Microbial contamination of bovine raw milk often occurs at the farm. To acquire a deeper knowledge of the microbiota of farm tank milk, we studied milk from 45 farms situated in 2 geographical areas in Norway. Each farm was visited on 3 different occasions, with at least 2 wk between visits. We combined both bacterial cell counts and a sequence variant inference method of amplicon-based high-throughput sequencing to achieve a high-resolution overview of the microbiota in each sample. Compositional variation of the farm milk microbiota was shown in relation to the 2 areas, between the farms and between the sampling times. Despite the near constant level of bacteria enumerated in milk from each individual farm, the dominant microbiota differed significantly between the samplings. The predominant microbiota was dominated by spoilage genera, such as Pseudomonas and Bacillus, as well as the dairy fermentation genus Lactococcus and mastitis-causing organisms (Streptococcus). Analysis of the identified sequence variants within these genera showed that the populations of Pseudomonas and Lactococcus in milk had similar composition between the farms, but that Bacillus and, in particular, Streptococcus populations changed between collection days from the same farm and between farms and geographical areas. Furthermore, the levels and composition of Bacillus and Paenibacillus were different between the 2 geographical areas. The results presented here provide new insight into the farm milk microbiota and show that this microbiota is a dynamic community highly subject to variation.