NIRS Technology in Animal Nutrition: How It Works and Its Impact on Formulation

Those who work with feed formulation know that the quality of a formula does not depend solely on a good linear programming equation. It depends, above all, on reliable information about what is entering the mixer. And it is precisely at this point that NIRS technology has been transforming the routine of laboratories, quality teams, and formulators at animal nutrition and production companies in Brazil and worldwide.

Near Infrared Reflectance Spectroscopy, known by the acronym NIRS, is not a new technology. It has existed since the 1960s and was widely used for decades in the grain and cereal industry in the United States and Europe. What changed in recent years was the combination of three factors: the advancement of chemometric mathematical models, greater equipment accessibility, and the maturity of analytical management software that allows transforming spectral data into operational decisions.

The practical result is that today a medium-sized feed mill in the interior of Paraná can access bromatological analyses in less than two minutes per sample, without generating chemical waste, without need for reagents, and at a cost per analysis significantly lower than conventional wet chemistry methods. But to leverage all this potential, it is necessary to understand how NIRS really works, what its limits are, and how to connect its results to the formulation process efficiently.

What NIRS technology is and why it has gained strategic relevance

NIRS is a vibrational spectroscopy technology that uses the interaction between electromagnetic radiation and matter to determine the chemical composition of a sample. The spectral range used is approximately 750 to 2500 nanometers, a region immediately adjacent to the visible spectrum and mid-infrared. In this range, radiation interacts with molecular bonds present in organic compounds, especially with hydrogen-containing functional groups: C-H, N-H, O-H, and S-H bonds. This explains why NIRS is so efficient for quantifying protein, moisture, fat, fiber, and carbohydrates in organic matrices such as grains, meals, oils, and feeds.

When a near infrared light beam strikes a sample, part of that energy is absorbed by the organic compounds present in the material and part is reflected back to the equipment's sensor. The resulting reflectance spectrum is essentially a molecular signature of the sample. Each ingredient has a unique spectral profile, and when chemical composition changes, the spectrum changes with it. From there, mathematical models are used to convert this spectral signature into quantitative values of the components of interest.

The role of chemometrics in NIRS analysis

Chemometrics is the set of statistical and mathematical methods that turn raw spectral data into interpretable information. It is what allows NIRS to go beyond a simple optical reading and deliver results such as "crude protein: 45.2% on an as-fed basis" or "moisture: 11.8%". To do this, the process involves building calibration models, which are mathematical equations trained on a set of reference samples whose values were previously determined by conventional analytical methods, such as AOAC methods or the Kjeldahl method for protein.

These calibration models use multivariate regression techniques, primarily Partial Least Squares (PLS) regression, to find the relationship between spectral variations and variations in analyzed parameter values. The coefficient of determination (R²) is the main indicator of model quality: the closer to 1.0, the better the model describes the relationship between spectrum and analyzed parameter. Models with R² above 0.95 are generally considered robust for quantitative prediction under industrial conditions.

Two other indicators are essential for evaluating model reliability: SECV (Standard Error of Cross Validation) and SEP (Standard Error of Prediction). The first measures model error during calibration and the second measures error when the model is applied to new samples, which is what really matters in daily use. A good calibration must present prediction errors compatible with the natural variation expected for that parameter in the use context.

Companies specializing in animal nutrition, such as AB Vista, Adisseo, and Evonik, have developed proprietary NIRS reading platforms with calibrations already built and validated for the main animal diet ingredients, including specific models for digestible amino acids and phosphorus digestibility. These platforms significantly reduce the time and effort required for a company to implement NIRS with analytical quality, especially when there is no internal expertise to develop their own calibrations.

Types of NIRS equipment and their applications in the animal nutrition industry

There are three major categories of NIRS equipment with relevant applications in the animal nutrition and production industry. Each serves different needs and implies distinct investments and analytical capabilities. Understanding the differences is fundamental for making an investment decision suited to each company's reality and objectives.

Portable NIR: agility at receiving and in the field

Portable NIRS equipment has gained considerable ground in recent years, especially with the popularization of compact devices that fit in a backpack and run on battery. They are particularly useful for rapid analyses at raw material receiving, allowing an analyst to screen a soybean meal or corn shipment before the truck is even unloaded. They also have relevant use on farms for silage, grass, and roughage analysis, and at slaughterhouses and farms for animal-origin meal evaluation.

The main limitation of portable devices lies in the spectral range they cover, generally narrower than benchtop equipment, which restricts the parameters that can be accurately analyzed. Spectral resolution is also typically lower. For analyses such as crude protein, moisture, and ether extract in high-homogeneity ingredients, results are reliable. For more complex analyses, such as free amino acids or contaminants at low concentrations, accuracy may be insufficient depending on the equipment and available calibration.

