Milk progesterone can be used to closely monitor the fertility status of dairy cows as it allows identification of prolonged postpartum anoestrus, oestrus, cysts, embryonic mortality and pregnancy. Although milk progesterone can be measured quite reliably on farm, the measured concentration is subject to sampling errors, assay variability, calibration issues and varying concentrations of milk components, all influencing the determined concentration. In order to interpret the values correctly, and subsequently provide valuable fertility information to the farmers with a high sensitivity and specificity, an appropriate data processing and monitoring algorithm is required. In this study, we compared two algorithms to monitor fertility based on milk progesterone, using both on-farm and simulated data. The first algorithm is considered as the current state of the art and consists of a smoothing multiprocess Kalman filter combined with a fixed threshold for decision making (MPKF+T). The second algorithm, progesterone monitoring algorithm using synergistic control (PMASC), models the progesterone time series using two sigmoidal growth functions, after which decisions are made based on individual process control charts. The on-farm data included 1,843 oestrus alerts, and number and timing of these were compared for both algorithms. Because of the time-lag inherent to the smoother, alerts by MPKF were given 20±16 h later than the first out-of-control measurement identified by PMASC. The analysis on well-controlled simulated data, in which variability was included both at cycle and at measurement level, showed that PMASC was less dependent on the actual level of progesterone during oestrus, and that using a model-based indicator calculated from PMASC allowed for a reliable and consistent estimation of true luteolysis time, even when the sampling rate during luteolysis was decreased with 66%.