Recommendations and Cross-selling: Pricing Strategies when Personalizing Firms Cross-sell

ABSTRACT

Recommender systems enable firms to target customers with products and services that better match their needs, as well as cross-sell products and services. Considering these factors in markets with monopoly and duopoly, we investigate (i) How do pricing strategies differ when firms cross-sell versus when they do not cross-sell, and (ii) How do these pricing strategies change when a firm improves its recommender system? We find that cross-selling can enable a monopolist to subsidize its price for the focal products, while maximizing its profit. In a duopoly, the price set by the firm with the inferior system (low-type firm) is always lower when the firms cross-sell than when the firms do not cross-sell; however, that does not necessarily hold for the high-type firm. When the high-type firm improves its recommender system, the low-type firm may decrease its price when firms cross-sell, which does not happen when firms do not cross-sell.

KEYWORDS:

• Recommender systems
• personalization
• duopoly
• online pricing
• online cross-selling
• game theory
• online targeting

Supplemental data

Supplemental data for this article can be accessed on the publisher’s website.

Notes

1. Pathak et al. (2010) [31] and Formisimo.com [13] note that customers recognize the benefits they obtain from recommender systems and cross-selling, respectively.

2. An ideal product is the one which a customer would like to select after examining (hypothetically) every single product offered by the firm.

3. Although products that are offered for cross-selling are likely to complement the focal product, such product characteristics are not relevant for our model setup.

4. If the firm sells products from multiple categories, the model can be applied to each category.

5. If the customer knows the exact details of the product she intends to purchase, then exploration or recommendations are not necessary as she can just purchase it, if available.

6. When $\theta =0$, the customer has zero search cost for any amount of effort. Then, she would search all the products and find her ideal product to incur zero mismatch cost. For such a customer, the presence of a recommender system does not impact her costs at all. Hence, this is an uninteresting case.

7. We also consider the possibility that the firm cross-sells its own products; however, the primary focus is on the case when the firm cross-sells third party products because the insights are new and particularly interesting for that case.

8. In practice, some, perhaps impulsive, customers purchase the first product displayed. In Section 6, we enhance our model to include such customers in our analysis.

9. We have also analyzed our results for $\theta \sim U\left[0,1\right]$. Our results regarding the pricing strategies remain qualitatively similar for both monopoly and duopoly.

10. When the firm cross-sells its own products, the results remain qualitatively similar to when the firm does not cross-sell. The solution is provided in the Appendix in Section F.1.1 and F.1.2.

11. The third parties would be naturally motivated to improve their service offerings and qualities of products since they will be able to sell more of their products.

12. When the firms cross-sell their own products, the results are qualitatively the same as when the firms do not cross-sell. The solution is provided in the Appendix in Sections F.1.2 and F.2.2.

13. We do not consider a game where the expected cross-selling revenues are strategic variables. As mentioned earlier, these revenues may be realized from profit sharing contracts with partner firms, or from showing advertisements and getting paid according to pay-per-impression or pay-per-click. Involvement of partner firms reduces the flexibility in controlling these revenues, making analysis of such a game less appealing for deriving practical insights.

14. We provide the analysis of only the pricing game in the main paper and do not treat recommender system effectiveness as strategic variables because our primary focus is on pricing strategies. For completeness, we also solve the game with recommender system effectiveness as strategic variables and provide the discussion in Section E of the Appendix.

15. Around 2006, Netflix was already operating a high-quality recommender system called Cinematch, which improved further through the competition entries. Blockbuster, on the other hand, was a new entrant in the online space with a relatively inferior recommender system.

16. In this context, the two firms provide competing services: the opportunity to watch movies that consumers could rent after paying the monthly subscription fee. Neither firms cross-sold other services or products alongside the subscription services at that time.

Additional information

Notes on contributors

Abhijeet Ghoshal

Abhijeet Ghoshal is an Assistant Professor of Information Systems at the Gies College of Business, University of Illinois Urbana Champaign. He received his Ph.D. from the University of Texas at Dallas. Dr. Goshal’s research interests include recommendation system design, data sharing, information privacy, and software support.

Vijay S. Mookerjee

Vijay S. Mookerjee is a Professor of Information Systems and Charles and Nancy Davidson Chair in Information Systems at Naveen Jindal School of Management, University of Texas at Dallas. His research areas include social networks, managerial issues in information security, optimal software development methodology, and the economic design of expert systems and machine learning systems. He has served as senior editor of Information Systems Research and serves as an associate editor of several elite journals.

Sumit Sarkar

Sumit Sarkar is Charles and Nancy Davidson Chair of Information Systems at the Naveen Jindal School of Management, University of Texas at Dallas. He received his Ph.D. from the Simon School of Business, University of Rochester. His research interests focus on machine learning, personalization and recommendation technologies, crowdsourcing, data privacy, sponsored search, information quality, heterogeneous databases, web services, and software release strategies.