Diet preferences in grazing animals are important for range management and ecology. However, estimating the composition of diet mixtures is challenging. The plant-wax marker technique, combined with nonnegative least squares (NNLS), has traditionally been used to estimate diet composition across herbivore species. More flexible methods, such as the normal compositional model (NCM) under Bayesian inference, offers an alternative approach for predicting diet composition. The efficiency of NCM was assessed by simulation. Mean n-alkane (C₂₇, C₂₉, C₃₁, C₃₃) and long-chain alcohol (LCOH; C₂₆OH, C₂₈OH, C₃₀OH) concentrations (mg/kg) were obtained for 11 subsamples for each of 2 forages – tall fescue and red clover – by gas chromatography. The CV of those measures was 10.7%, which was used to derive a common SD (20 mg/kg). Forage mixtures ranging from 0.10 to 0.90 fescue, in 0.10 increments, were simulated from a Gaussian distribution. Values for each n-alkane and LCOH, using their respective mean and the common SD, were drawn independently. For each mixture, 100 sets of observations were generated. Data were analyzed using NNLS and NCM. For the n-alkanes and LCOH alone and in combination, efficiency was assessed by normalized mean squared error (NMSE), mean average difference (MAD), and coverage of 95% CI. For both types of markers, both methods accurately predicted the forage mixtures, although more so with NCM. The NMSE for the NNLS were 0.032%, 0.019% and 0.012% for n-alkanes, LCOH and their combination, respectively; for NCM, those respective values were 0.028%, 0.011% and 0.008%. Similarly, MAD for NNLS were 1.29%, 0.92% and 0.75% for n-alkanes, LCOH and their combination, respectively; for NCM, the respective values were 1.22%, 0.74% and 0.63%. Coverage was high for the combined markers: 97.1% for NCM and 91.2% for NNLS. Given the scenario used, the NCM more accurately predicted forage mixtures. In the current study, a common SD was assumed. Such is doubtfully the case in practice, and the NCM methodology can accommodate non-trivial covariance structures among markers and better model differences in variation for individual markers. Such extensions are underway. The NCM approach provides a robust tool for estimating forage mixtures.