Radiotherapy workload is poorly represented by simple parameters of patients, fractions or fields treated because these do not contain any measure of treatment complexity. However, complexity is increasing and there is an urgent need to quantify this. We have evaluated the basic treatment equivalent (BTE) model as a measure of radiotherapy workload and complexity. Radiotherapy treatment times, from the patient entering to exiting the treatment room maze, were measured for 1298 treatment sessions on 269 patients. The data were used to assess the original model and derive three new models for predicting treatment duration. The most complicated, the 'Addenbrooke's complex model', contained two additional predictor variables, including 'site/technique', in a linear additive form. Before the study, the department used a standard treatment appointment time of 10 minutes. However, 50% of the measured treatments took longer than 10 minutes, (mean 10.9). Summed over the working day, this discrepancy indicates that a standard 10-minute appointment is a poor basis for scheduling radiotherapy. The original BTE model was effective in predicting treatment times, although this was improved by refinement of the model. The Addenbrooke's complex model correctly predicted 70% of treatment times to within 2 minutes (55% for the original BTE model), 80% to within 2.5 minutes and 95% to within 4.7 minutes. The percentage of the variation in observed times accounted for by the model is 59.4%. The models can represent radiotherapy complexity, can improve scheduling on linear accelerators, and are likely to be applicable to other departments. They are thus tools to assess the impact of changes in complexity from new techniques, trial protocols (e.g. the Medical Research Council prostate radiotherapy trial RTO1), and possible time saving from advanced technology such as multileaf collimators (MLCs) or automated machine set-up. The replacement of manually-lifted shielding blocks by MLCs should save 1.1-1.5 minutes for a three- or four-field pelvic plan (i.e. 12%-13%). The models could also be used to aid planning for future linear accelerator provision and for costing radiotherapy treatment.