Milk production data were collected from 150 river buffalo cows raised in six herds in Granma, Cuba, and having from 1 to 10 production cycles, during the 1997-2005 period. The data originated from monthly samples collected in the morning with the presence of the calf. SAS 9.0 was used to analyze the data by means of a model containing the effects of herd, calving year, calving season and calf sex, plus the linear effect of age at first calving, in order to classify the cows according to total milk yield adjusted to 7% fat content (TMYA) or total unadjusted yield (TMY). The individual values of TMYA were then adjusted according to the effects of herd, calving year and calf sex, which had shown significance in the variance analysis. TMY was adjusted for only the first two effects. Spearman's correlation between these two variables was 0.95. The final classification reflects the present state of each cow, taking into account the whole evaluation and the number of lactations. This work has practical interest, since it allows the selection of the best animals for becoming mothers of future bulls and the culling of the least productive cows. Furthermore, information is given on the herds productiveness, based on overall means and their dispersion values. In conclusion, buffalo cow classification can be done either on the basis of total milk yield or on the basis of total milk yield adjusted for fat content, because of the high correlation between both variables; however, adjusted yield showed a higher variability.