OBJECTIVE To assess the meaningfulness of a year's worth of audit data relating to all the inpatients of one consultant general surgeon and to question the usefulness of certain outcome measures. METHODS Analysis of records entered on to audit computer (Dunnfile) and relating to inpatient episodes for one consultant general surgeon over one year. Data obtained were compared with ward records and the patient administration system to check their accuracy. METHODS The three hospitals and 12 wards in Brighton health district where the surgeon admitted patients. METHODS 859 Records relating to inpatient episodes from 1 January to 31 December 1988. These covered 655 main procedures and 79 secondary procedures performed at the same time. METHODS Procedures were analysed by complexity of operation (BUPA code) and grade of surgeon; complications were counted and rates constructed by surgeon and by BUPA code: returns to theatre were analysed. RESULTS Simple counts revealed some data, such as the fact that one registrar performed more major operations (32) than the senior registrars (22 and 14), and an analysis of complications showed that he had a lower complication rate (11.4% v 20.0% and 19.4%). But the simple complication rate disclosed nothing about whether the complication was avoidable. Likewise, the number of returns to theatre needed further qualification. Analysis of data collection for February to April 1988 showed a 30% deficit of information on the audit system compared with ward records and prompted a re-examination of everyone's role in collecting data. After the year's audit there was still a 17% shortfall compared with the district's patient administration system, though some of this was accounted for by a backlog of work. CONCLUSIONS It is difficult to ensure adequate data collection and entails everyone in an unfamiliar discipline. Connecting the audit system to the patient administration system would help. Despite the limitations of crude analyses of workload and complications rates, the audit data helped to measure activity and in the management of the firm. Nevertheless, time and care have to be taken in presenting and interpreting audit data carefully. CONCLUSIONS Counting is not enough.