The Impact of Product Recommendation Networks on Sales: The Moderating Influence of Product Age

In many online stores, recommender systems create a product recommendation network by placing links to recommended products on a product page. The importance of a product in such a product recommendation network can be measured by its PageRank centrality. Previous research found that the PageRank of books in a product recommendation network is a predictor for its expected sales. However, it is unclear whether this relationship is affected by the age of a book. Consequently, this chapter studies the influence of book age on the relationship between PageRank of a book and expected sales (SalesRank). We use various multivariate regressions to analyze how the relationship between PageRank centrality and SalesRank is moderated by product age. We collected data from a large online bookstore over a period of 35 days and find that the PageRank of a book is a significant predictor of expected sales if only books younger than 170 days are considered. When including also books in the regression models older than 170 days, we find no significant influence of PageRank on expected sales. Thus, book age moderates the influence of PageRank on SalesRank. We do not only find a direct influence of book age on expected sales but also confirm a significant quadratic moderating effect of book age. As a result, if online stores want to increase the demand of a specific product, recommendations should also take the age of a book into account as for very new as well as old books a high centrality of a book in a product network does not lead to higher sales.