Ration Calculation on Shaky Ground: Why Hay Analyses Vary Significantly Between Labs – and What This Means for Ration Calculations

This article was translated using AI.

Key Points at a Glance

• A ration calculation is only as accurate as its two input variables: the nutrient content of the hay and the amount actually consumed. If both are uncertain, a result calculated to two decimal places is deceptive.
• In a practical test, the same carefully homogenized hay sample was sent to three different laboratories. Crude fiber was almost identical (39.1 to 40.4% of dry matter), while sugar (5.8 to 9.5%), crude ash (4.6 to 7.0%), and prececal digestible protein (2.0 to 3.3%) diverged significantly.
• The sugar values, which are crucial for horses at risk of EMS, Cushing’s, and laminitis, fluctuated the most – by more than 60 percent between the lowest and highest laboratory results.
• The laboratories state their own measurement uncertainty: One lab specifies ±31.6 percent relative for sugar, another ±3.5 percentage points absolute. A measured sugar value of 5.8% can therefore "in reality" lie anywhere between around 2% and 9%.
• The main causes are sampling from inhomogeneous hay, different analytical methods (wet chemistry versus NIRS), and biological changes in the sample itself – elevated yeast and mold counts can subsequently break down sugar and fructan.
• Hay analyses remain useful – but as an orientation regarding orders of magnitude, not as exact truth. Those who consistently use the same accredited laboratory, calculate with safety margins, and observe the horse itself (weight, body condition, manure, coat) as a "living measured value" make better decisions than any calculation program.

Hay analyses have long been standard in good feeding advice – and that is fundamentally correct. Anyone who wants to know what ends up in the trough can hardly avoid the laboratory. From the analytical values, feeding programs then calculate a ration and provide figures to the second decimal place: so much energy, so much protein, so much sugar per day. This conveys a reassuring sense of control. But exactly this feeling can be deceptive. Because a calculation is only as reliable as the numbers put into it – and in the case of hay analyses, these numbers are significantly more uncertain than the clean appearance of a laboratory report would suggest.

The impetus for this article was a specific practical case: due to a misunderstanding, the same batch of hay was sent to two different laboratories at the same time – with completely different results. What initially looked like a coincidence was subsequently systematically simulated and confirmed. The results seriously call into question the significance of pinpoint ration calculations.

What a Ration Calculation Actually Requires

At its core, a ration calculation is based on a simple multiplication: nutrient content per kilogram of feed, times the daily amount of feed consumed, equals the horse's supply. From this, it is derived whether energy, protein, sugar, or minerals are on target or whether supplementation is necessary. The procedure is popular and established among many feed advisors – the requirement values of the Society of Nutrition Physiology (GfE) provide a sound basis for this.

The problem lies not so much in the formula, but in its two input variables. Both are subject to considerable uncertainty in practice: the "nutrient content per kilogram" comes from a laboratory analysis whose accuracy – as will be shown – is limited. And hardly anyone knows the "amount eaten" exactly, especially in group housing. If both factors are shaky, then the result inevitably shakes too – regardless of how precise the decimal places look at the end.

The Practical Test: Same Hay Batch, Three Labs, Three Results

To rule out that the deviations were only due to different hay, a single batch of hay was sampled particularly carefully: the stalks were even cut to obtain as homogeneous a sample as possible. From this well-mixed material, three almost identical sub-samples were sent to three different feed laboratories. Ideally, three very similar results should have come out. In fact, they looked like this:

* pcv crude protein = prececal (in the small intestine) digestible crude protein – the protein value that actually enters the calculations for meeting the horse's requirements. "Range" column: Ratio of highest to lowest laboratory value.

The pattern is revealing. Crude fiber – the structural substance component – is practically identical at 39.1 to 40.4%; the three laboratories agree here to within a few percent. For almost all other values, however, the results diverge: digestible protein fluctuates by a factor of 1.65, crude ash by 1.52, and the sugar content – the most important single value for horses with metabolic diseases – by a factor of 1.63. In absolute numbers, this means: depending on the laboratory, the same hay appears with 5.8, 7.9, or 9.5 percent sugar.

Why Crude Fiber Remains Stable – and Sugar, Protein, and Ash Wander

It is no coincidence that crude fiber of all things lies so close together. Upon inquiry, one of the participating laboratories explained the relationships in a plausible manner. Crude fiber primarily reflects the degree of maturity and the cutting date of the grass – properties that are largely uniform in a batch. Therefore, this value is robust and well comparable.

It's different for protein and sugar: here, even small differences in the botanical composition have an impact. Meadow hay is not a homogeneous powder, but a mixture of many grasses and herbs with different leaf-to-stem ratios. Depending on which stalks happen to end up in the analysis sample weighing only a few grams, the result shifts. The laboratory put it in a nutshell: with inhomogeneous growth, even with good preparation, a single stalk can be enough to noticeably change the measured value. Deviations in ash content, in turn, usually indicate contamination with sand or differences in mineral content.

