A mechanistic broiler growth model was evaluated to determine its predictive performance for body weight (BW), body weight gain (BWG), feed intake (FI), and feed conversion ratio (FCR). A database was compiled from eight controlled research trials and three published studies, representing two commercial strains (75% Ross 308; 25% Cobb 500) and yielding 199 phase level pairs of observed and model predicted outcomes. Each study was simulated using the same diet formulations, ingredient profiles, and rearing conditions described in the original experiments, and predictions were compared to observations by phase. Adjacent diets within each study were paired, and differences in performance (ΔObserved and ΔPredicted) were calculated to assess whether the model correctly reproduced the direction and magnitude of responses to changes in standardized ileal digestible (SID) lysine and metabolizable energy (ME). Precision was quantified by the coefficient of determination (R2) from simple linear regression of observed on predicted values, and accuracy was summarized by the mean absolute percentage error (MAPE). To establish a reference threshold for accuracy, baseline biological variation was estimated from the same database. Across the full dataset, the model closely matched observed performance: BW MAPE 5.39% with R2=0.993; BWG 6.55% with R2=0.988; FI 10.0% with R2=0.990; and FCR 7.01% with R2=0.909. Delta analyses confirmed that the model reproduced both direction and magnitude of responses to diet changes. For SID Lys (n=126), MAPE and R2 were: BW 4.17%, 0.99; BWG 8.02%, 0.96; FI 10.44%, 0.98; FCR 6.94%, 0.90. For ME (n=81), results were: BW 4.4%, R2=0.999; BWG 5.58%, 0.997; FI 9.06%, 0.991; FCR5.9%, 0.954. As a benchmark, the natural between trial variation under standardized conditions averaged MAPE ≈10.6% for BW, 9.56% for BWG, 10.8% for FI, and 15.4% for FCR, indicating that the reported model errors where within typical biological variability. Overall, the model reproduced growth performance and nutrient response patterns with high precision and satisfactory accuracy across diverse datasets. These findings support its application for comparing feeding strategies, projecting performance, and conducting “what if” nutrition scenarios in broiler production.

Reis, M., A. Garcia-Ruiz3, N. Jaworski and  N. Ferguson. 2026. Evaluation of a mechanistic broiler growth model. 2026 International Poultry Scientific Forum, abstract 329P.