Benchtop NIR: the analytical standard of industrial laboratories

Benchtop NIR is the most established type in the animal nutrition industry and can be found in laboratories of feed mills, premix plants, and large integrators throughout the country. It operates across the full near infrared spectral range, from 750 to 2500 nm, with high spectral resolution, giving it greater prediction power and accuracy for a broader range of parameters. Under adequate calibration and sampling conditions, a well-configured benchtop NIR can reliably predict parameters such as crude protein, moisture, ash, ether extract, crude fiber, starch, NDF, ADF, total amino acids, and in some cases, digestible amino acids and phosphorus digestibility.

For the equipment to deliver reliable results on a daily basis, it is necessary to maintain a disciplined routine of instrument performance verification, calibration model updating, and cross-validation with wet chemistry reference analyses. Many companies neglect this analytical maintenance and end up using outdated calibrations that no longer reflect the actual composition of the ingredients they are purchasing, especially when they change suppliers or crop seasons.

Inline NIR: continuous analysis integrated into the production process

Inline NIR represents the most advanced stage of integration between spectral analysis and industrial process. In this configuration, the NIRS sensor is installed directly on the production line, whether in the mixer, feeding chute, silo output, or scale conveyor, to perform continuous real-time analyses without the need to collect and transport samples to the laboratory.

The practical implications are significant. An inline NIR installed in a feed mill mixer can generate nutritional profiles for each produced lot, monitor mix homogeneity, and detect deviations from the target formulation in seconds. If a lot's protein content is below specifications, the system can alert the operator before the product is packaged or dispatched. In terms of process control, it is a paradigm shift: we move from a reactive model, where the problem is identified after the fact, to a predictive and corrective real-time model.

Investment in inline NIR is higher than in other categories and implementation requires a customized parameterization process for each production line. But the results in terms of rework reduction, product uniformity, and raw material cost management typically justify the investment in a relatively short period for companies with relevant production volume.

What NIRS can actually analyze in animal production

One of the most common questions among professionals evaluating NIRS adoption is clearly understanding which parameters can be reliably analyzed and which still depend on conventional methods. The answer is not unique because it depends on available calibration quality, the type of matrix analyzed, and the equipment used. But it is possible to draw a reasonably clear picture of the state of the art in the animal nutrition industry.

Classic bromatological parameters

For conventional bromatological parameters, such as crude protein, moisture, ether extract, ash, crude fiber, NDF, and ADF, NIRS is already a mature, widely validated technology for the main ingredients used in production animal diets. Soybean meal, corn, sorghum, wheat, fish meal, meat and bone meal, wheat bran, corn gluten, and other frequently used ingredients have available calibrations with good robustness on various commercial platforms and quality benchtop equipment.

For these parameters, NIRS delivers results in less than two minutes per sample, without waste generation, with analysis cost far lower than wet chemistry, and with the possibility of archiving the sample spectrum for retrospective analyses. This means that if in the future a question arises about the quality of a raw material lot that has already been consumed, it is possible to reprocess the archived spectrum with a new calibration model and obtain additional information.

Digestible amino acids and nutrient digestibility

This is one of the most promising and also most demanding fields of NIRS in animal nutrition. Prediction of digestible amino acids, especially digestible lysine, digestible methionine, and digestible threonine in protein meals and animal-origin meals, is already possible with calibrations developed by specialized companies. The importance of this for formulation is enormous. When a formulator works with total amino acid values instead of digestible values, they are essentially assuming all ingredients have the same bioavailability for that amino acid, which is known not to be true. A soybean meal with excessive thermal processing will have significantly lower lysine digestibility than properly processed meal, even if total lysine content is similar. If formulation does not capture this difference, animal performance will fall below expectations or the nutritionist will need to work with artificially elevated safety margins, which increases cost without providing real benefit.

Mycotoxins, contaminants, and other safety parameters

Mycotoxin detection by NIRS is a field that has advanced considerably in recent years, but still with important limitations that need to be understood. Near infrared spectroscopy techniques combined with robust chemometric models are already capable of detecting aflatoxins, deoxynivalenol (DON), fumonisins, and zearalenone in corn and cereals, especially at higher concentrations. However, for very low detection limits, which are the ones relevant from a regulatory and safety standpoint for poultry and swine feeds, NIRS analytical sensitivity may still be insufficient depending on the equipment and calibration model.

In this context, NIRS works best as a rapid screening tool: samples showing negative results or well below the limit of interest can be released more quickly, while samples showing positive signals or close to the limit are forwarded for confirmation by ELISA, chromatography, or other reference methods. This combined approach optimizes laboratory resource use and accelerates raw material receiving flow without compromising process safety.

