For years, promoting products from the same category that the customers previously purchased from has always been the mainstream in recommendation systems, since it is more likely that products from the same category will match and correspond with customers existing purchase preferences. However, the focus of the recommender system research community has shifted their attention from accuracy-based recommendation strategy towards diversity-based recommendation strategy, this entails offering products from new category that the customer never purchased from. The shifted interest is mainly caused by the ever-increasing impact of long-tail phenomenon on the current e-commerce practice, instead of choosing the mainstream products people tend to search for niche products that can better serve their needs, the advance of recommendation systems is considered as a great stimuli of this effect since it reduces the search effort greatly.

In the article it was empirically tested that both the purchase quantity and the variety of purchased product categories are positively correlated with customer retention rate. Moreover, the retention rate increases dramatically if the purchase is made in new product category instead of an existing product category (Park & Han, 2014). According to Kamakura et al (2003), this is mainly due to the fact that switching cost increases as the number of interaction points between company and customer increases. Furthermore, the continuity of customer-company long-term relationship is build on company’s potential in fulfilling current and future customer’s needs. If a single E-tailor can offer basically all the desired products, then it is logical that the consumers will stay with that E-tailor, there is no need to search for different E-tailor. Amazon and Taobao (Chinese equivalent of Amazon) are great example of such E-tailor.

Additionally, out of curiosity and to test the findings of this article in another real life setting. I went to Qidian, largest E-book publishing website of China, to test out the personalized recommendation system which bases its recommendation from the books stored in my virtual bookshelf. Despite the fact that my virtual bookshelf mainly consists of books of horror genre, the recommendation system still suggested books of other genres, such as romance, history and fantasy (figure.1). This implies that the recommendation system does not solely base its recommendation on genre (the main attribute) but it also takes the tags (humor, thrilling, adventure, etc) that these books associate with in to the account. For example, a fantasy book might be as adventurous and thrilling as a horror book can be. Therefore this fantasy book might still suits my taste despite the fact that it is different in main attribute.

Another example could be found at Amazon; I never searched or purchased any toy related items from Amazon, yet Amazon recommended me toys (figure 2). This might bring me to the thoughts to purchase toys for my little cousin through Amazon from now on instead of visiting brick-and-mortar shops. Because the diversity-based recommendation system heightens my awareness of the potential products categories that could be bought from Amazon and thus my retention rate is being fostered.

A final remark, even though recommending diverse product might be beneficial in discovering latent needs of costumer and consequently providing better service to them, it might also lead to negative emotions such as annoyance and frustration if the recommended item deviates too much from customer’s preferences or conflicts with their purchase behavior, therefore it is necessary for firms to be extra cautious in applying diversity-based recommendation system.

 

Author: 373227yj

References:

• Park, S.H., Han, S.P. (2014) ‘From Accuracy to Diversity in Product Recommendations: Relationship Between Diversity and Customer Retention’ International Journal Electronic Commerce, 14, 18: pp. 51-71
• Kamakura, W.A., Wedel, M., Rosa, F.D., Mazzon, J.A. (2003) ‘Cross-selling through database marketing: A mixed data factor analyzer for data augmentation and prediction’, International Journal of Research in Marketing, 20, 1: pp. 45-65
• http://www.Qidian.com
• http://www.Amazon.com