This paper proposes a two-stage beamforming (TSB) scheme with user scheduling for FDD massive MIMO. The developed TSB scheme designs the analog pre-beamformer and schedules the users using statistical channel state information (S-CSI), reducing the overhead of the pilot and the feedback. Particularly, in the one-ring local scattering channel model, the pre-beamformer design and user scheduling problem is formulated as a 0-1 quadratic constrained quadratic programming (QCQP), which is further linearized to a mixed integer linear programming (MILP). In the multiple scattering clusters channel model, we design the pre-beamformer and schedule the users based on graph theory, where the chromatic number of the equivalent matrix represents the minimum number of orthogonal pilots. Then, we propose an iterative beam selection and user scheduling (I-BSUS) scheme that approximates the minimum pilot constraint by the maximum vertex degree. Moreover, the net spectrum efficiency (NSE) is improved using a multi-user digital precoder, which depends on the effective instantaneous CSI (EI-CSI). Simulation results validate the superiority of the proposed scheme in enhancing the NSE over the existing schemes.