Robotic milking machines can milk more than 99% of all dairy cows

At present, 10% of cows in the UK are milked through robots, and Lely now have 3000 installations in the UK, which is estimated to double to 6000 by 2028.

But while robotic milking may appear to be a totally different concept, such machines are equally hospitable to the same cows put through a parlour. As it is, less than 1% of cows are unable be milked through robots when moving from a conventional milking system. However, improving certain traits to fine-tune milking efficiency can enhance milking speed through quicker attachment and milking times.

"Don’t throw the baby out with the bath water. Carry on using your current breeding criteria, because it has done you well until now, and a robot can milk 99.5% of your cows," said Ben Nottage, Dairy XL account manager for Lely Atlantic, speaking at the Cattle Breeders Conference in Telford.

However, he said making small incremental ‘tweaks’ to certain traits, such as milking speed, udder depth and cow stature, could equate to big gains and make robotic systems more efficient.

Mr Nottage said the more farmers intervened with cows, the less milk they made. He added that focusing on newer genomic traits for mastitis and lameness resistance could also help reduce the number of cows that need collecting, saving time. “None of these traits are do or die, but when you start adding them up, they make a big difference.”

Robots work on most large units too, but correct shed design is fundamental to capitalise on the labour savings that can be made when installing robots

One Lely customer in the USA is now milking 900 cows, using 12 robots and just three labour units.

“The real difference with larger herds is they don’t build to milk cows anymore; they build to manage cows. If you build to manage, that is where the true labour savings and extra cow longevity comes in."

He said using segregation gates – that send cows that need to be checked to a management pen – were key to avoiding staff having to traipse around the shed looking for animals.

“The minute you dispense labour around the whole shed, that is where your time goes,” he explained.

For this reason, he said it is best for robots to be sited in the middle of sheds rather than on the outside parameters. Post-installation, he said, the first step is to re-assess how labour is assigned by working out what tasks need doing daily, monthly and annually, and planning new rotas around these tasks.

“Staff will probably come in, do one shift and go home,” he explained. Labour can be refined over the next 12-24 months, he added.

“It is an ongoing process. We can train the cow to use the robot in 2-3 days, but you will have to manage staff, and your time, differently.”

Three main principles of good shed design:

1. Give cows more time to be cows – to lay down and eat with minimal interference.

2. Organise labour around the cow, not around people – think about how many cow ‘touches’ are required (calving, drying off, etc) and design your routines around this.

3. Focus on processes not on outcome – if you must deal with a cow, how can you do it in an easy way? Consider how gates and collection areas can be used alongside designated areas for drying off, breeding and pregnancy diagnosis.

Breeding trait advice:

1. Teat length

Robots are better at milking cows with shorter teats than parlours, provided the laser can detect them. This is because clusters tend to slip in conventional parlours, whereas the robot has an advantage, because there is no weight on the claw.

But improved teat placement and length can speed up attachment time, improving overall milking efficiency.

Do not select extreme bulls (below -1).

Heritability: 29%

2. Udder depth and stature

Robots can find it hard to detect very shallow udders on extreme stature cows.

“If you have been breeding for type and selecting plus 3 bulls for stature and udder depth, you end up with really tall cows and really shallow udders and the robot can’t find the udders.”

Avoid extreme scores (+3 or +4) for body depth, in combination with stature and extreme shallow udders.

Heritability: 35%

3. Milking speed

Anecdotal evidence suggests cows that milk quickly through parlours often have high cell counts, but this trend is not replicated in robotic herds because variable vacuum removes milk away from the teat quickly and cows are not waiting to be milked in collecting yards.

Milking speed also varies hugely within a herd. For example, data from one robotic herd shows three cows that gave the same milk had a range of milking times, with the best milking out in 3.15 minutes and the worst taking 7.28 minutes. There is huge scope to improve this at a herd level by selecting bulls that are positive for milking speed.

Heritability: 21%

4. Wellness traits and temperament

Mastitis and lameness in robotic herds are a challenge to manage because those cows must be segregated and treated. You can make labour and cost savings by reducing cases.

Start genomic testing heifers and assessing their resistance to disease and use this data alongside wellness traits for bulls.

Bad temperament cows can be challenging, too.

“Robot cows are instantly quieter, so these poor disposition cows show up quickly in a robotic system.

“The robot arm will far outlive the cow if she wants to kick it, but there is more to temperament than just being in the box.”

Bad temperament cows can be more difficult to train, collect and will move in the box affecting attachment time.

Take-home messages:

• Look at herd data and filter out the bottom 10% of the herd based on milking speed and attachment time

• When you have filtered your bull choice to your final 30, using current breeding criteria, start to look at 1-2 other traits

• If you want to install robots, start talking to your breeding adviser beforehand to make small tweaks to your breeding programme

• Aim for 0% of cows that need fetching. You will never get to that, because there will be a cow that has a sore foot, for example, but you should target that.