No, you won't come close to the precision of a university trial. If unbiased research already exists, you should probably go with that instead. But, in some situations, you need more information.

An 1,800-cow dairy in central California is now feeding brown mid-rib corn silage, thanks to an on-farm trial.

The farm was large enough to have three high-producing milking strings with cows in their second lactation or greater. The 392 cows in the experimental group, which received the brown mid-rib silage, averaged 8.6 pounds more milk per day than the 203 cows in the control group. Dry matter intake in the experimental group was 1.5 pounds per day higher, on average.

The experiment was set up to minimize the chance of outside bias. Brown mid-rib corn silage was grown in the same field as the regular silage fed to the control group. After ac-counting for any possible differences between groups, the 8.6-pound difference in milk production was just too great to discount the probable effect of the silages.

"During that period of time, we did not change the other ingredients in the ration," points out Ray Hinders, the herd's consulting nutritionist, who has had extensive experience in running on-farm trials.

You don't have to be a big dairy, like the 1,800-cow operation in California, to run an on-farm feeding trial. Farms of all sizes can benefit, assuming the trials are run correctly and the farms go about it with reasonable expectations.

Decision-making tool
An on-farm feeding trial requires precision. Unless it is run correctly, the farm may arrive at false or misleading conclusions. Some professional nutritionists are adamant about it: A farm can never gather useful information from an on-farm trial because there's just too much variability involved.

But, no one is expecting you to meet the same strict standards required of professional nutritionists who publish their work in scientific journals. They must control their experiments so precisely that the chances of reaching a wrong conclusion are no more than one out of 20.

You're not publishing an article in a scientific journal. You're simply trying to improve your odds of making the right decision. It's a decision that you have to make regardless of whether you run a feed trial or not.

Cost of a bad decision
What is the cost of being wrong? If a 50-cow herd considered using a product that cost 1 cent per cow per day, it would be looking at spending $152.50 per year, assuming the cows receive the feed additive throughout a 305-day lactation. In that case, it would cost more to run the trial than any potential savings.

At some point, however, you cross an economic threshold. You need to know with some certainty that the benefits of using a product will more than cover the added cost.

If the cost of a product is less than 5 cents per cow per day, it's usually OK to rely on "collateral" information from outside resources, including published research and nutritional consultants, says Normand St-Pierre, dairy scientist at Ohio State University. If the cost is between 5 cents and 20 cents per cow per day, then you're in a "gray area," he says. Hopefully, enough research exists that a trial won't be necessary. But, once the cost exceeds 20 cents per cow per day, and little in the way of research exists, then you need the ability to ascertain the response in your herd with an on-farm trial, he adds.

Peter Robinson, extension dairy nutrition and management specialist at the University of California-Davis, says part of it depends on how readily you can see the benefits of a product and the amount of research that has been published on it. "There is little point in doing an on-farm trial on BST, since the benefits of BST are evident and the published (research) base is strong," he says. Similarly, sodium bicarbonate and some yeast products have a strong research base, as well.

"In contrast, a trial on an RP methionine product may be advised since the benefits will not likely be evident on farm and the published research base is weak," Robinson says.

Herd size may be another factor in deciding whether to run a feeding trial on-farm. The larger the herd, the larger the impact in terms of total dollars. But, a small farm can suffer just as much - on a relative basis - by making the wrong decision.

For small farms, too
Both Robinson and St-Pierre agree that farms of all sizes can benefit from on-farm feeding trials if the studies are designed properly.

A large farm has an advantage because there are more cows to choose from in a particular production group. For instance, it has more cows in a "high" milk production group, which is important because feed additives often produce their strongest response in high-producing animals. In contrast, half of the animals on a 50-cow dairy may be past peak lactation at any given time.

But, large farms have certain disadvantages, as well. Since they keep animals in pens or large groups, it's more convenient to feed the animals by group. And, that can lead to erroneous observations, since pens of animals have their own unique environments that can confound the experimental results. For instance, when the animals in a feeding group are run through the milking parlor, they may encounter a different set of conditions - such as a change in the milking crew - than those encountered by another feeding group.

One way to overcome this problem, St-Pierre says, is to switch the pens back and forth from "experimental" to "control" status in successive months. That way, if the "experimental" group consistently does better - regardless of which pen or pens happen to be the "experimental" group - then you have a good indication that the feed additive or treatment works, despite any differences between pen environments.

A small farm with 50 tie-stalls is actually a good research environment, because the producer can simply designate every other stall as "experimental" or "control." That helps make the experiment truly random.

And, you don't need a large number of cows. You need 40 cows - 20 in the experimental group and 20 in the control group - to be able to say with 95 percent confidence that a feeding change produced a 2-pound difference in milk production, St-Pierre says. Therefore, even a 50-cow herd would have enough animals to run some comparisons, as long as the comparisons did not require that all cows be in a "high" production group or a transition group.

While an on-farm trial may not be as precise and accurate as a university trial, it can still yield important clues about the cost-effectiveness of a feeding change.

Avoid these problems

These are among the problems that can plague on-farm feeding trials:

  • Lack of randomization. This occurs when animals are selected for "experimental" or "control" groups, but the selection is not done on a random basis. In other words, there's something about the way the animals are chosen - such as location within a building - which biases the results.
  • Not accounting for other factors. If cows in the experimental group go up in milk production by 2 pounds per day, on average, was the increase caused by a new feed additive or a change in temperature, the quality of forage, or some other factor?
  • Normal variation in the herd. Cows are not computers - they do not always perform the same way day in and day out. And, the people managing them are the same way. Many herds have random "background noise," where the bulk tank average will vary by a couple of pounds per day simply because of random variation. You can't always predict it, nor can you explain it. It just happens.