Model-Based Purchase Predictions for Large Assortments

Abstract

Being able to accurately predict what a customer will purchase next is of paramount importance to successful online retailing. In practice, customer purchase history data is readily available to make such predictions, sometimes complemented with customer characteristics. Given the large assortments maintained by online retailers, scalability of the prediction method is just as important as its accuracy. We study two classes of models that use such data to predict what a customer will buy next: A novel approach that uses latent Dirichlet allocation (LDA), and mixtures of Dirichlet-Multinomials (MDM). A key benefit of a model-based approach is the potential to accommodate observed customer heterogeneity through the inclusion of predictor variables. We show that LDA can be extended in this direction while retaining its scalability. We apply the models to purchase data from an online retailer and contrast their predictive performance with that of a collaborative filter and a discrete choice model. Both LDA and MDM outperform the other methods. Moreover, LDA attains performance similar to that of MDM while being far more scalable, rendering it a promising approach to purchase prediction in large assortments.

Topic modeling in marketing: recent advances and research opportunities

“…While the former three represent about 89% of publications, the latter four comprise of 7 papers. In general, 10 of 61 papers were published in (mostly quantitative) marketing journals, of which 5 can be assigned to sales / retailing (Blanchard et al 2017;Hruschka 2014;Jacobs et al 2016;Schröder 2017;Trusov et al 2016), 3 to online textual consumer reviews and services research (Büschken and Allenby 2016;Calheiros et al 2017;Tirullinai and Tellis 2014), 1 to social media (Song et al 2017), and 1 to research in marketing literature (Amado et al 2017). Furthermore, we transferred the sum (numbers), and the relative importance (color) of the methodological strategies exerted by scholars into a matrix (Table 6) to better detect patterns and predict trends in research.…”

“…1015Galyardt 2015, pp. 42), a substantial amount of published research tend to be merely experimental than focusing on substantial results (e.g., Jacobs et al 2016;Phuong and Phuong 2012;Wang et al 2015). Additionally, because of the large variation of models (Airoldi et al 2015, pp.…”

Editorial—Marketing Science and Big Data

“…For example, latent Dirichlet allocation (LDA) is normally used in text processing to identify "buckets of words." Jacobs et al (2016) turn LDA on its head to identify, from the consumer's perspective, sets of products that tend to be purchased together. Their analyses have the potential to improve product recommendations and thus contribute to the recommendation system literature.…”

Learning Preferences with Side Information

“…Recommendations have also been studied in the context of promotions (Garfinkel et al (2008)) and the firm's profit (Hosanagar et al (2008)). Finally, and Jacobs et al (2016) seek to predict future purchase probabilities; these estimates can have important applications such as demand estimation.…”