Robotic Milking: Farm test - the cow's time spent in the milking robot

Which milking robot manufacturer to choose is for many a careful evaluation based on several factors: cost, annual running costs, capacity, reliability and service options, to mention some. All parameters are important when making this kind of investment. The total time the cow stays in the milking robot affects the number of cows that can pass through the milking robot, affecting the utilization of the milking robot’s capacity. An optimal capacity utilization is important to ensure a good economy with milking robots.

Milking robot capacity utilization is important for the results of the investment, and this kind of test is not the only one for testing capacity utilization; number of kg milk per day is also an important parameter. FarmTest visited 34 dairy farmers with the objective to measure the cows’ visit time in the milking robot and assess the quality of the teat cleaning.

Chart 1. DeLaval had shorter process time per kg milk.

Seconds per kg milk from the start of the cleaning until the last cup is removed.

Table 1: Registered cell count from Performance Control on the tested cows.

Average cell count from the Performance Control


tested cows

tested herds
DeLaval VMS 154.000 286.000
Lely A3 143.000 320.000
Merlin 80.000 253.000
SAC Futureline 151.000 271.000

 Chart 2: National average somatic cell count on July 6th 2011

Compared to the national average (Chart 2) which is tank cell count, there is nothing to suggest that the fast milking of the tested DeLaval cows influences cell counts. The reason for this is that the cell counts in that case would have to be at least above the average cell count for milking robots.The tank cell counts were typically 30-40000 under the Performance Control’s cell counts, as fresh cows, sick cows and dry cows were not milked in the tank.

Double cleaning, which means that the teat cleaning runs two times after each other can be an opportunity to better stimulate the cow for milking.

A farmer with Merlin milking robots runs a double teat cleaning. He has learned that the double cleaning gives better let-down of milk and hence faster and more gentle milking. A farmer with Lely milking robots expressed the same opinion, but the double cleaning was not used on his herd.

See also section. 2.1.

Image 1: Closely set teats are a challenge to the milking robot 

Be aware of 'time thieves’

The cow in Image 1 has a flaccid udder and very closely set rear teats.

This cow is approx. 51 weeks from last calving, which was its first calving, and will calve again in approx. 16 weeks.

The cow's daily yield is around the time of testing. 12.0 kg, and falls to approx. 6.0 kg before drying off (figures from the Performance Control). The cow spent about 8 minutes on cleaning and attachment.

When daily yield is taken into account the cow in Image 1 is a time thief in the milking robot. Focus on the cow's lactation period, so that the cow becomes pregnant earlier, can help in shortening the preparation time, since a firmer udders can give a better teat placement. This cow was at the first Performance test after the second calving yielding 35.0 kg, which could mean she had a more appropriate teat placement than what we can see in these pictures, and thus a quicker preparation.

Avoiding 'elephant races'

Mads Nielsen from Jutland Agricultural Advisory informs that the cow flow through the milking robots are just as important as the milk flow, and compares 'time thieves’ with the familiar concept from the highway: 'elephant races'.

After a slow-milking cow has pushed herself in front of the other cows, and walks in to the milking robot, there will be queue if it takes too long before this cow is milked. Some cows will give up and go back to the cubicles.

Is there a focus on robotic capacity?

Farmers' opinions on their experiences and reflections on the utilizationof milking robot capacity:

Farmers with Lely milking robots:

• More than 67-68 cows per milking robot requires more management

• Milking robot utilization belongs to the breeding management.

Farmers with DeLaval milking robots:

• Have tried 133 cows to two milking robots, which led to too many cows to fetch.

• More than 65 cows per robot is too many, number of milkings go down

• Have chosen guided cow traffic to have fewer rejections

• 65 cows per robot works well, but I also fetch cows four times a day

• More than 65 cows per robot leads to many stressed cows and thus fewer milkings.

Farmer with SAC milking robots:

• Have more focus on teat placement than milk flow.

Mads Nielsen from Jutland Agricultural Advisory offers the following recommendations to farmers with milking robots:

• Technically, milking robots must be efficient, which means that the manufacturer’s recommended service routines should be followed and there should be regular checks for leaks.

