As part of the quality assurance and quality control activities within the CarboEurope-IP network, a comparison of eddy-covariance software was conducted. For four five-day datasets, CO₂ flux estimates were calculated by seven commonly used software packages to assess the uncertainty of CO ₂ flux estimates due to differences in post-processing. The datasets originated from different sites representing different commonly applied instrumentation and different canopy structures to cover a wide range of realistic conditions. Data preparation, coordinate rotation and the implementation of the correction for high frequency spectral losses were identified as crucial processing steps leading to significant discrepancies in the CO₂ flux results. The overall comparison indicated a good although not yet perfect agreement among the different software within 5–10% difference for 30-min CO₂ flux values. Conceptually different ideas about the selection and application of processing steps were a main reason for the differences in the CO₂ flux estimates observed. A balance should be aspired between scientific freedom on the one hand, in order to advance methodical issues, and standardisation of procedures on the other hand, in order to obtain comparable fluxes for multi-site synthesis studies.