Predictive performance and farmer satisfaction of

Background and objectives: Methodology: The tool was tested by 40 volunteer dairy farmers in the Bobo-Dioulasso milkshed (Burkina Faso) on 82 lactating cows (65 zebu, 17 crossbred). Implementation fidelity was assessed from the mean difference (predicted ration, R2 - actual ration, R3), and predictive performance from robust linear regression, reported overall and by cow type using R², the slope, and error indicators (mean absolute error, MAE; root mean square error, RMSE; mean bias error, MBE), complemented by Bland-Altman analyses. A structured survey assessed the satisfaction of dairy farmers. Results: Dairy farmers implemented R3 ration close to R2 (mean differences -0.37 to +0.04 kg GM/cow/day across feed types). Predicted and actual values were strongly associated for most variables (R² > 0.70), though weaker for milk production among crossbred cows; bias remained limited (MBE = -0.27 L/d/cow for milk production). Agreement on milk production was the most independent performance indicator. Crossbred cows produced more milk (10.6 vs 1.2 L/d/cow), whereas the relative change was greater in zebu cows (33%vs 8% increase). Over 90% of dairy farmers reported increased production and were satisfied. Contributions: This first field assessment is encouraging: under supervised conditions, farmers implemented