This study aimed to evaluate the feasibility of mid-infrared spectroscopy (MIRS) to predict rennet coagulation time (RCT), curd-firming time (k₂₀) and curd firmness (a₃₀) of buffalo milk. One hundred and sixteen milk samples were collected and analysed using both reference analysis and MIRS. Reference measures of RCT, k₂₀ and a₃₀ were matched with milk spectra information (5,000 to 900 cm^-1 wavenumber) and prediction equations were developed for each trait using i) cross-validation on the whole dataset and ii) external validation on a subset of the entire data. The prediction models were evaluated through the coefficient of determination (R²) and the residual predictive deviation (RPD). The most accurate prediction model was developed for RCT (R² = 0.72; RPD = 1.88) in cross-validation. Models developed for RCT and a₃₀ allowed a quite satisfactory prediction of milk coagulation properties. Nevertheless, the accuracy was not enough to suggest their application in milk payment systems. Instead, they might be interesting for breeding purposes.