Quality of fats, oils, and liquid ingredients

NIRS also has application in liquid ingredients, such as vegetable oils and animal fats, for determination of parameters such as acidity, peroxide value, fatty acid content, and oxidation degree. For companies using large volumes of soybean oil, poultry fat, or beef tallow in feed formulation, this is an application with relevant economic potential, since variation in the quality of these ingredients has a direct impact on diet metabolizable energy and, consequently, on animal performance and feed efficiency.

The impact of NIRS on precision formulation

For those who formulate feed — whether for broilers, swine, cattle, aquaculture, or pets — the value of NIRS lies in the quality of information it provides to feed the formulation process. And when we talk about information quality, we are talking about three dimensions that are often treated separately but must be seen in an integrated way: precision, timeliness, and representativeness.

Updating nutritional matrices with real operational data

One of the most common and at the same time most problematic practices in feed formulation is the use of nutritional composition tables as if they were fixed, universal values. Corn from supplier A, from this year's crop, in the state of Mato Grosso, does not necessarily have the same composition as the corn described in Rostagno or Sakomura tables, which were built from samples collected under different conditions, different years, and different regions. Variability is real, relevant, and systematically underestimated in many operations.

When a company implements NIRS and begins systematically analyzing the raw materials it receives, it begins building its own nutritional composition database, based on real operational conditions: its suppliers, its region, its season. Over time, this database allows calculating real means and standard deviations for each ingredient and supplier, and using these values to feed formulation software nutritional matrices with much greater precision than would be possible using only composition tables.

The practical impact on formulation is twofold. On one hand, the formulator can reduce safety coefficients applied to critical nutrients because uncertainty about the actual ingredient composition is lower. On the other hand, formulation more consistently delivers the specified nutritional levels, which translates into better animal performance and lower variability between produced lots.

Reducing variability between the formulated and produced product

There is a difference every experienced formulator knows, but that is not always rigorously measured: the difference between what is in the formula and what is actually in the feed bag. This difference has multiple origins. Weighing errors, ingredient segregation during transport, mixing process problems, and raw material composition variation are all sources of deviation between formulated and produced.

NIRS acts directly on the last of these sources. When ingredients enter the process already with a known and updated composition, formula calculation starts from more solid premises. And when the finished product is also analyzed by NIRS before dispatch, it is possible to verify whether guarantee levels are being met and take corrective actions quickly when a deviation is identified. This is especially relevant for nutrients with high economic or regulatory impact, such as crude protein and amino acids in poultry and swine feeds.

Supplier qualification based on consistent analytical data

Raw material supplier qualification is an activity that frequently suffers from lack of objective data. Many companies still make purchasing decisions based primarily on price per ton, without systematically considering composition differences between suppliers. The result is that in practice, the cost per nutrient unit can be completely different from cost per ton, and the company is paying more for the same nutritional performance without realizing it.

With systematic NIRS analyses by supplier over time, it is possible to build a quality history that supports smarter purchasing decisions. A soybean meal with average protein of 47% from one supplier is not equivalent to a meal of 45% from another, even if both declare the same value on the report. And when the composition difference is captured and fed into the formulation process, feed cost can be dynamically adjusted to better leverage each available ingredient in the market.

NIRS, laboratory, and formulation: integrating data for smarter decisions

One of the most common problems observed in companies that already use NIRS is that the generated data stays trapped inside the laboratory. The analyst makes the reading, records the result in a spreadsheet or the equipment's own system, and the data stops there. The formulator has no real-time access, the quality manager cannot visualize trends over time, and procurement has no visibility into the variability of the ingredients being acquired. This isolation of analytical data is a waste of potential that is unfortunately still very common.

Integration between laboratory and formulation process requires, first, structuring the data flow: from the moment the sample arrives at the laboratory to the moment the result is available for the formulator to consult. This includes standardizing the sampling process, correctly identifying lots, recording results in a centralized system accessible to the teams that need them. Laboratory management platforms, such as Labinfy, have been adopted by companies seeking this integration, allowing NIRS-generated analytical results to be organized, queried, and compared in a structured way, simplifying the matrix update process and ingredient variability monitoring over time.

When the formulator has access to a structured analytical history of their ingredients, they can make much more informed decisions. If soybean meal from a specific supplier has been showing crude protein systematically 1.5 percentage points below the declared specification over the last 30 days, this is information that must reach the formulator so they can adjust the matrix or take another action. Without a system that consolidates and presents this history in an accessible way, this perception is unlikely to happen proactively.

On the formulation side, systems like Formulamix were developed to integrate with this laboratory data flow, allowing nutritional matrices to be updated based on actual analytical results and giving formulators clear visibility into the impact of composition variations on formula costs and nutritional quality. This laboratory-formulation integration is what transforms NIRS from an analytical control tool into a strategic production management tool.

