Evolution of Referrals Over Customers' Life Cycle : Evidence from a Ride-Sharing Platform

Online platforms often ask their users to refer friends in exchange for a reward. This paper addresses how referral decisions and referral quality evolve throughout the customer's life cycle as a function of service usage, experience level, and past referral behavior. Our analysis is based on a longitudinal dataset that comprises the transactions and referral actions of 400,000 users in a ride-sharing platform over a year. The richness of our dataset allows us to address two shortcomings from previous studies: modeling dynamic behavior (i.e., the relationship between past and future referrals) and accounting for unobserved heterogeneity across users. Our results show that users make more referrals when they use the service intensively and frequently. For example, users become 8% less likely to make referrals for each week they have not used the service. Furthermore, users make more high-quality referrals as they become more experienced with the service. We find that the top 10% most experienced users are 13% more likely to make referrals, and their referral quality is 19% higher relative to when they first used the service. Finally, as users make more referrals, they become more likely to run out of friends to whom they can refer the service, leading to less referrals in the future. On average, the probability of making additional referrals decreases by 73% after a user makes a referral, and the quality of subsequent referrals reduces by 19%. These results offer managerial insights that can be leveraged to tailor referral programs throughout the customer's life cycle