June 25, 2019

Arla's new artificial intelligence tool predicts cows' milk output


Arla Foods has developed a new artificial intelligence (AI) tool to better predict their milk intake from farmer-owners.

This means that 200 million kilos of milk can now be utilised better each year making Arla's value chain even more sustainable, the dairy co-op said.

Every year, Arla collects around 13 billion kilos of milk from its 10,300 farmer-owners across Northern Europe. As such, the cooperative is keeping a close eye on emerging technologies that can make the dairy production even more efficient.

Arla's new AI-based tool will predict how much milk 1.5 million cows will produce in the future, according to the co-op.

Previously, this type of forecast would take days to create manually from piles of Excel data sheets. With the new tool, it will only takes a few hours to produce a forecast that is 1.4% more accurate.

"The better we are at predicting what our milk intake will be, the better we can plan and optimise our entire value chain, which both improves profitability for our farmer-owners and drives sustainability. The new AI tool provides us with an insight into our supply of milk that we have never had before," said Michael Bøgh Linde Vinther, Arla's director of global milk planning.

The improved milk intake forecast means that 200 million kilos of milk can be utilised better each year.

"By using AI technology, Arla is able to create the milk intake forecast from a much richer data foundation. It includes things like seasonal changes, the number of farmers converting to new milk types, the farmers' geographical characteristics as well as how much milk they produce on a daily basis," Arla explained.

"This means that Michael (Vinther) and his colleagues can handle the milk planning at a more detailed level than they could before."

Vinther said: "We are now able to make important strategic decisions on a more informed basis. The data has become more valid as it is now formalised in a 'bulletproof' system rather than based on individual knowledge. It's amazing to see how this new technology is able to optimise and improve a (up until now) very time-consuming task."

For example, it is now possible to make the distinction between how much milk should be collected from farmers in North Germany and West Germany three to five months ahead of time.

"This kind of knowledge is very valuable, as it makes us able to plan and adjust the number of Arla trucks travelling across the country. In that way, we can both reduce cost and save the environment for unnecessary CO2 emissions," Vinther explained.

The new milk intake forecasting tool is implemented in all Arla's markets across Europe, including Denmark, Germany, Sweden, the UK, Belgium, Luxembourg and the Netherlands.

- Arla