The milk analyse result reliability is important for assurance of foodstuff chain quality. There are more direct and indirect methods for milk composition measurement (fat (F), protein (P), lactose (L) and solids non fat (SNF) content). The goal was to evaluate some reference and routine milk analytical procedures on result basis. The direct reference analyses were: F, fat content (Röse-Gottlieb method); P, crude protein content (Kjeldahl method); L, lactose (monohydrate, polarimetric method); SNF, solids non fat (gravimetric method). F, P, L and SNF were determined also by various indirect methods: -MIR (infrared (IR) technology with optical fi lters), 7 instruments in 4 labs; -MIR-FT (IR spectroscopy with Fourier's transformations), 10 in 6; -ultrasonic method (UM), 3 in 1; -analysis by the blue and red box (BRB), 1 v 1. There were used 10 reference milk samples. Coeffi cient of determination (R 2 ), correlation coeffi cient (r) and standard deviation of the mean of individual diff erences (MDsd, for n) were evaluated. All correlations (r; for all indirect and alternative methods and all milk components) were signifi cant (P ≤ 0.001). MIR and MIR-FT (conventional) methods explained considerably higher proportion of the variability in reference results than the UM and BRB methods (alternative).