A typical banking, financial and administrative system involves specific characteristics: a large number of devices around a processor, with several different kinds of work stations (displays, keyboards, printers, badge and document readers...), a heterogeneous workload (by linkage of specialized micro-transactions using local or remote files), versatile operating facilities on displays for untrained administrative personnel (form-loading on the display, selecting key words, spotting errors, generating operational messages...), and working with several sets of typical functions (savings operations, cheque accounting, fund transfer, deposits, withdrawals, and mainly data entry).In this case it was mandatory to approach the system performance evaluation study by first building and observing a typical workload model in the forecast operating enwironment. Measurement steps were then scheduled from outside to inside operating procedures to get analysis from the user's point of view (a bank teller's operations, for example).Then, overall performance results were derived by direct measurement, which established relationships between throughput, response time, processor overhead, and space and time parameters related to system behavior.That was done by progressively increasing the number of terminals and exercising the workload on two levels of technical and functional saturation.Simultaneously, a simulation model used the same description of the workload, and after validation with the preceding direct measurement results, was used to extend the previous relationships on various systems. (The full range of Erlang distribution parameters is assumed with unknown servers; the trace-driven method was not possible.)The final results are shown in tables and charts which exhibit system boundaries, providing useful guidelines for designing network stations and performing workload forecasting.