€9.30 per cow per year: what a feed-pushing robot actually changes in the barn

Feed a total mixed ration once or twice a day and you build a mound that, after the main feeding bout, lies too far from the feed fence for the cows to reach. On top of that they sort it and push it away themselves. So it gets pushed back: conventionally with a yard tractor and a sweeping tyre, a blade or a feed auger, in a fixed routine, six to twelve times a day.

LFA MV measured what that costs, and the number is small: experienced tractor operators need around 3.5 labour-minutes to push the feed on both sides of a feed table, including walking, climbing in and out, and attaching and detaching the implement. This is not a task anyone cuts for lack of time.

It gets skipped all the same — not because it is long, but because it looks unimportant. The institute describes the mechanism plainly: a calving, a breakdown in the milking parlour or the general labour shortage interrupts the routine, “and one push or another is missed”. Many farms work with changing staff, and whether the evening push actually happened once the herd manager has gone home is something nobody knows the next morning.

None of this would matter if it made no difference. LFA MV cites two findings that together carry the case for the robot. Nydegger et al. (2005) examined the relationship between pushing frequency and performance and found that dry-matter intake and milk yield are clearly higher at six or more pushes a day than at four. And according to Bonsels (2019), cows take almost half of the feed they need during the night.

The second point is the one a daytime routine cannot cover. The pushes most likely to be missed are the evening and night ones — exactly the hours in which half the feed is eaten. And the animals hit first are the ones that reach the table last anyway: LFA names low-ranking and first-lactation cows.

A robot changes nothing technical about this. It changes the fact that the question “was the feed pushed at 10 p.m.” no longer has to be asked.

LFA MV built the comparison on a working farm that had run both methods: 330 dairy cows, one barn with one main feed table. The assumptions are laid out openly and transfer to any farm.

• Conventional: tractor at €30,000, eight years of service life, 3.5 percent interest, 3.5 labour-minutes per push, six pushes a day, €20 per labour-hour, 4 percent of purchase price for repairs, 12 litres of diesel an hour. The tractor works around 14 hours a day on the farm and only 3 percent of that time goes to feed pushing — only that share is charged.
• Automatic: robot at €20,000, 7 percent depreciation, 3.5 percent interest, 3 percent repairs, 800 kilowatt-hours of electricity a year at €0.25, and 30 labour-minutes a week.

The result: €8.38 per cow per year for six pushes a day with the tractor, €9.30 for the robot. LFA calls the robot “only marginally more expensive”, adding that, set up accordingly, it pushes more than six times a day, “and with relative certainty at night as well”.

The 30 minutes a week in the robot line deserve a second look. They cover maintenance and cleaning and “getting it going again after a stoppage”. The institute, in other words, prices in a robot that occasionally stops and has to be nudged. That is more realistic than any availability figure on a datasheet, and it is why the objection “the tractor is there anyway” does not get the last word: LFA answers it with the observation that the cows are supplied at every hour of the day “provided the number of breakdowns stays within limits”. What such a stoppage costs an operation and who resolves it is the subject of progress detection as an operating metric for service robots generally.

LFA asked six suppliers for their technical data, and the systems differ first in what they do to the feed.

The roller mixes the ration through on every pass as it pushes it towards the feed fence. That is how the Butler from Wasserbauer and the OptiDuo from DeLaval work. The drum pushes the feed compactly to the fence without loosening it: Moov 2.0 from JOZ, Stallboy feed from Hetwin, Faro from Schauer, Juno from Lely.

Both designs push dynamically, which according to LFA matters when feed of differing structure — hay, straw, fresh grass, TMR — and in varying quantities lies on the table. Some systems also offer concentrate or liquid attractant dosing (Wasserbauer, Hetwin, DeLaval, Faro).

For the decision, the mixing question is not a side issue. Anyone feeding a TMR designed against sorting may not want it mixed again at the table. Anyone feeding forage and concentrate separately may want exactly that. This is a feeding decision, not a robot decision, and it should be taken before the quotation.