• If the recommended service routines are followed, it will primarily be biological factors that cause problems, which can be poor teat placement, nervous cows etc.

• It is important to focus on breeding. Go for the cows with high yield and high milk flow - these traits generally come together.

• At start up of robot milking it is important to have good service routines from day one. Bad habits are hard to change.

One farmer said: "Robot capacity is slipping somewhat in the background due to the satisfaction that the work is less tiring now, but milking cows that are time thieves must go. Cows have to match the robot’s requirements for teat placement".

Consider using the check list in Appendix 1 to be aware of capacity utilization.

2. Results

2.1 Teat cleaning

Teat cleaning is done, depending on the brand, with rotating brushes or in a separate milking cup called preparation cup, which in addition to cleaning the teat also foremilks the cow. The cleaning can be adjusted according to individual needs, for time as well as for number of cleanings.

The measuring of the teat cleaning is recorded from the time when the brushes or preparation cup start searching for the teats, until the last teat is washed. This means that the time used for searching for the teats is included in the measurement. Teat placement is critical for how much time is spent on cleaning. The cleaning process is the same regardless of teat placement, but the search time can vary a lot.

Cows with three teats and cows with double teat cleaning are excluded in the comparison, but for the Merlin robots there is a comparison made between single/double teat cleaning, since for one of the Merlin herds it was reported that double cleaning gave a shorter milking time.

Image 2: Cleaning brushes on a Merlin from Fullwood.

Image 3 SAC robot ready for teat cleaning

Charts 3 and 4: Time measuring for the start-to-end teat cleaning.

Average cleaning time in seconds Lely and Merlin

Average cleaning time in seconds for DeLaval and SAC

Image 4: DeLaval cleaning cup doing a nice job here.

Image 5: Teat cleaning with Merlin rotating brushes

Image 6: Teats that bend from the brushes are not properly cleaned.

Lely and Merlin milking robots clean the teats with rotating brushes, while SAC and DeLaval have preparing cups. Time comparison for teat cleaning is not possible since DeLaval and SAC not only clean but also foremilk the teat in this process. As shown in Charts 3 and 4 Merlin and SAC Futureline are the fastest in their category.

The quality of the teat cleaning is assessed visually in the Farm test. The quality is rated by the cleanliness of the teat on a scale from 1 to 5, where 1  is the lowest. (Chart 5) In addition to the cleaning quality, the lowest rate was given when at least one teat was not cleaned, since it must be expected as a minimum that each teat should be cleaned.

In some cases the cleaning was set off because the robot spent too much time searching for the teats. Some applied double cleaning the first week after calving, and others used double cleaning consistently.

A farmer with Merlin milking robots experienced that a double cleaning led to better milk let-down and a faster and more gentle milking.

Merlin milking robots are compared for single and double cleaning with the following results:

  Single cleaning
25 measurings
Double cleaning
8 measurings
Cleaning time 23,3 sec 36,9 sec 16,6 sec
Milking time/ kg milk 35,8 sec 30,7 sec 5,1 sec

If, for example, the herd performance is 30 kg milk/day at 2.5 milkings, the cow’s box time is reduced 2 minutes/day when applying double cleaning. The results are taken from the actual measuring. Possible differences in the cows’ breeding index for milking time is not taken into account.

Image 7: DeLaval in the cleaning process - but without having found the teat.

Image 8: DeLaval cleaning cup has completed cleaning the front teats. There is a clear difference between cleaned front teats and not cleaned rear teats.

If a teat is difficult to find in the cleaning process, there is a risk that this teat will not be cleaned, or that another teat will be cleaned twice.

Around 80% of the milking robots perform the teat cleaning to a grade 5 (Chart 6), which can indicate that the number of cows with uncomplicated teats and udders is high. Even if DeLaval shows high incidence of cleaning with score 1, they are nevertheless at the top in scoring 5, for ‘best quality of cleaning’, which indicates that if the cleaning cup finds the teat it cleans it very well.