Key considerations for implementing NIRS with quality

Adopting NIRS technology brings real gains, but only when implemented with methodological rigor. There are some common pitfalls that compromise result quality and, consequently, the confidence the company places in the equipment. Knowing them in advance is important for avoiding frustration.

Calibration: the foundation of everything

Calibration model quality is the most determining factor for NIRS result reliability. High-resolution equipment with poor calibration will deliver poor results. Calibration must be built or validated with representative samples of the raw materials that will be analyzed daily: actual suppliers, actual crops, typical process variations. Generic calibrations or those built from samples from a different geographic reality or industrial context may show systematic deviations that go unnoticed for months.

In addition, calibrations need to be maintained and updated periodically. The composition of ingredients such as soybean meal and corn varies between crops, production regions, and climatic conditions. A calibration model built three years ago may not adequately represent the ingredient being received today. The practical recommendation is to establish a model performance verification routine at least at each new crop season or whenever there is a significant supplier change.

Representative sampling: the critical link in the chain

There is no reliable NIRS analysis without adequate sampling. The generated spectrum represents the composition of the sample placed in the reading cell, not necessarily the composition of the lot as a whole. If sampling is done inadequately — for example, collecting material only from the surface of a truck or a single silo point — the result will represent that specific point, not the entire lot.

Heterogeneous ingredients, such as animal-origin meals, silages, and mixtures, require even more rigorous sampling protocols. The general rule is to collect multiple sub-samples throughout the lot, homogenize them, and prepare a representative composite sample before any analysis. For inline analyses, the continuous sampling process is the material flow itself passing through the sensor, which in theory guarantees representativeness but still requires attention to correct sensor positioning and flow homogeneity.

Cross-validation with reference chemistry

NIRS should never be seen as a complete substitute for wet chemistry reference methods, but as a complement that increases the speed and frequency of analyses. Maintaining a routine of parallel reference analyses is essential for monitoring calibration performance and identifying any deviations before they impact operational decisions. A good practice is to define a cross-validation protocol — for example, analyzing by wet chemistry a fraction of all samples analyzed by NIRS and systematically monitoring differences between the two methods over time.

Perspectives and trends of NIRS in the animal nutrition industry

The global market for NIRS equipment and solutions for the food and animal nutrition industry continues to expand. Technological advances are making sensors increasingly smaller, more robust, and connected. New equipment already comes integrated with data management platforms that facilitate spectrum archiving, calibration management, and trend visualization over time.

One of the most relevant trends is the integration of NIRS with artificial intelligence and machine learning platforms for building more robust and adaptive calibration models. Models that continuously learn from new samples and update semi-automatically have potential to significantly reduce the workload associated with calibration maintenance, which is today one of the main operational obstacles for many companies.

Another important trend is the democratization of access. Equipment that ten years ago was restricted to large integrators and multinationals is now within reach of cooperatives, regional feed mills, and medium-sized producers. The combination of more accessible equipment with calibration-as-a-service platforms, where the company uses models developed and maintained by outsourced specialists, is opening NIRS technology to a much broader segment of the industry.

And as more companies adopt NIRS, the standard demanded from raw material suppliers also rises. When a buyer has the analytical capacity to quickly verify the composition of each received lot and compare it with what the supplier declared, pressure for real quality — not just reports — naturally increases. This has a positive effect on the entire chain: more traceability, more consistency, and less variability in ingredients arriving at feed mills.

From analysis to result: why NIRS data must leave the laboratory

The discussion about NIRS in animal nutrition frequently ends at the level of equipment and analysis. What is the best equipment, what is the best calibration, which parameter has the highest accuracy. These are important questions, but not sufficient. The technology's real potential is realized when the data it generates is integrated into the company's decision-making process in a structured and routine manner.

Formulators with systematic access to NIRS analytical data can update their nutritional matrices based on the actual composition of the ingredients they are using, reduce artificial safety margins that increase formulation cost without proportional benefit, and make ingredient substitution decisions with more safety and speed. Quality teams that visualize trends in analytical history can identify problematic suppliers before the final product quality is compromised. Industrial managers with visibility into ingredient variability can plan the production process more assertively.

All of this is only possible when data leaves the equipment, passes through a system that organizes and contextualizes it, and reaches the right people at the right moment. NIRS technology is the starting point for generating this data. Intelligent management of this data is what determines the real impact it will have on company results.

For companies on this path, the combination of a solid analytical process in the laboratory and a formulation system that consumes and reflects this data in an integrated way defines the difference between using NIRS as a compliance cost and using it as a real competitive advantage.