This study applies qualitative behaviour assessment (QBA) for the first time to dairy buffaloes, using three groups of observers with different cultural backgrounds and different levels of experience in animal behaviour observation and buffalo farming. Eight buffalo heifers aged 16–18 months were subjected to two isolation tests, one performed in the indoor part of their home environment, and one in a novel outdoor paddock. Animals were filmed individually for 2.5min, and the resulting 16 video clips were shown to three observer panels, consisting of 11 applied animal behaviour scientists from 6 European countries, 11 Italian animal scientists with a background in buffalo farming but no experience in behavioural observation, and 14 Italian undergraduate animal science students with no particular experience. A free choice profiling method was used to instruct observers in QBA, and data for the three panels were analysed separately using Generalised Procrustes Analysis. All three panels showed significant inter-observer agreement (p<0.001) and generated two main consensus dimensions characterised as ‘calm-agitated’ and ‘curious-shy’. There were significant correlations between buffalo scores provided by each of the three observer panels on both these dimensions (dim1: Kendall W=0.96, n=3, χ2=43.28, p<0.001; dim2: W=0.68, n=3, χ2=30.73, p<0.01). Buffaloes viewed in the familiar indoor pen were assessed by all three panels as more calm and less agitated (dimension 1) than animals viewed in the novel outdoor pen (Wilcoxon z=−2.52, p<0.01, z=−2.52, p<0.01, z=−2.38, p<0.01 for Panels 1, 2, and 3, respectively). Scores on dimension 1 for the same animals viewed in either indoor or outdoor pen were correlated at r=0.60 (p<0.10), 0.74 (p<0.05) and 0.71 (p<0.05) for Panels 1, 2, and 3, respectively. Quantitatively, buffalo in the outdoor pen displayed longer bouts of running and higher frequencies of sniffing (both p<0.05) than those in the indoor pen. Principal component analysis showed meaningful associations between qualitative and quantitative assessments, allowing qualitative dimensions to play a valuable role in interpreting the animals’ state. The main outcomes of this study are that QBA can be usefully applied to scientific studies of dairy buffalo, and that substantial differences in observer background do not appear to diminish the reliability of QBA.