# Balancing Quality and Yield in Beef Carcasses - The Natural Valley Experiment

Here I recount the very successful experience in beef cattle marketing when I was involved with a new medium sized packing plant called Natural Valley Farms located near Wolsley, Saskatchewan. I think this story should be told because, in a nutshell, we demonstrated that a packer could purchase slaughter cattle on the basis of both quality and yield without creating any difficulties for the packer and with great advantage to the producers who participated in this experience.

In the years leading up to 2004 I was assisting Julie Stitt in the development of the National Cattle Identification Program. In that capacity I met frequently with Kathy Martin while attending meetings to explain and promote the program. We both saw national ID as an opportunity to gather and to return to producers the grade information on individual cattle. This had not been possible before National ID and we often commented on this as a new opportunity for the industry. From Kathy, I learned of the establishment of a new medium sized packing plant called Natural Valley Farms being set up Wolsley, Saskatchewan. Shortly thereafter Kathy was hired by Natural Valley Farms and was put in charge of cattle procurement. Because of our previous discussions Kathy invited me to make a proposal to the Principals of the packing plant. I described a proposal that I will shortly lay out here. I recall making two presentations about a month apart. Some skepticism arose during the first presentation and in the interim I fleshed out my proposal in more detail. At the end of the second presentation I was given permission to proceed.

What we did is best explained by describing the proposal.

I first explained that there is considerable variation in the “yield” of beef carcasses. By yield I was referring, not to “dressing percent “, but to some representation of saleable product from a carcass. The industry has long recognized this as evidenced by the three yield classes included within the present grading system. I need to digress a bit in order to explain the present yield class system.

When the grading regulations were revised in 1993 a yield classification system was created that was based on a formula developed a by researchers at the Lacombe Research Station in Alberta. That formula, based on Hot Carcass Weight (HCW), Rib Eye Area (REA) and Fat Thickness, shown below provided an estimate of the closely trimmed lean meat yield from the five major primal cuts of a carcass.

*Lean meat yield % is:*

I have no doubt that this was a reasonably accurate formula in delivering a lean meat yield percent. In fact, it still is the basis for the yield estimates made today by graders. But from my perspective it is flawed in at least two important respects.

As noted above the estimate applies only to the five major primals; Chuck, Rib, Loin and Short Loin and Round. Thus, it relates to only 86% of the carcass weight.

The reference to closely trimmed lean meat yield underestimates the saleable yield of a carcass. We know now that the average saleable yield of a carcass is between 73 and 75% of the carcass weight. The understatement of saleable yield creates an unfortunate impression in the minds of many that the average yield of a beef carcass is under 60%

A greater problem was created when the researchers attempted to translate the formula into the “Graders Rule” which is still used by graders today. I have written a separate article which outlines the very serious flaws in the graders rule. These flaws are so great that they render the yield estimates almost useless. If this were not so the industry would not need computer and camera assisted grading.

I emphasize, as I always do, that the flaws in the yield grading system are not in any sense the fault of the grader. They are the fault of an imperfect graders rule that was devised almost 25 years ago and *was intended as an interim measure* pending the expected immediate introduction of camera and computer assisted grading. That expectation and capability did not materialize until just recently and is still not in use for grading purposes.

So, when I was given the opportunity to proceed I decided not to use the graders rule. Instead I asked the Grading Agency to have their graders measure the carcasses for us. This involved measuring the length and width of the Longissimus Dorsi muscle (from which to estimate Rib Eye Area) and measuring fat thickness. I was disappointed when the Agency declined to cooperate. They continued to estimate yield percent but declined to share the measurement data with us.

I therefore decided, with the help of Kathy Martin, to train one staff member to make the measurements for us. A few years earlier I had done some work using Rib Eye Tracings of known area to work out that a close fit for actual area was, Length x Width x 0.78. This was not a perfect method but was far superior to the Grader’s Rule, as will be demonstrated shortly. This formula also avoided the laborious and time consuming task of actually measuring the REA. Note that the graders do not measure REA either but just use Length and Width to create a “Muscle Score” which, combined with a “Fat Class”, gives a % yield estimate. Therein lies the source of most of the errors made in estimating yield percent.

