The objective of this study was to characterize variability in both individual animal and pen data in data streams on a well-managed commercial dairy farm. Additional objectives were to assess the effect of outside events on the variability of data streams and utilize multi-variate statistical process control (SPC) to improve detection time of out-of-control data streams. A 1,400-cow dairy in Eden, Wisconsin, was equipped with milk meters and rumination/activity collars to record individual cow milk production, rumination, physical activity, and pen-based feed intake. Data were collected over a 3 mo period. Milk production was analyzed for out-of-control data points via the Shewhart procedure, and rumination, physical activity and feed intake were analyzed multivariately via the MVP procedures of SAS. On this dairy, milk production averaged 45.1 kg, with a standard deviation of 1.3 kg among days within a pen, 11.5 kg among individual cows within a pen, and 20.1 kg among days within individual cows. Rumination (min per day) averaged 441.6 min, with a standard deviation of 14.0 among days within a pen and 120.7 min among individual cows within a pen. Physical activity (measured in arbitrary units) averaged 489.2, with a standard deviation of 17.1 among days within a pen, 103.4 among individual cows within a pen. Feed intake (DMI kg/cow/d) averaged 26.6 kg, with a standard deviation of 1.8 kg among days within pens. Multi-variate SPC increased sensitivity when compared with individual single-variate SPC analyses. Out-of-control milk production values were preceded by a deviation from normal variance in the multi-variate analysis of rumination, physical activity, and feed intake. Thus, multi-variate SPC could be used as an early determinant of extreme variability in data streams on a commercial dairy.