To give some idea of the differences in behaviour in response to real-world disturbances, we have simulated responses to a typical retail sales pattern. The corresponding ORATE variances have already been listed in Table 4 (σ2 calculated from the time series), but we also include here to enable a visual comparison. Note that if the retail sales pattern had been random white noise (or an independently and identically distributed normal distribution), then this σ2 would be exactly equal to that evaluated via noise bandwidth. So the ORATE variances would lead us to expect significantly different responses for the various systems. This is indeed the case, as is quite obvious from Fig. 7, which shows ORATE and the associated stock movements about the target inventory level.