On two farms, three milking groups of buffalo cows were used to assess the consistency of entrance order and the preference for one side of the milking parlour. On Farm 1 (F1) all animals were primiparous (n=57). On Farm 2, three primiparous, 16 secondiparous and 36 multiparous cows (range 1–8) constituted group F2G1; whereas group F2G2 had 12 primiparous, 10 secondiparous and 14 multiparous cows (range 1–10). Animals were milked in auto-tandem milking parlours (2×5 and 2×6 for Farms 1 and 2, respectively). For each cow, entrance order into milking parlour, side where she was milked, milk yield, time and duration of milking were recorded. These data were derived from the computerised identification of cows. The sequence in which the cows entered the milking parlour ranged from 1 to 57 for group F1, from 1 to 55 for group F2G1 and from 1 to 36 for group F2G2. The analysis of data was conducted on 130, 120 and 92 consecutive milkings for groups F1, F2G1 and F2G2, respectively. Kendall's coefficients of concordance showed a strong constancy of the entrance order into milking parlour for groups F1 (W=0·658; χ2=4792·81; P<0·001), F2G1 (W=0·779; χ2=5046·81; P<0·001) and F2G2 (W=0·624; χ2=2030·48; P<0·001). Spearman rank correlation coefficients indicated that the more productive cows in groups F1 and F2G1 tended to enter the milking parlour first (rs=−0·221 and rs=−0·215; P<0·10; respectively). In group F2G1, a negative correlation was found between duration of milking and order of entry in the milking parlour (rs=−0·265; P<0·05). Animals in group F2G2 (rs=0·334; P<0·05) with higher days in milk entered the milking parlour latterly. In all three groups, 68 cows (45·9%) preferred the right side of the milking parlour, 73 the left side (49·3%) and the remaining seven (4·8%) showed no preference. Finally, negative correlations were found between mean entrance order and parity for both groups of Farm 2 (rs=−0·319; P<0·05 and rs=−0·325; P<0·05 for F2G1 and F2G2, respectively). As buffaloes showed higher entrance order consistency and side preference than other domestic ruminants, it is concluded that management practices that disturb their choice should be avoided in order to minimise stress during farming routines.