This observation is consistent with the international literature on roughage analysis: there is a consensus that sampling is by far the largest source of error – even before anything happens in the laboratory. A representative sample from several bales of a batch, ideally taken with a hay probe, is therefore more important than the choice of laboratory.

Measurement Uncertainty: What the Labs Say Themselves About Their Numbers

An often-overlooked point is found in the laboratory reports themselves – usually in small print on the back pages: the so-called expanded measurement uncertainty. It indicates the range in which the "true" value lies with approximately 95 percent certainty. And these figures are significant.

One lab quantifies the uncertainty of its sugar determination as ±3.5 percentage points – not relative, but absolute. For a reported sugar value of 5.8%, this means: the actual content can mathematically lie anywhere between around 2.3 and 9.3%. Another laboratory states the uncertainty relatively, citing ±31.6% for sugar and ±21.2% for crude fat. Both laboratories are basically saying the same thing: sugar values in particular are the most unstable of all. For comparison, the uncertainties for crude protein and crude fiber are significantly lower at ±1.6 to ±3.7 percentage points or ±7.7% respectively.

NIRS or Wet Chemistry? Why the Method Shifts the Sugar Value

A significant part of the differences between the laboratories can be explained by the methods used. Basically, there are two ways. In wet chemistry, the ingredient is determined directly chemically. In Near-Infrared Spectroscopy (NIRS), on the other hand, the sample is transilluminated with infrared light and the content is estimated using a stored mathematical model – fast and inexpensive, but ultimately a prediction and not a direct measurement. In the present test, one laboratory determined sugar and fructan via wet chemistry, another via NIRS. Exactly here was the largest difference: 5.8% (wet chemistry) versus 9.5% (NIRS).

A horse-specific study shows that this is not an isolated case. A research team compared wet chemical sugar determination with four commercial NIRS methods on 64 roughage samples. The result: large deviations between all methods – and for water-soluble sugar (WSC), all four NIRS methods systematically overestimated the content, on average by around 15 to 27 grams per kilogram of dry matter. Exactly in this direction – and in a comparable order of magnitude – the NIRS value in the practical test also deviates upwards. For feeding, this means: if the sugar content is the decisive variable, for example in horses at risk of laminitis, one should prefer a wet chemical determination if in doubt.

When the Sample Changes Itself: Sugar, Fructan, and Microbiology

Even the same laboratory using the same method does not always deliver identical values – and there is a biological explanation for that too. In repeated analyses of hay from the same source, protein, fiber, and energy remained close together, while sugar and fructan fluctuated the most. Noteworthy: one of the samples had significantly elevated yeast and mold counts – over 1.5 million colony-forming units per gram, with proven mold infestation.

This is not a contradiction, but the explanation. Yeasts and molds feed on exactly the sugars and fructans that are supposed to be analyzed. If a sample is microbially contaminated – perhaps because the hay was not optimally dried or stored – the microorganisms simply break down part of these sugars before or while the sample reaches the laboratory. The measured sugar value then drops not because the hay is "lower in sugar," but because the sample has biologically changed. Thus, even repeatability on the same hay is not a matter of course.

The Second Big Unknown: How Much Does the Horse Really Eat?

Even if the analysis were exact, the second input variable of the calculation would remain uncertain: the amount actually consumed. Ration programs calculate with an assumed consumption – for example, with the minimum amount of 1.5 kg of hay per 100 kg of body mass or with sample values like 9 kg per day. The laboratory reports themselves explicitly point out that with free roughage access, actual intake must be checked. In practice, however, this is hardly possible.

In species-appropriate group housing – which is clearly to be preferred for reasons of horse health – it is not possible to weigh how much each individual animal eats. High-ranking horses eat more, low-ranking ones less; sugar-addicted or insulin-resistant horses often eat significantly more of sugar-rich hay than healthy horses; at feeders, hay nets, and slow feeders, losses occur through sorting and trampling; and intake fluctuates considerably throughout the day. Anyone who infers individual intake from the simple amount provided is guessing – nothing more. Thus, the second factor of the ration formula also carries an uncertainty of easily 20 to 30 percent or more.

Mathematically, the problem is exacerbated: if you multiply two quantities that are each uncertain by 30 to 60 percent, then the final result can be off by many times. A ration that is reported to two decimal places can in truth easily deviate from reality by a third or more.

Pseudo-accuracy: Why Two Decimal Places Are Deceptive

The actual risk lies not in the calculation itself, but in the false trust in the result. A cleanly formatted printout with precise numbers looks objective and final – and tempts one to lose sight of the living horse. Experienced feeding experts have therefore long proceeded differently: they treat calculated rations as a rough orientation and consciously adjust the ration again and again to the horse's development instead of relying on point values. The international specialist discussion is also moving in this direction – the long-customary NSC value is increasingly seen critically, precisely because it turns out so differently depending on the method and calculation.

