Smallholder dairy farms (SDF) producers are being challenged by market demands combined with competition and limited access to productive and market inputs, contributing for reducing its participation in the dairy activity. Thus a sustainability assessment for SDF development is necessary, opening the opportunity for recognizing the diversity between production systems. Farm typologies are often used for this purpose. The objective was to develop a typological analysis of the smallholder dairy farms in the Brazilian semiarid region and evaluate characteristics of raw milk and water used in milking. A semi-structured questionnaire containing social, physical, livestock, herd management, technological, productive, and economic indicators with 49 variables was applied in 28 SDF. Of these, 23 were selected, and collections of milk and water used for milking management were performed. Three multivariate analysis techniques were used: exploratory factor analysis, hierarchical cluster analysis, and canonical discriminant analysis. Three production systems have been identified: Conventional system: Mature illiterate and rural farmers, small herds and farms, low technological level, less intensive management, and low milk production with in nature marketing (15 farms; 53.6%); Traditional system: Mature farmers with basic education, farms managed with hired labor, large herds and areas, extensive management, low technological level, high production of milk for the manufacture of cheese (6 farms; 21.43%); and Emerging system: Young farmers with a high level of education, small herds and farms, intensive management, intermediate technological level, and high milk production and productivity with in nature marketing (7 farms; 25.0%). The social, physical, livestock, herd management and technological, and productive indicators showed discriminatory power (P < 0.05) to differentiate the typologies. Milk and water traits were similar (P > 0.05) among the production systems. The average values of the milk composition are in the threshold of the Brazilian legislation, unlike the somatic cell count (CSS) and standard plate count (SPC). Antibiotic residue was not found in milk from any SDF.The water used in milking management was characterized by high counts of aerobic mesophilic bacteria and total and thermotolerant coliforms, hard water, and neutral pH. For the continuity of SDF in the Brazilian semiarid, it is essential (i) participation of young people in the sector; (ii) feed planning for the dry period; (iii) herd health calendar according to legislation; (iv) water treatment (chlorination) and use of alkaline detergent for use in milking management, and (v) formal marketing of milk and its derivatives. Finally, our results can be used to inform and support public decision-makers responsible for the formulation and implementation of livestock development policy in the Brazilian semiarid region and being useful for other drylands.