An easy decision-making graphic tool to improve herd level milk yield in a local scale dairy farming system

Abstract

Several features prevent dairy farms from reaching their full potential milk yield levels. A plurality of methods are available to analyse a farm’s yield gap, but in practice, farmers rarely use them to understand their main constraints to production. We propose a simple and graphical approach to tune the limiting (feed-related) or reducing (management-related) factors to evaluate the likelihood of being a high-yielding farm. We gathered data from 32 farms within a local-scale dairy system in Northern Italy. Data regarded milk yield (MY), dry matter intake (DMI), feeding ration’s homogeneity index (Hi), feed sorting (Si) index, ration’s geometric mean particle length (GMPL), ration digestibility, income over feed cost (IOFC) and MY summer-winter ratio (SWR). Farms were classified according to their MY levels into high (H) and low or medium (L + M), with a 36.7 kg × cow−1 day−1 threshold. At an ANOVA model for MY class, H farms resulted in higher IOFC (p < 0.001), GMPL (p = 0.046), DMI (p = 0.006), digestible DM (DDM, p = 0.013), digestible crude protein (DCP, p = 0.011), digestible starch (Dstarch, p = 0.001), and feed efficiency (FE, p = 0.003). At a logistic AIC stepwise regression, the GMPL (odds = 6.528, 95% CI = 1.11-64.2) and DMI (odds = 3.889, 95% CI = 1.43-16.5) favoured farms being classified in the H production class. The nomogram was used to calculate a confusion matrix, achieving an overall accuracy of 0.70, demonstrating its ability to transform predictive models into a graphical, realisable tool.