To evaluate the performance of an automated sorter typically requires testing of the material to be separated on a full-scale device. In many cases limitations on the quantity of material available for testing, and the costs of the test programme, may limit the range of operating conditions examined. In this paper we propose a method for prediction of automated sorter performance, based on material and machine characteristics, which can be used to evaluate how changes to these characteristics affect performance. The distribution of particles at the point of identification is a key parameter in determining sorter efficiency. A Monte Carlo simulation was used to predict the position of particles on a feed conveyor used to deliver particles to the identification zone. The results produced show that this technique can be used to simulate the particle distribution achieved by an operating automated sorter (TiTech Combisense). Sorter efficiency depends on firstly accurate identification of the target product (typically deflected) and secondly the precision of selective deflection, using compressed air jets or mechanical actuators. A reduction of deflection efficiency is generally caused by the co-deflection of "non-target" particles resulting from their close proximity to particles being deflected. In order to fine tune the sorter model data from automated sorting experiments were analysed to provide information on deflection efficiency over a wide range of operating conditions. The sorter model developed for the -10 + 6 mm size fraction was shown to provide a realistic prediction for coarser (-15 + 10 mm) particles. The sorter model was used to demonstrate that sorter performance could be significantly improved by improved particle feed distribution. When a high separation efficiency is required for a sorting application a 3 stage process was predicted to have a higher separation efficiency than a single stage process using the same number of sorters. (C) 2015 Elsevier Ltd. All rights reserved.