We estimate demand for residential broadband to study the efficiency properties of usage-based billing. Using detailed-high frequency internet protocol data records, we exploit variation in the intertemporal tradeoffs faced by subscribers under usage-based billing to nonparmetrically estimate subscribers-preferences for different characteristics of service; access and overage fees, usage allowances, and connection speeds. We find significant heterogeneity in tastes along each dimension of service. Using these estimates, we examine the efficiency of various 3-part tarriff pricing schedules, weighing subscriber welfare against the costs of accomodating traffic while maintaining a target level for quality (utilization of network resources). Our results show that under the efficient picing schedules; 1) a very small percentage of subscribers are impacted, 2) the loss in welfare for those impacted is relatively small compared to network cost savings. Collectively, our results provide strong support for the FCC's claim that usage-based pricing is effective at increasing e¢ ciency in broadband networks.