We study a fundamental question that arises in social recommender systems: whether it is possible to simultaneously maximize: 1) an individual's benefit from using a social network, and 2) the efficiency of the network in disseminating information. To tackle this question, our study consists of three components. First, we introduce a stylized stochastic model for recommendation diffusion. Such a model allows us to highlight the connection between user experience at the individual level, and network efficiency at the macroscopic level. We also propose a set of metrics for quantifying both user experience and network efficiency. Second, based on these metrics, we extensively study the tradeoff between the two factors in a Yelp dataset, concluding that Yelp's social network is surprisingly efficient, though not optimal. Finally, we design a friend recommendation and news feed curation algorithm that can simultaneously address individuals' need to connect to high-quality friends, and service providers' need to maximize network efficiency in information propagation.