Implementing linear appraisal systems (LAS) may reduce time, personnel and resource costs when performing large-scale zoometric collection. However, optimizing complex zoometric variable panels and validating the resulting reduced outputs may still be necessary. The lack of cross-validation may result in the loss of accuracy and value of the practices implemented. Special attention should be paid when zoometric panels are connected to economically-relevant traits such as dairy performance. This methodological proposal aims to optimize and validate LAS in opposition to the traditional measuring protocols routinely implemented in Murciano-Granadina goats. The sample comprises 41,323 LAS and traditional measuring records from 22,727 herdbook-registered primipara does, 17,111 multipara does and 1485 bucks. Each record includes information on 17 linear traits for primipara/multipara does and 10 traits for bucks. All zoometric parameters are scored on a nine-point scale. Cronbach’s alpha values suggest a high internal consistency of the optimized variable panels. Model fit, variability explanation power and predictive power (mean square error (MSE), Akaike (AIC)/corrected Akaike (AICc) and Bayesian information criteria (BIC), respectively) suggest the model comprising zoometric LAS scores performs better than traditional zoometry. Optimized reduced models are able to capture variability for dairy-related zoometric traits without noticeable detrimental effects on model validity properties.