The part of the comparison that brochures cover most briefly is the one that matters most for commissioning: what the machine steers by. LFA lists three methods.

• Magnets sunk into the floor — Butler Gold, Stallboy feed. The track is fixed; the Butler offsets it itself depending on the amount of feed, and for very large quantities a second magnet track is recommended.
• Induction sensors — Moov 2.0.
• Ultrasound against the feed fence or neck rail — Juno. The robot feels its way along the fence. Outside the barn, where there is no fence, a metal rail has to be bolted on so that it can orient itself there too.

That is the line a project hangs on: anyone who wants to send the robot across the yard into a second barn builds either a magnet track or a rail, depending on the system. All the listed robots can serve several barns according to LFA; the manufacturers quote 500 to 1,000 metres of maximum travel per round, and gradient capability varies between 10 and 15 percent by manufacturer. The route between barns is therefore a planning quantity with a length, a gradient and a surface, not a given.

Then there is the charging station. Among its six questions before selection, LFA explicitly asks whether the station can sensibly be integrated into the barn and whether it needs protection from rain and sun — and how far the robot can travel before it has to charge. That site configuration is an asset of the building rather than of the robot holds in the barn just as it does in the office building described in the article on robot markers.

LFA's most interesting question is about the failure case, and the manufacturers' answers differ. Schauer's Faro detects through overcurrent sensing that it cannot get through and starts an evasive manoeuvre. The Butler shifts its track. The Moov 2.0 moves up to 120 kilograms according to its maker; if there is more, it drives around the pile and carries on a little further along.

For the farm this means: the robot does not replace feeding out, and it does not clear up feeding-out mistakes. The practical reports agrarheute collected from a farmer survey on 5 January 2025 turn on exactly this point: users of a Juno 100 report oversized feed quantities and repeated faults; a Wasserbauer user from 2018 to 2020 reports large quantities, software errors and broken displays. On the other side stands a farm whose Juno runs “pretty well” after 14 years with minimal repair costs, and a Moov user who has feed pushed hourly for 60 cows and “rarely has problems”.

The spread is not a contradiction. It describes what the 30 minutes a week in the LFA calculation mean: at a table with even feeding-out they go unused; at a table with mounds they become hours.

The base prices LFA quotes sit close together: Wasserbauer's Butler Gold at €19,500, Lely's Juno at around €22,500 including freight, installation and first programming, JOZ's Moov 2.0 from €17,450, or €1,550 more with lifter and door opener. At those margins the price decides nothing.

What does decide is stated by the institute with unusual bluntness: “The most important question of all, however, is whether a dealer and a reliable service team are available for the chosen favourite.” The farmers in the agrarheute survey reached the same conclusion independently: proximity of service was the primary purchasing factor, ahead of any technical specification.

LFA's second sentence belongs beside it: once robotics has been invested in, “it has to be used, which means familiarising yourself with the technology”. A pushing robot that stands in the corner for two weeks after its first stoppage because nobody knows how to restart it is the most expensive variant in the comparison — and the only one that appears in no table.

Three caveats, so that the numbers are not asked to carry more than they can.

The calculation is for one farm. 330 cows, one main feed table, an existing tractor. On a farm with 80 cows the robot is spread over fewer animals and the tractor over fewer pushes; LFA lays the assumptions open so that they can be replaced, not so that the result can be copied.

Manure collection is the other round in the barn, and no such comparison exists for it. Collecting and scraping robots for alleys are offered by the same manufacturers and run a fixed loop on the same principle. We know of no independent cost calculation for them of the kind LFA produced for feed pushing; anyone automating both rounds prices the second one themselves.

The manufacturer data in the overview is manufacturer data. Range, gradient capability and prices come from LFA's survey of the six suppliers, not from measurement. What was measured is the 3.5 minutes at the tractor and the cost line — and those suffice for the decision that is actually pending.