This explicitly does not mean that hay analyses are superfluous – on the contrary. They provide valuable orders of magnitude: whether hay is fundamentally low or high in energy, whether the sugar content is rather uncritical or rather delicate, whether abnormalities such as sand contamination or mold are present. This classification is worth its weight in gold. What the analysis cannot do is the gram-precise control of a ration – and exactly this expectation should not be placed on it.

What This Means for Practice

Concrete, actionable principles for horse owners, stable operators, and experts can be derived from the findings:

• Continue to use hay analyses – but as an orientation regarding orders of magnitude, not as exact truth. They show tendencies, not pinpoint results.
• Stay with the same accredited laboratory and the same method permanently. This way, the values over time are at least comparable with each other, even if they are not absolutely "correct."
• Trust the robust parameters (crude fiber, rough energy classification) more than the volatile ones (sugar, fructan, digestible protein) – and read the latter with the stated measurement uncertainty.
• Sample representatively: with a hay probe from many bales of a batch, mix well, as close as possible to the time of feeding. Sampling determines the result more strongly than the laboratory.
• If sugar content is the decisive variable (EMS, Cushing's, laminitis), prefer a wet chemical determination and, in case of doubt, act conservatively – soak hay, control access via hay nets, re-determine sugar yourself using a refractometer.
• Work with safety margins instead of pseudo-accuracy. Better to consciously plan a buffer than to rely on the second decimal place.
• Take the horse seriously as a "living measured value": observe body weight and fat or lymph deposits; manure consistency, coat, musculature, and performance willingness often reveal more about the actual supply than any table.
  
  

Calculate, Yes – But with Humility Regarding the Numbers

Hay analysis is a valuable tool and helps to think about feeding systematically. It only becomes problematic when the clean look of the numbers fakes an accuracy that neither the analysis nor the knowledge of the feed quantity provides. The same batch of hay can appear as low in sugar or borderline, as low in protein or adequately supplied, depending on the laboratory – and in group housing, no one knows to the kilogram exactly how much each individual horse eats of it.

The most honest conclusion is: calculate, yes, but with humility regarding the numbers instead of blind faith in the ration calculator. Those who use the orders of magnitude from the analysis, consistently maintain the same method, plan safety margins, and above all observe the horse itself attentively, ultimately feed more safely and healthily than someone who blindly trusts a ration calculated to two decimal places.

Sources

1.   Le Cocq K, Harris P, Bell N, Burden FA, Lee MRF, Davies DR. „Comparisons of commercially available NIRS-based analyte predictions of haylage quality for equid nutrition.“ Animal Feed Science and Technology, 2022;283:115158. DOI: 10.1016/j.anifeedsci.2021.115158

2.   „Comparison of NIRS and Wet Chemistry Methods for the Nutritional Analysis of Haylages for Horses.“ Journal of Equine Veterinary Science, 2018. DOI: 10.1016/j.jevs.2018.08.013

3.   Hoffman RM, Wilson JA, Kronfeld DS, et al. „Hydrolyzable carbohydrates in pasture, hay, and horse feeds: direct assay and seasonal variation.“ Journal of Animal Science, 2001;79(2):500–506. DOI: 10.2527/2001.792500x

4.   University of Wisconsin – Soil & Forage Analysis Lab: „The largest error in forage analysis is improper sampling methods on the farm.“  uwlab.soils.wisc.edu/forage

5.   Undersander D. et al. „Interpreting forage quality tests.“ Progressive Cattle / Ag Proud (Sampling as the largest source of error; NFTA certification for CP, ADF, NDF). https://www.agproud.com/articles/48120-interpreting-forage-quality-tests

6.   Oregon State University Extension (EM 9415): „Understanding Sugar and Nonstructural Carbohydrates in Equine Pasture and Hay.“  https://extension.oregonstate.edu/catalog/em-9415-understanding-sugar-nonstructural-carbohydrates-equine-pasture-hay

7.   The Horse: „Changing Carbohydrate Evaluations in Animal Diets“ – regarding method-related variation and the withdrawal of the NSC value. https://thehorse.com/127958/changing-carbohydrate-evaluations-in-animal-diets/

8.   Gesellschaft für Ernährungsphysiologie (GfE): Recommendations for energy and nutrient supply of horses, 2014. Basis of the feeding recommendations in the laboratory reports used.

9. Data basis: five real feed test reports (LUFA Nord-West, LKS, Raiffeisen-Laborservice) of the same or origin-identical batches of hay, 2025/2026, as well as the technical statement of one of the laboratories on the question of deviations.