An objective method for determining the coagulum cutting time is needed to improve consistency and processing efficiency of goats' cheese. A fibre optic sensor was used to measure the backscatter of near-infrared radiation at 880 nm during the coagulation of skimmed goats' milk for the purpose of predicting coagulum cutting time. A randomised block design, replicated three times, was used to test the effect of three protein concentrations (3, 5 and 7% (w/w)) on diffuse reflectance parameters for cutting time prediction of milk coagulated at five different temperatures (20, 25, 30, 35 and 40°C) to assure a wide range of coagulation rates. The inclusion of a protein term in the existing algorithms was essential to reduce the standard error of prediction to under 6·2 min. An algorithm including a time-based parameter and a protein term, Tcut=β0T2 min (1+γ% Protein) was found to predict cutting time with a SEP of 2·42 min and an R2 of 0·98. γ was considered constant (γ=−0·0674, goats' milk) representing the protein effect on β0. Algorithms using response-based parameters (such as change in reflectance ratio) and the composition parameter protein required additional regression parameters such as temperature and an intercept term to predict the cutting time with the same precision as algorithms using only time-based parameters. Time-based parameters were found to decrease proportionally with increasing temperature and decreasing protein concentration. Response-based and mixed-based parameters were found to decrease with decreasing temperature. Reflectance ratio at cutting time did not significantly change with protein concentration for skimmed goats' milk. The activation energy of κ-casein hydrolysis was calculated based on changes in reflectance profile parameters and was found to be in the range 63–72 kJ mol−1.