In this paper, we perform a coverage analysis of a cellular massive multiple-input multiple-output (MaMIMO) network which adopts 3D beamforming. In contrast to the previous works on 3D beamforming which assume that all users are placed on the ground (i.e, in a 2D environment), we consider a more practical scenario where the users of the network are dropped in a 3D environment with different heights. In this scenario, by adopting a stochastic geometry framework, first we calculate the coverage probability of the network as a function of the users height. Then considering this coverage probability as the objective function, the optimum tilt angle of the base station antenna pattern is found and the effect of users' heights on this tilt angle is investigated. Our numerical results show that by taking the users' heights into account and using the proposed method, a considerable improvement is achieved in the performance of the network compared to other similar 3D beamforming methods that ignore the users'height distribution.