When is Concentration a Good Thing? The Dynamic Effect of Recommendation Concentration on Consumer Purchase

41 Pages Posted: 30 May 2019

See all articles by Jilei Zhou

Jilei Zhou

Peking University - Guanghua School of Management

Fei Ren

Peking University - Guanghua School of Management

Yingfei Wang

University of Washington - Michael G. Foster School of Business

Xiaona Zheng

Guanghua School of Management Peking University

Date Written: April 10, 2019

Abstract

The effect of recommendation concentration is critical to profit maximization of e-commerce platforms. However, recommendation with high concentration level may not always be a good thing to purchase behavior which researchers know little about. To fill this gap, we develop a stochastic model to capture the state-dependent effect of recommendation concentration on purchase behavior. Specifically, we introduce a hidden Markov model (HMM) to investigate the following research questions: (1) is there any hidden state that potentially governs the dynamic purchase journey? (2) If so, how will concentration-level factors affect purchase behavior given a consumer's hidden state? (3) How does recommendation concentration play a role in transferring consumers among different hidden states? Would concentration levels in different recommendation systems (popularity recommender, content-based recommender and collaborative filtering recommender) make a difference? Based on Bayesian estimation of the model with a granular click-stream dataset from a vertical e-commerce platform of liquors, we identify three consumer states governing the purchase process, which are termed as "awareness'', "interest'' and "desire'' respectively. Our results show that the effects of recommendation concentration on purchase are state-dependent. Concentrated recommendation with focal product is effective to boost the focal product's conversion rate in awareness state, whereas it would depress the conversion rate of the focal product in interest state. In addition, recommendation concentration has a significant impact on moving consumer into a deeper state in which the intrinsic purchase intention is higher. Furthermore, through counterfactual simulations of different concentration strategies, our findings offer implications for e-commerce platforms on when and how recommendation concentration should be introduced dynamically, which stimulates more purchase actions eventually.

Keywords: Recommendation Concentration, Purchase Behavior, Dynamics of Consumer State, Hidden Markov Model

JEL Classification: C32, C52

Suggested Citation

Zhou, Jilei and Ren, Fei and Wang, Yingfei and Zheng, Xiaona, When is Concentration a Good Thing? The Dynamic Effect of Recommendation Concentration on Consumer Purchase (April 10, 2019). Available at SSRN: https://ssrn.com/abstract=3384324 or http://dx.doi.org/10.2139/ssrn.3384324

Jilei Zhou

Peking University - Guanghua School of Management ( email )

Peking University
Beijing, Beijing 100871
China

Fei Ren (Contact Author)

Peking University - Guanghua School of Management ( email )

Peking University
Beijing, Beijing 100871
China

Yingfei Wang

University of Washington - Michael G. Foster School of Business ( email )

Box 353200
Seattle, WA 98195-3200
United States

Xiaona Zheng

Guanghua School of Management Peking University ( email )

Peking University
Beijing, Beijing 100871
China

Do you have negative results from your research you’d like to share?

Paper statistics

Downloads
104
Abstract Views
733
Rank
466,166
PlumX Metrics