Proactive caching has emerged as a promising means to accommodate increased demands for wireless capacity. However, studies of proactive caching usually focus on minimizing the overall load of cache-aided networks. How to calculate each user’s caching gain is still an open problem. In this paper, a two-phase cache-aided network is investigated, in which users with heterogeneous preferences are served by a base station through a shared link. Effective throughput is considered as a performance metric, which describes the reduction in each user’s transmission cost. All possible values of effective throughputs achieved by legitimate caching policies form an achievable domain. It is proved that the achievable domain is a convex set and can be characterized by its boundary. A special type of caching policies, termed uncoded placement absolutely-fair (UPAF) caching, is studied. For the two-user case, games are formulated to allocate effective throughput gains for the two users. For the general multiuser case, a UPAF policy is proposed to organize user cooperation. It is shown that users with more concentrated preferences can obtain higher effective throughputs.