FarmTest No. 82, Cleaning Procedures in AMS (Kromann 2011) has investigated 28 herds and a total of 488 cows, looking at how different robot manufacturers clean the teats, and how good they are cleaned.

Chart 5: Teat cleaning in cups seemingly results in more teats not being cleaned.

Percentage with grade 1 for teat cleaning -  where at least one teat is not cleaned

Chart 6: The percentage of grade 5 was similar for all milking robots.

Percent with grade 5 - for best cleaning quality

Table 2: Percentage of 1-5 ratings in tested cows.

  Grade 1 Grade 2 Grade 3 Grade 4 Grade 5
Lely A3 2 2 7 7 82
VMS 18       82
SAC 20   2   78
Merlin 6   12   82

Image 9: Good teat placement makes attachment easier.

Image 10: Closely set teats are not robot-friendly.

Image 11: Cleaning the laser head can increase the search speed

Image 12: A dirty udder can be the cause of errors in attachment and lead to extended attachment time.

Image 13: Cows that lie on the floors place greater demands on the teat cleaning

Image 14 Tool for removing udder hair.

2.2 Attaching milking cups

The time of milking cup attachment is measured from the start of the search for the first teat until the fourth milking cup is attached.

As with cleaning, teat placement is crucial for how much time is spent on attachment and some farmers expressed that they will focus more on breeding 'robot cows', which means that the udder and teats must be placed so that cleaning and attachment can be done without too much time wasted on searching for teats.

If the udder is dirty or hairs very long, this may be the cause for longer attachment times since the robot reacts for things sticking out on the udder. Impurities such as straw or very long hair can be perceived as a teat by the laser. The same applies to long-haired and dirty tail ends. If a long milking time is recorded for a cow, the obvious thing to do first is to check udder cleanliness and possibly cut or burn udder hairs and cut the tail.

Burning of udders is made with cold burning, which is a gas burner with 'cold' flame. The 'cold' flame of only 300° C occurs when gas is burned with too little oxygen.

Improper attachments are time-consuming, Image 16 shows a milking cup attached to the udder and not the teat. Attachment as shown in the image is on a dirty udder, which also points out another problem than time waste, it is also a hygiene problem and measures need to be taken to prevent this from happening.

Improper attachment as shown in Image 17, where the rear cups are attached to the front teats, indicates it might be time to perform a cleaning of the laser and if necessary calibrate the equipment.

Chart 7: Attatchment time from when the milking cup starts searching for the first teat until the fourth cup is attached.

Average attachment time in seconds

Image 15: Merlin from Fullwood is fastest at attaching.

Image 16: Improper attachment extends milking time.

Image 17: Improper attachment.

2.3 Preparation

The total preparation includes the previously mentioned cleaning and attachment, but also fore-milking. In fact, the preparation is the stimulation for the cow to let down the milk. It is essential for an optimal milking, seen from the cow's point of view, that there is a maximum milk flow from all four glands quickly after the cluster is attached to the cow (veterinarian Michael Farre, 2010).

In AMS-stalls the cows who voluntarily go to the milking robot are often stimulated by this process itself. It is not entirely clear how general this effect is, therefore a good stimulus in the milking robot is preferable.

The cow in images 18 and 19 has a flaccid udder and very closely set rear teats. The cow is approx. 51 weeks from last calving, which was its first calving, and should calve again next in approx. 16 weeks. The cow's daily output is around 12.0 kg at the time of testing, and drops to approx. 6.0 kg before drying off (figures from the Performance Control).

The cow used approx. 8 min. for cleaning and attachment.

Image 18: Closely set teats is a challenge to the milking robot.

Image 19: Cleaning and attachment to the cow in images18 and 19 took approx. 8 minutes.

Charts 8 and 9: Average times for cleaning and attachment.

Average attachment time Lely and Merlin

Average attachment time DeLaval and SAC

Looking at daily milk yield the cow in Image18 is a ‘time-thief’. Focus on the cow's lactation period - that the cow becomes pregnant earlier, may shorten the preparation time since a firmer udder will give a better teat placement. This cow yielded 35.0 kg at first performance control after second calving, which maybe meant that she had a better teat placement than in these images and therefore a faster preparation.

