Introduction/Objective: To prevent the development of infectious diseases in sheep fattening houses, antibiotics are systematically administered in food during the fattening start phase. This practice is challenged in France due to its potential role in the emergence of antimicrobial resistance in food‐producing animals. One alternative to the systematic delivery of medicated food for preventive purpose is the metaphylactic use of antibiotics consisting in a collective treatment as soon as the disease is detected in some animals. The collective delivery of antibiotics via drinking water is a promising strategy in this context, but the actually ingested doses are dependent on the individual drinking behaviours which could led to an under‐exposure of some animals to the antibiotics potentially responsible for therapeutic failures. In this study, to optimize this strategy, we characterized the individual drinking behaviours of fattening lambs and measured their impacts on the interindividual variations of the ingested doses. Materials and Methods: About 800 lambs were studied in a fattening house over 4 periods. For each period, the individual real‐time water consumptions of 200 lambs were recorded using water meters connected to drinking troughs that detected the RFID chips in the lamb ear tags. One hundred of these lambs received an antibiotic: sulfadimethoxine/trimethoprim (SMD/TMP, period 1), sulfadimethoxine (SDM, period 2), oxytetracycline or tilmicosin (OTC or TIL, period 3) or flumequine (FLU, period 4), and antibiotics were measured in plasma and in the drinking water. In parallel, the pharmacokinetic parameters of the antibiotics were determined in lambs during classical pharmacokinetic studies performed in the laboratory. Results and Conclusions: The average daily water consumption was surprisingly much lower than the expected one (2.5 l per day) and highly variable between lambs (up to 8 l). Concomitantly, antibiotic plasma concentrations varied between lambs from 4‐fold (SDM) to 10‐fold (TIL, FLU). A pharmaco‐statistical model was built by incorporating the time series of water consumption in a classical pharmacokinetic model which was able to predict plasma concentrations actually observed in fattening lambs. Such model will allow to predict for any antibiotic the impact of individual drinking behaviours on individual plasma exposures, and to propose improved dosage regimens adapted to the specificity of animal behaviours, whatever the animal species.