A decentralised control structure is often preferred to a centralised structure for controlling a multi-unit process. The use of a decentralised control structure requires the plant decomposition. Many plant decompositions may be generated from a particular multi-unit process, even if a decentralised controller employs the same pairing. Therefore, the selection of the best plant decomposition becomes an important task in decentralised control design. In this paper, a new method for evaluating a given set of plant decompositions is proposed. The proposed method treats the interactions between subsystems as a multiplicative output uncertainty. Since the different plant decomposition may result in a different uncertainty bound, the uncertainty bound is used as an indicator to evaluate a set of plant decompositions from which the best plant decomposition may be determined. The calculation of this indicator requires only the open-loop system, so the selection of the best plant decomposition can be performed before the controllers are designed.