The milking robots from the four different manufacturers are not directly comparable on the average times for cleaning and attachment, since DeLaval and SAC also fore-milk in the cleaning process, while Lely and Merlin fore-milk in the milking process.

Merlin and SAC Futureline are the fastest in their respective category.

Image 20: Merlin is fastest in preparation compared to Lely (Cleaning and attachment).

2.4 Milking time

Milking time depends on kg milk, milk flow and whether the cow is calm during milking. A calm cow in the milking robot provides good milking, but is also important for the milking robot avoiding damages and system crashes, but also so that the cow doesn’t get injured. A calm cow is attributed to the skills of the operator, and it is the general impression from this test that the operators have great knowledge of the cows’ behavior.

Chart 10: Recorded average milking times.

Milking in seconds from start of attachment until finished milking

Chart 11: DeLaval VMS had the shortest milking time in the test.

Seconds per kg milk from start of attachment until finished milking.

2.5 Process time (Preparation and milking time)

The total time for preparation (cleaning and attachment) and milking time indicates the actual processing time the cow stays in the milking robot. After milking the teats are sprayed with a disinfectant, a process of approx. a few seconds. The disinfection process is not measured separately.

Process time (cleaning, fitting and milking) see Chart 12, shall be seen in relation to the milk yield. Chart 13 shows seconds per kg milk.

Table 3 Variations in the measured time and quantities of milk.

Variations in measured milking times in seconds from start of cleaning until milking is finished and variations in milking time/kg milk
  Shortest and longest milking time (sec) Shortest and longest milking time/kg milk (sec)
Lely A3 194 774 15,7 71,6
DeLaval VMS 233 860 23,0 132,3
Merlin 221 681 19,9 97,9
SAC Futurline 230 961 23,4 123,8

Chart 12 Measured process time per cow (cleaning, attachment and milking).

Milking in seconds from start of attachment until finished milking

Chart 13 DeLaval VMS has the lowest time per kg milk.

Seconds per kg milk from start of attachment until finished milking

2.6 Box times

The box time reflects the time spent, from the cow's entering the milking robot, until the cow has left it again. There has been no comparison of the box times since there can be conditions that can be attributed to the cows' behavior rather than the milking robot, that affect the box times.

Even if a cow gets out quickly and the next cow enters quickly, it might be the case that the milking robot is not ready, which could be due to cleaning of the laser or emptying the milk receiver, so the milking robot might not be ready to perform the teat cleaning. The time from when the cow is inside the milking robot, until the cleaning starts is of course important for the robot capacity.

As can be seen in Chart 14, SAC Futureline has a high time consumption, which can be attributed to the fact that the milking robot does not start the cleaning until the milk receiver has been emptied from the previous milking. The time from when the cow is fully milked until it has left the milking robot and the gate has closed behind it, includes teat spraying and how fast the cow leaves the milking robot.

If there is concentrate left in the trough the cow will not exit, so a calibration of the feed allocation is important.

SAC Futureline has mounted metal weights that pushes the cow after a few seconds. These seem to work according to plan, since, as can be seen in Chart 15, the SAC Futureline has the shortest time from when the cow is milked until it has left the milking robot.

Chart 14: The time from the gate is closed until the cleaning begins.

Chart 15: Time spent on exiting is shortest in the SAC Futureline.

Seconds from finished milking until the cow is out from the robot

2.7 Farmers' focus on robotic capacity

The farmers' attitude to the utilization of milking robot capacity was for the majority of respondents that the cows and the milking robot should operate without too much assistance from the farmer. Some farmers have experienced that there are subtle limits to how many cows can be miles per milking robot, and therefore choose not to challenge that number more than necessary, since it would only lead to extra work with cows that need to be fetched because they do not go to be milked.

As one farmer said: "Robot capacity is slipping a little into the background due to the satisfaction that there is less tiring work, but milking cows that are 'time-thieves', must go. Cows must have the teat placement that the robot requires".

