In this study three different laboratory procedures were compared as predictors of net energy content of raw materials and compound feeds for ruminants. Laboratory measurements included an enzymatic technique (pepsin‐cellulase method) and a rumen fluid technique (Hohenheim gas test). An estimate based on crude nutrients only was also tested. A total of 98 feed samples with knownin vivodigestibility were available. By multivariate regression analysis equations for the prediction of the net energy content were calculated. The results suggest that in case of compound feeds for dairy the cellulase technique and the gas test are suitablein vitrotechniques. With both procedures the relationships between energy contents estimatedin vitroand energy contents measuredin vivoshowed residual errors of less than 4 percent. However for the evaluation of raw materials the gas test was superior. Compared to the cellulase technique the risk of extreme under‐ or overestimations is apparently reduced and this was found to be the case for own prediction equations as well as equations from literature. Crude nutrients proved to be very poor predictors for all feed samples.