Variability in forage components affects the accuracy of delivered diets and dairy farm sustainability. Optimal sampling and monitoring protocols can inform diet reformulation to improve the accuracy of delivered diets. The objectives of this research were to: 1) quantify and identify the main sources of nutrient composition variability in corn silage and alfalfa-grass haylage at harvest and feed-out; 2) estimate the stable time (1⁄λ) and magnitude of change in forage composition between stable periods (∆) for corn silage and haylage, 3) optimize farm-specific sampling and monitoring practices (number of samples (n), sampling interval (h) and control limits (L)); and 4) evaluate the impact of optimized sampling and monitoring protocols on the accuracy of delivered diets and production outcomes in a commercial setting. For the first three objectives, we collected corn silage and grass-legume haylage samples at harvest and feed-out on 8 NY dairy farms. Results from a mixed-model analysis showed silo unit (storage structure for ensiling forages) is the most important source of variation for corn silage and haylage composition within farms, followed by day-to-day variation; the field where forage was grown was the least influential source of variation. For the second objective, we used k-means clustering analysis to estimate 1⁄λ and ∆ and optimized sampling and monitoring protocols via a renewal reward model (RRM) for objective 3. Results indicated shorter (1⁄λ) ̂ and larger ∆ ̂ estimates than values proposed in previous studies. The larger ∆ ̂ estimates decreased the number of identified shifts and false alarms (components are signaled out of the control limits while they are within the control limits) of corn silage and haylage composition in a quality-control-analysis. For the fourth objective, we applied our optimized sampling and monitoring protocol to corn silage and haylage on a 2000-cow NY dairy to inform the timing of diet reformulation (treatment protocol). The milk yield, feed efficiency, and income over feed costs of the treatment protocols were compared with the standard diet reformulation practices (control protocol). Results showed shorter diet reformulation intervals for the treatment protocols than the control protocol (median of 8 vs. 22 d). The treatment protocol improved the crude protein (CP) and acid detergent fiber (ADF) accuracy of the targeted and mixed diet and likely drove the observed tendency for increased milk yield in the treatment protocol.