The timeline for a farm to adjust from parlor milking to robotic milking is often described as three days – three weeks – three months – three years. It gets easier as each of those markers is crossed. That makes sense because there are many factors involved and they don’t all adjust at the same rate. It takes three days to train the cows. Not every cow will be trained in three days. Some cows never learn. Some cows learn but they are too stubborn to conform. It takes three weeks to train the people. Yes, it is often easier to train cows than it is to train people. To be fair, people have a lot more to learn. It takes three months to develop routines and systems to make things go smoothly. Finally, it takes three years to select the best robot cows. Almost all cows can be milked in robots. Robot barns become more efficient when less efficient cows are removed from the herd and replaced with more efficient cows. Understanding the evolution from a parlor herd to a robot herd and identifying the most efficient robot cows can keep the adjustment on track.
The best robot cows combine four factors – milk production, milkability, milking speed and trips to the robot. Each of these factors can be measured, and the measurements can be combined to rank the best robot cows.
Milk production
Milk production is the easiest factor to measure because everyone has been measuring it for a long time. Robot software and herd management software include a variety of metrics. In most markets, it is best to include components in the production measurement if available. A standardized complete lactation projection is more useful than a daily or seven-day average milk weight. Age adjustments are important. If age adjustments are not available, consider making your comparisons within lactation groups.
Milkability
Milkability is measured in terms of incomplete or failed milkings. Ultimately, an incomplete or failed milking occurs when a quarter produces less than the expected yield of milk. This can happen when a teat is not found, a teat cup is kicked off, a quarter does not milk out completely or a quarter milks out completely but produces less than expected. Failures and incompletes can be caused by machine factors like dirty cameras or delayed maintenance, or they can be caused by cow factors like kicking or poor udder conformation. Occasional incomplete or failed milkings can be expected. They occur in parlors too, but we are less likely to know about them. Cows that have frequent incomplete or failed milkings take longer to milk, produce less milk and are at higher risk for mastitis.
Milking speed
The amount of time a cow spends in the robot depends on how long it takes to prepare her for milking, how fast milk flows from the teat, and how much milk she produces. A cow can be slow because she dances and kicks while cups are attaching, because her teats restrict milk flow or because she has a lot of milk. The lowest-producing cows usually have the shortest milking duration, so shorter isn’t necessarily better. Box time and milking duration both include prep time and milking time but don’t reflect quantity. Average flow is a better measure of milking speed because it combines time and volume. Yield per box time and harvest flow are the preferred comprehensive measures because they combine prep time, milking rate and total production. Those metrics don’t penalize high-producing cows just because they take longer or give low-producing cows a pass just for being fast.
Robot visits
Trips to the robot may be the hardest thing to measure and predict. We need an objective measure of how much the cow moves around the barn and how many opportunities she gives us to milk her. The number of milkings is not the best measure because the software does not differentiate between a fetched milking and a voluntary milking. In free-flow barns, if a cow comes to the robot but is not milked because she does not have milking permission, that is called a rejection or a refusal. Rejections are almost always voluntary. Cows with more rejections move around the barn more and give the robot more opportunities to milk them. Gate passages provide the same information in a guided-flow barn. If a cow does not have milking permission, the sort gate will send her to the feedbunk. The more times she goes to feed, the more opportunities we have to milk her. A cow that gives us more opportunities to milk her is a better robot cow. Rejections, refusals and gate passages indicate how many chances a cow gives us to milk her.
Ranking your cows
You can use the same principles used in Net Merit and TPI indexes to combine production, milkability, milking speed and trips to the robot in an index to pick your best robot cows. The actual number is arbitrary because it is only for comparison. Choose appropriate factors to apply to production, milk speed, milkability and trips to the robot to generate a single score for each cow. Set the factor to account for both the magnitude of the value and the importance of each trait.
Table 1 shows a simple comparison of two cows. Bessie has a higher 305-day predicted production and an index value of 3,227. Elsie’s index value of 3,418 is higher than Bessie’s even though Bessie makes more milk because Elsie milks faster, with fewer failures and more refusals. To place more emphasis on production, increase the production factor from .1 to .2. Then Bessie ranks higher with an index value of 6,427 compared to Elsie’s 6,218 in spite of Elsie’s other advantages.
Dairy farmers talk about the advantages of having a parlor and robots because cows that don’t work in the robots can be sent to the parlor. They are not wrong, but it is also true that every robot farmer can choose to milk their best robot cows and find other homes for the rest. Balanced voluntary culling based on production, robot milkability, milking speed and trips to the robot will result in the right robot cows and the most efficient robot barn.
