Using Multiple Correspondence Analysis to Evaluate Milking Parlour Performance

Modern milking parlours allow the automated collection of many data for each cow being milked that can potentially be used to monitor the overall performance of milking process. A group of 24 dairy farms located in Lombardy (Northern Italy) with RFID technology for electronic identification of cows, milk meters to measure individual milk yield and the same milking machine settings (42 kPa system vacuum, 60 cycles/min pulsator rate, and 60% pulsator ratio) were involved in the study. Cows milked/stall per hour [n], milk yield/stall per hour [kg], milking efficiency [%] were downloaded by the herd management software and combined with additional information such as number of milking stalls, stalls/milker, type of exit from milking parlour (rapid or conventional), and type of milking routine (full or partial). Relationships among variables were evaluated through a multiple correspondence analysis (MCA) that identified three main groups of related parameters. Large parlours (>32 milking stalls) were associated with a high number of stalls/milker (>15), low number of cows milked/stall per hour (<2.5), low milking efficiency (<31%), low milk yield/stall per hour (<39 kg). On the contrary, small parlours (<16 milking stalls) were associated with a low ratio stall/milker (<8), high performance in terms of cows milked and milk yield per stall per hour (>3.5 and > 52 kg) and milking efficiency (>39%). Parlours of medium size (16–32 milking stalls) were associated with intermediate performance. Results suggest that MCA can potentially be used to evaluate milking parlours performance.