The following is the farmers' opinion on their experiences and reflections with utilization of milking robot capacity:

Farmers with Lely milking robots:

  • More than 67-68 cows per milking robot requires more management
  • Milking robot utilization is part of the breeding work.

Farmers with DeLaval milking robots:

  • Have tried 133 cows for 2 robots, which led to many fetch cows
  • More than 65 cows are too many, number of milkings goes down
  • Has selected guided cow traffic to have fewer rejections
  • 65 cows works well, but he also fetches cows 4 times a day
  • More than 65 cows give too many stressed cows and thus fewer milkings.

Farmer with SAC milking robots:

  • Focuses more on teat placement than milk flow.

3 Farm Test methods

The purpose of the Farm test is to measure the time that the cow spends in the milking robot. Primarily, it measures the time for cleaning, attachment and milking. Cleaning and attachment is in the test expressed as preparation time, and preparation time can be decisive for the length of the milking time.

The farmer is interviewed about the attention that is directed towards the cows' time in the milking robot and what is done to have a good utilization of the milking robot capacity.

Farm tests are exploratory studies of new technology and new methods for Danish agriculture. The studies are done in the field. The surveys are conducted in close collaboration between the Research Centre for Agriculture, suppliers of new technology, research and testing institutions, local advisors and last but not least, the farmers. FarmTest results are primarily meant to give the manager help with decision support.

The goal of this farm test is to disclose the measurements and do some calculations based on the measurements made, comparing the different milking robots for both information to farmers with existing milking robots as well as farmers who are considering investing in milking robots.

Companies and building consultants have made references available, and contributed with recommendations for capacity utilization.

The report is based on measurement data from 34 herds from four milking robot manufacturers. In total, measurements are made for 213 cows, divided as follows:

Merlin - 38 measurements, of which 23 were suitable for calculations

SAC Futureline - 55 measurements, of which 35 were suitable for calculations

Lely A3 - 72 measurements, of which 23 were suitable for calculations

DeLaval VMS - 48 measurements of which 31 were suitable for calculations.

Cows with three teats and cows with double cleaning were discarded. However, a test was done on Merlin for double cleaning compared to single cleaning. Suitable measurings were thoses where all interim times are listed, milkings are complete, the cows can be found in the Performance Control and milk yield is noted.

Grading in Table 2, however, includes al registered measurings, because this assessment is independent of the measurement results.

Companies who have participated:

Lely Scandinavia A / S, Red Bank 114, 7000 Fredericia, Tel. 7366 1654

SA Christensen & Co., Nordre Havnevej 2, 6000 Kolding Tel. 7552 3666

Merlin Dairy Services Ltd., St. Nørlundvej 14, 7361 Ejstrupholm, Tel. 7577 3249

DeLaval A / S, Tårnvej 100, Old Hill, 7100 Vejle, Tel. 7941 3188

4 Discussion

In general, farmers are aware that problematic milkings, including time-thieves, require some adjustments of the milking robot. But with the many chores that need to be done with a dairy herd, not all deviations lead to calibration or a new introduction of the cluster. Observations on 488 cows in a test conducted by FarmTest (Kromann, 2011) in 28 herds, concludes that there are management challenges in the tested herds' milking systems. It is mainly regarding the daily routines and the control of equipment. Especially with Lely milking robots, many of the observed errors are due to the fact that the robotic arm and cleaning brushes are not adjusted.

For most farmers, the robot is an investment in better working conditions and the capacity utilization does not come first. Too much operator work (control, calibration, etc.) doesn't make the investment a work relief.

Therefore it is understandable that the capacity utilization of the robot is not prioritized, but one must not forget that optimal capacity utilization is important to ensure a good economy with milking robots.

Long working days are typical for farms with conventional milking systems, but these farmers also have the opinion that when the milking is over, there is freedom to do other things. In other words, there are no alarms that disturb. A milking robot requires daily monitoring, but it is also up to the individual farmer's temperament how much daily surveillance there should be, or in other words, what alarms you dare to give lower priority to than others. Prioritization of alarms can also be clarified with the AMS consultant, the milking robot supplier or experienced farmer colleagues.