The formula I used was the same as the formula used by the grading agency, with two changes.

First, I quite arbitrarily made the average yield 60%. This may seem odd but substantial data existed to indicate that the average yield was much closer to 60% than to 57.96%. It is also apparent that the solution to the three bracketed components in the equation is not affected by the base yield.

Second, I used the length and width measurements to directly estimate the impact of Rib Eye Area on yield.

Therefore, the formula I used was:

**How Natural Valley Paid for the Cattle**

An important discussion was held about how to pay for the cattle. Some of the Principals wanted to figure out a way to pay a “premium” for high yielding cattle and to deduct a “discount” for lower yielding cattle. This made sense to them because quality is rewarded with a premium over the base price for AAA and Prime Carcasses and a “discount for A carcasses”. I argued against this approach pointing out that premiums and discounts were appropriate for quality because quality is a subjective thing that cannot be measured very well numerically whereas yield can actually be measured and can therefore be rewarded directly. The method I proposed and that was adopted was quite simple as the following illustration will show.

1. Set a competitive Base Price for an AA carcass, example $160.00/.cwt. (the approximate price at the time)

2. Determine a competitive AAA and Prime premium (example, + $5.00 for AAA and + $7.00 for Prime) and an appropriate discount for A and B carcasses.

*E***xample 1**

Carcass 1 is graded AAA so attracts a quality based price of $165.00 per cwt. We now apply the yield factor. Carcass 1 had a yield of 62%. Therefore, since the base yield was 60% the settlement value for Carcass 1 is $165.00 x 62/60= $170.50.

**Explanation and Proof**

When the packer decided to pay $165.00/cwt for an AAA carcass, and assuming an average saleable meat yield of 60%, he was actually paying $165.00/60 = $2.75 per pound of saleable meat yield. But Carcass 1 had a yield of 62%. Thus 62 x $2.75= $170.50. In this way, the packer is paying exactly the same for every pound of saleable beef for every carcass within each quality grade as example 2 will confirm.

*Example 2*

Carcass 2 is also AAA so has a quality based price of $165.00/cwt. But this carcass yields only 58% so the settlement is $165.00 x 58/60 = $159.50. Again, the packer paid $2.75 for every saleable pound of carcass weight but there were just 58 pounds. So, the settlement would be 58 x$ 2.75 = $159.50/cwt

This is fair. The seller of Carcass A is not subsidizing the seller of carcass B and each is getting paid exactly the same for each pound of lean meat yield of an AAA carcass. Instead of averaging his purchases the packer is paying for what he gets. And, best of all, the producer discovers what he was selling and gets a valuable signal about what the packer is looking for in terms of both quality and yield.

Thanks to the confidence of management in my proposal this method of payment was instituted. To be sure there were no mistakes made, I had all the measurements emailed to me until I was convinced that the people working out the settlements knew exactly how to proceed. Meanwhile, my greatest concern was how the producers who marketed lower yielding cattle would react to lower settlements on their cattle. They had been informed of the payment method but I was still unsure how those with lower payouts would react. I need not have been concerned. Certainly some were disappointed but, as one such producer stated to me, this was the first time that he had learned anything about yield variations. As a result, he set about to change his feeding practices and to consider the breeding of the cattle he purchased. Also, to some extent the impact of low yield in some cattle was offset by higher yielding cattle in the same shipment. Something else happened quite quickly. When producers became aware that they were being paid for both Quality and Yield, they started to market higher yielding carcasses.

I do not have an accurate count of how many cattle were marketed under this pricing structure. But I do have a record of 2,050 cattle purchased in 32 Lots. The total payment for these cattle purchased under the payment method described above was $2,303,871.74. I also calculated what the payments would have been if the cattle had been bought solely on the basis of their Quality Grade with no adjustment for yield. That payment was $2,279,895.04 for a difference of -$23,976.70 or -1.04%. I found the difference to be almost negligible and in fact the plant paid slightly more. The difference could as easily been in the other direction. But this fortified my assertion that paying for yield variations would not cost the packers significantly more (or less) than would disregarding yield.

