Test-day (TD) models are used in most countries to perform national genetic evaluations for dairy cattle. The TD models estimate lactation curves and their changes as well as variation in populations. Although potentially useful, little attention has been given to the application of TD models for management pur- poses. The potential of the TD model for management use depends on its ability to describe within- or be- tween-herd variation that can be linked to specific management practices. The aim of this study was to estimate variance components for milk yield, milk component yields, and somatic cell score (SCS) of dairy cows in the Ragusa and Vicenza areas of Italy, such thatthemostrelevantsources ofvariationcanbeiden- tified for the development of management parameters. The available data set contained 1,080,637 TD records of42,817cowsin471herds.Variancecomponentswere estimated with a multilactation, random-regression, TD animal model by using the software adopted by NRS for the Dutch national genetic evaluation. The model comprised 5 fixed effects (region × parity × days in milk (DIM), parity × year of calving × season of calving × DIM, parity × age at calving × year of calving, parity × calving interval × stage of pregnancy, and year of test × calendar week of test) and random herd × test date, regressions forherd lactation curve (HCUR), the animal additive genetic effect, and the permanent environmental effect by using fourth-order Legendre polynomials. The HCUR variances for milk and pro- tein yields were highest around the time of peak yield (DIM 50 to 150), whereas for fat yield the HCUR vari- ance was relatively constant throughout first lactation and decreased following the peak around 40 to 90 DIM for lactations 2 and 3. For SCS, the HCUR variances were relatively small compared with the genetic, per- manent environmental, and residual variances. For