During the inception period of approx. three months, it's a good idea to get to know all the alarms and then with this knowledge assess the risk parameters.

Appendix 1

Check list for optimal robot utilization:

  • List of ‘slow cows’ (cows that use more than the average time in robot) - check weekly
  • List of the 10% ‘slowest cows’ – check daily
  • Calibration of robot arm to be done as a routine – daily/weekly/monthly
  • Weight of concentrate calibrated with every new delivery, so that volume and weight are correct
  • Cows in late lactation with low yield that spends long time in robot to be dried off
  • Monitor two whole milkings daily (cleaning, attachment, take-offs, teat spray etc.)
  • Check brushes and cleaning cups daily

Appendix 2

The companies' comments to the FarmTest:


Lely's overall recommendation is to cut/burn udder hairs 2-3 times a year. The need for this depends on weather and temperature around the cows. Remember to burn udder with a cold flame. Every day should the laser, the liners and the air tubes be checked, so that the milking process is performed optimally.

Lely recommends that customers daily use a few 'standard printouts' to optimize the robot's time to ensure that the customer spends the time optimally. It is important that the customer focuses on cows that spend more time than average in the robot for attachment. Finally, we recommend that customers actively use the Lely certified consultants and engineers to optimize production and operation.


SAC has a lot of focus on development of milking robots that can comply with what the farmers want. SAC's milking robot concept is based on the potential for adaptation to different herd sizes. There are individual boxes with a standard capacity same as the other on the market and double boxes that have a capacity of 100-110 cows or a combination of these boxes as needed. With these combinations an investment in SAC's milking robots does not necessarily require an expansion of the number of dairy cows.

Milking robots are also very easy to integrate into existing barn, since the space required for the robot is very small. SAC pays a lot of attention to milking time per kg mil, but basically the intention has been to develop opportunities for our customers, since we sell a milking robot concept at a very competitive price, and are therefore more focused on price per kg milk than on the time. Awareness of robotic capacity is far from set in background, and therefore our management programs are developed continuously and usually SAC releases new updates every year. We do not know if the tested herds have the latest versions installed.

Preparation times and application times for cows milked in AMS depends on many factors: Cow teat placement, lactation time, udder hair and cleanliness, and the equipment to detect and attach the cups. There is thus many factors that influence dairy results and therefore we always give the farmer thorough advice in the startup period, to enable him to be a professional operator of the new robots. The farm tests are conducted on SAC's milking robots in the RDS Futureline series. SAC's newest milking robot RDS Futureline MAX is further optimized for the preparation of the cow for milking. This means less time spent in preparation of the cow and thus also an opportunity for higher capacity with the RDS Future Line MAX.


General recommendations for high capacity utilization of the robot

Reproduction and hoof health are two of many important parameters affecting the utilization of milking robot capacity. Selection of the herd's newly calved first-calf cows on the basis of milk flow and a relatively smaller proportion of young cows in the herd affects the utilization of the milking robot positively.

Box flow (milk yield in relation to time in the milking robot, where preparation, milking and teat spray is included) is an indispensable tool to identify the herd's slow cows. Box flow is higher for older cows than for first-calf cows and to be in the best quarter, heavy breeds need to have a box flow of at least 1.7 kg milk/minute, and Jersey herds at least 1.3 kg.

There is often a very clear link between good hoof health in the individual herd, and high utilization of milking robot capacity.


Morten  Lindgaard Jensen

Morten Lindgaard Jensen
1 articles

Advisor, Videncentret for Landbrug, Kvæg, (Knowledge Centre for Agriculture, Cattle), Denmark


Read more »

Videncentret for Landbrug, Kvæg

Videncentret for Landbrug, Kvæg

This article was published with permission from Videncentret for Landbrug, Kvæg (Knowledge Centre for Agriculture, Cattle), Denmark.

Translation from Danish done by