In a densely deployed wireless downlink, user selection provides opportunities for higher spectral efficiency. By proper selection of a group of L users from a larger K-size pool, interference can be mitigated improving achievable rate. However, if the group is too large or selection criteria too strict, complexity may be high and performance may be reduced. Previously, greedy user selection methods have been proposed to avoid computationally prohibitive full search. Here, an alternative low-complexity random sequential user group selection scheme for a multi-user multi-input single-output (MU-MISO) downlink system is investigated that utilizes a proposed pairwise user orthogonality metric. The MU-MISO system under consideration contains single-antenna receivers. The cumulative distribution function (CDF) of the proposed metric is shown to characterize system performance and can be computed numerically via Monte Carlo integration and applied to analyze performance-complexity trade-offs in densely deployed precoded MU-MISO systems. Performance and complexity of algorithms that select the best G user group out of the K choose L available in the selection pool are assessed and its merits are compared to current greedy user selection methods. Tradeoffs among achievable rate, group size, pool size, computation, and transmit beamforming for different channel models are presented that vary in numbers of antennas and spatial correlation.