Animal proteins are naturally enriched in 15N relative to the consumed diet and the 15N enrichment of animal proteins over the diet (Δ15N) has been recently shown to capture the between-animal variation in N use efficiency (NUE). The aim of this study was to evaluate the performance of a previous model to predict between-animal variability in NUE from measured Δ15N in lactating cows. The data included 990 records from 17 experiments comprising 164 contemporary groups (CG; group of cows fed the same diet at the same place and measuring period) with an average of 6 cows per CG. Samples of milk or plasma and feed were analysed for 15N and Δ15N was calculated as the 15N difference between animal proteins and diet. Measured NUE and Δ15N of each cow were expressed as their deviations from the average value of the CG they belonged to and noted as observed deviations. Predicted differences in NUE between each cow and its CG average value was calculated as 0.035 × observed deviations of Δ15N (according to the previous published model) and noted as predicted deviations. The coefficient of variation was 10 and 13% for observed NUE and Δ15N within CG, respectively. The correlation between observed and predicted deviations of NUE was positive in 84% of CG. Overall correlation coefficient between predicted and observed deviation of NUE was moderate but significant (r=0.50, P<0.001). The model error showed that Δ15N was able to identify between-animal variability in NUE when the phenotypic deviations among cows surpassed 2.90 percent units. This study confirmed that regardless of the modest prediction accuracy of the validated model, the Δ15N biomarker was able to identify phenotypes of contrasting NUE.