OBJECTIVE The accuracy of central axis dose calculation was evaluated for 48 photon beams from 28 linear accelerators at nine centres in Finland. In addition, inter-accelerator consistency of beam data was evaluated for Varian Clinac 600 CDs and 2100 CDs, and averaged data sets were generated for output factors (OFs) and percentage depth doses (PDDs). The averaged data sets obtained were used to identify potential dosimetry reasons for local errors. METHODS Agreement between measured and calculated doses was determined at isocentre at 10 cm depth in water for nine different sized open square and rectangular fields. Averaged OFs were determined for nominal energies of 4, 6, 10, 15 and 18 MV both at d(max) and at a 10-cm depth. In order to develop a function for the OF data, OFs for square fields were parameterised through empirical model fitting. The feasibility of a simple equivalent square collimator formula was also evaluated for the presentation of OFs for rectangular fields. Averaged PDDs were determined at a 10-cm depth. RESULTS The difference between measured and calculated doses exceeded +/-3%, +/-2% and +/-1% for 3, 6 and 35 of the investigated 48 beams, respectively. The differences were due to errors observed in both OFs and depth dose data. When the agreement between dose calculation and measurement was within +/-1%, inter-accelerator differences in OFs were within +/-1.0% at both the depth of dose maximum and at 10 cm for Clinac 600 CDs and also for 2100 CDs. Differences in PDDs were within +/-1.2%. CONCLUSIONS The importance of quality control for beam data was demonstrated by showing significant errors in measured data. For Clinac 600 and 2100 CDs, the quality control can be accurately performed by comparing local data to averaged reference data. Robust averaged data sets were obtained for 6, 15 and 18 MV beams of Clinac 2100 CDs.