More Personalized, More Useful? Reinvestigating Recommendation Mechanisms in E-Commerce

To what extent should firms invest in personalized recommendation mechanisms, and are all personalized recommendations equally welcomed by online consumers? To answer these questions through the lens of resource matching theory, we investigate users' perceptions of three types of personalized recommendations: one-to-all (nonpersonalized), one-to-many (partially personalized), and one-to-one (most personalized). Using both experimental and configurational analysis approaches, our study posits that online consumers differently experience each type of personalized recommendation and their resource matching sources (familiarity, complexity, external information) in various shopping contexts. Our study abductively formulates several theoretical propositions regarding the usefulness of each personalized recommendation. We show empirical evidence that the most personalized recommendation is not always perceived to be as useful as conventionally believed. In particular, highly personalized recommendation is found to be useful for recommending simple technology products for experienced customers. Ironically, a partially personalized recommendation, one-to-many, is perceived as the most useful mechanism for recommending complicated technology products. Based on our findings, we suggest that e-commerce vendors consider the three resource matching dimensions to avoid collecting more than enough customer data, thus enabling adequately personalized recommendation results on their online digital platforms.