It is now commonplace to base performance comparison on productivity measurements. Yates recommended the use of Barber-Johnson diagrams to compare activity in acute specialties, but he did not commend their use in specialties with large numbers of beds and long lengths of stay. Performance comparison in geriatric medicine is difficult because factors such as the case mix and the discharge destination influence throughput. We used two mathematical models to compare the performance of five consultants in the same department. A Monte Carlo simulation based on historic-frequency tables indicated that the mean annual admissions to forty beds by the five consultants would vary between 592 and 748 admissions per year and a two-sample t-test indicated that only two of the consultants were the same, whereas the two-compartment flow model based on the average bed occupancy predicted the annual admissions to be between 650 and 856 and it indicated that the two consultants treated longer-stay patients differently. Thus the models identified differences in consultant performance which were not shown by simple statistics. The information required to decide if the variation in consultant performance is due to clinical or other differences is not yet routinely available. However, the benefit of modelling is that it identifies areas of variation and facilitates exploratory discussion.