This large scale study investigated the spatial variability of Pb, As, and Cd contents in raw milk within and between the 10 main milk producing areas in China. A total of 997 raw milk samples were analysed by inductively coupled plasma mass spectrometry (ICP-MS). Mean values of Pb, As, and Cd in milk were 1.75 μg/L, 0.31 μg/L, and 0.05 μg/L, respectively. The highest level of Pb and As was present in area C, and Cd was highest in area J. The standard deviation suggested a higher heterogeneity of milk heavy metal contamination within area than between areas. Levels of Pb, As, and Cd showed significant differences between studied areas. The estimated root mean squared standardised error obtained by the cross-validation suggested a differentiated quality of Pb, As, and Cd modelling between areas: the predictions obtained were sometimes overestimated or underestimated. These results can be used to define a more appropriate sampling procedure for heavy metal contaminate distribution in raw milk for improved future control of milk contamination by heavy metals in the studied areas. The significant positive correlations between concentrations of Pb-Cd, As-Cd, and Pb-As were observed in nine, six and five areas, respectively. No significant negative correlations were observed. The observed variability of correlation values suggested a different pollution source for Pb, As, and Cd in milk between areas. Further studies are required to clarify the relationships between the contamination of raw milk by heavy metals and the herd environment.