and off promotions.