Economically, cheese yield is very important. At the level of cheese manufacturers, an improvement of cheese yield by 1% increases the benefit without increasing milk consumption. Recently, methods to predict cheese yield based on the mid infrared (MIR) spectrum were described. Based on individual bovine milk samples collected in Walloon Region during 4 years, an equation to predict cheese yield, expressed as g of coagulum per 100 g of processed milk, was developed. This equation was applied on more than 850,000 standardized MIR spectra from the Walloon bulk milk spectral database generated during the process of milk quality analysis for payment between January 2014 and December 2015. The data were analyzed using mixed model procedures in order to study the variability of predicted cheese yield at the level of the Walloon Region. As expected, first results highlighted the effect of the seasons, with a difference of about 3% between least square means of summer and winter. Independently of the season effect, the provincial origin of the milk samples had a highly significant effect. Differences between provincial origins could be related to the main feed managements practised in each Walloon Province, e.g. in the Provinces of Liege and Luxembourg, cattle have more often and longer access to the pasture than in the other Walloon provinces. Furthermore, maize silage is incorporated more often and with a higher level in the ration in the Province of Hainaut. This first investigation highlighted the usefulness of MIR spectroscopy to predict cheese yield from milk bulk sample, which is of great interest to all cheese manufacturers. However, further researches on the variability at the Walloon level will be conducted applying innovative modeling approaches.