But there is more to tell. The average yield as determined by the Natural Valley formula was 59.82% and as determined by the Graders was 60.12%. This merits a comment. I have said that the yield percent as determined by the graders is inaccurate so how do I explain the close agreement between the CBGA average yield and the Natural Valley average yield? The answer is found in the fact that a large number of carcasses were involved. In our approach we used a consistent technique based on actual measurements and a formula. Therefore even though we may have erred slightly, and we undoubtedly did, the errors would be consistent. In other words, our method would have tended to rank the carcasses correctly. But the graders were obliged to estimate the yield using a graders rule which is a poor proxy for a formula based approach. So errors in estimates made by the graders would tend to be self cancelling since overestimates were no more likely than underestimates.

There is a story worth telling here that illustrates the point made above. Nearly two centuries ago Sir Francis Dalton, one of the fathers of modern statistical analysis, attended a country fair and was attracted to a side show where, for a small fee, one could guess the weight of a large ox. The proceeds of the wagers would go to whomever came closest to the mark. Many estimates were made. Some of the people were very familiar with livestock and their guesses were undoubtedly close. Some who knew nothing at all about livestock would have guessed wildly high while others would have guessed ridiculously low.

Sir Fransis Dalton may or may not have made his own guess but his interest was not in winning the pot. Instead he asked if, after the guesses were tabulated and the winner chosen, he could have the several ballots to study. His wish was granted and the Dalton proceeded to plot them, to calculate their distribution and ultimately their average. As the story is recounted, the distribution made an almost perfect “bell curve” with a few stragglers high above and far below the average forming the base of the curve with increasing numbers as the average was approached. And the Average? To everyone’s surprise, except Dalton, the average estimate was 1,197 lbs and was almost precisely the same as the 1,198 lb actual weight of the ox. The overestimates had cancelled out the underestimates.

So, the average estimate of yield as determined by CBGA graders can be relied upon as accurate but no such claim can be made about the individual carcasses that make up that average.

To demonstrate this point I insert here a chart which shows the yield of 1,844 carcasses as measured by the Natural Valley method, the smooth curve, and estimates by CBGA graders which dance around above and below the Natural Valley estimates. To be fair to the graders, part of the reason for the wide variations is that the Graders Rule which they use to estimate a grade cannot give a yield greater than 65% or lower than 49%. The Natural Valley formula found that 6% of the carcasses had yields above 65.5% and a small number had yields below 49%. Nonetheless, the deviations are so great as to render the CBGA estimates completely unreliable, again the fault of the measuring device, not the grader. This, I think is a part of the reason that packers have paid little attention to yield estimates made by the graders. They know that the grader’s rule is unreliable.

**Differences in NVF and CBGA Yield**

**1844 Carcasses**

I pointed out earlier that total payouts were almost exactly the same for cattle bought on the Natural Valley system or bought without reference to yield. But that hides an important truth.

I measured the difference in payouts on an individual lot basis. There were 32 lots and lot numbers varied from 203 head to a small lot of 6 head. The lot with the highest average yield contained 45 cattle and the average extra payout due to high yield was $64.77 per head. Another large lot or 128 head received an extra $64.56 per head due to high yield. The lowest yielding lot was a lot of 65 head that was paid out at $38.23 per head less than it would have received if paid on a quality basis only.

This experiment demonstrated three important things.

First, that under the present method of sale high yielding lots of cattle are penalized and low yielding lots benefit from a payment system that ignores variations in yield.

Second, that those packers that measured yield more accurately, especially with modern technology can pay for cattle more fairly without paying any more (or any less) on average.

Third, and most importantly, that it is possible to combine high yield and high quality in the same animal and produce better carcasses to the benefit of the entire industry. But to do this the producer needs two things. The producer needs to be paid for the yield that is produced and also needs to receive the information needed to alter breeding, production and marketing practices.

This exercise also demonstrates quite conclusively that the current method of estimating yield percent in beef carcasses is hopelessly flawed and that any information that does flow to the producer concerning yield is unreliable. Yield and quality are the two most important characteristics of a beef carcass. How does an industry improve its product when it persists in measuring yield in a flawed manner?