Customer-Product Matches in Online Social Referrals

Abstract

New social media platforms have significantly increased the impact of social referrals on product diffusion. However, empirical evidence on how effectively referrers actively match products with the customers they refer, thereby increasing the returns on online social referrals, is limited. To address this, we collected and analyzed data from both field and lab settings. In the field study, we analyzed 137,622 referrals that involved 20,169 instant apps and 1,141,363 unique users including referrers, recipients and 100 contacts randomly selected from each referrer's local social network. Using a graph embedding framework, we estimated users' latent preferences for these apps based on the massive-scale historical user-app interaction data. In the lab study, we examined the social referrals of three apps among 221 participants from four networked communities, collecting data on their product preferences and local social networks. Our findings consistently reveal that referrers recommend products to their contacts who have significantly stronger preferences for these products than non-referred contacts. This supports the effectiveness of active matching. We also found that the effectiveness of active matching is greater for products with a narrower appeal and when referrers are more engaged with the products or platform, or have smaller local social networks. Our study provides some of the first empirical evidence on the effectiveness of active matching in social referrals.

Keywords: Social referral, social networks, matching, graph embedding, product preferences