In this letter, we propose a novel user clustering algorithm for downlink multiuser multiple-input single-output (MISO) non-orthogonal multiple access (NOMA) systems along with its implementation in a real-time testing environment. The proposed method selects the clusters by considering users' spatial channel properties against the generated orthogonal directions. This is then followed by power allocation and zero-forcing precoding steps to mitigate the interference between the selected clusters. Performance comparisons are provided in terms of both real-time tests and simulations. It is demonstrated that a notable improvement in capacity and reliability can be obtained through the proposed approach in multiuser MISO-NOMA systems with reduced complexity.