Interference alignment (IA) has recently emerged as a promising interference mitigation technique for interference networks. In this paper, we focus on the IA non-iterative transceiver design problem in a multiple-input-multiple-output interfering broadcast channel (MIMO-IBC), and observed that there is previously unexploited flexibility in different permutations of user ordering. By choosing a good user ordering for a predetermined IA inter-channel-interference allocation, an improved transceiver design can be accomplished. In order to achieve a more practical performance-complexity tradeoff, a suboptimal user ordering algorithm is presented. Simulation shows the proposed suboptimal user ordering algorithm can achieve near-optimal performance compared to the optimal ordering while exhibiting only moderate computational complexity.