How to Reduce Animal Nutrition Costs with Feed Formulation Software

Least-cost formulation is the core method of any feed formulation software. The algorithm finds the combination of ingredients that meets all nutritional constraints at the lowest possible cost. But the software's real value goes beyond that single solution: it lies in the indicators the model produces alongside the formula, in the variability data that feeds the model with precision, and in the ability to repeat and scale the process across multiple products simultaneously. This article covers each of these dimensions in depth.

The optimizer's logic: cost per unit of nutrient, not per kilogram of ingredient

The most common mistake in subjectively evaluating ingredients for formulation is comparing them by price per kilogram. That's not how the optimizer works. It evaluates each ingredient by the cost it represents per unit of each nutrient it delivers, simultaneously considering all nutrients relevant to the diet and all interactions between ingredients in the model.

Two protein ingredients with a similar price per kilogram can have completely different economic values in the model, depending on protein concentration, amino acid profile, digestibility, the presence of antinutritional factors that limit maximum inclusion, and the correlation between the nutrients they supply and those that are limiting in the formula at that moment. Soybean meal at R$2.20/kg may be cheaper than sunflower meal at R$1.60/kg if the diet's limiting nutrient is lysine and soybean meal delivers more lysine per real spent than sunflower meal. The model calculates this automatically for all nutrients and all ingredients at once.

Understanding this logic is the prerequisite for correctly interpreting the optimizer's outputs. Shadow price, optimal price, and sensitivity analysis only make sense once you understand that the optimizer is solving a system of simultaneous equations, not comparing pairs of ingredients in isolation.

The three pieces of data that determine an ingredient's role in the solutionThe three pieces of data that determine an ingredient's role in the solution

For the optimizer to correctly assess an ingredient's value, it needs three precise pieces of data: the ingredient's nutritional composition as it actually is from the supplier the plant uses, the acquisition price per kilogram updated to the moment of formulation, and the stock availability or maximum volume that can be purchased for that production cycle. When any of these three pieces of data is inaccurate or outdated, the calculated formula is locally optimal for data that doesn't reflect reality, and the calculated cost diverges from the actual cost of production.

Proximate composition versus digestible composition: which data feeds the model

One of the most important advances in monogastric formulation in recent decades is the shift from formulating on a total-nutrient basis to formulating on a digestible-nutrient basis, particularly for amino acids. This distinction matters for cost reduction because it changes the relative economic value of ingredients in the model.

Formulating on a total crude protein and total amino acid basis, a meal with 48% crude protein looks superior to one with 44%, and the optimizer tends to favor it. But if the second meal's standardized ileal amino acid digestibility is substantially higher, the digestible amino acid delivered per real spent may be higher for the second ingredient, even though its total concentration is lower. Formulating on a digestible amino acid basis, the optimizer automatically captures this distinction. Formulating on a total amino acid basis, it doesn't see the digestibility difference and may be systematically choosing the wrong ingredient from an economic standpoint.

The practical implication is that shifting to digestible-amino-acid-based formulation, besides improving nutritional precision, often opens room for alternative ingredients that have good digestibility but apparently inferior total composition. Modern-processed canola and sunflower meals, with well-standardized extraction temperature, can have amino acid digestibility competitive with soybean meal under specific processing conditions. Without digestibility data in the model, the optimizer can't capture that value. With it, these ingredients can enter the optimal solution at prices that would otherwise exclude them on a total-nutrient basis.

Metabolizable energy and its influence on energy ingredients

The same logic applies to energy. Formulating using gross energy values instead of metabolizable or net energy means the optimizer isn't correctly capturing how much usable energy each ingredient delivers to the animal. Corn, sorghum, and DDGS have relatively close gross energy values, but their metabolizable energy values for poultry and swine differ significantly and vary with processing and composition. The optimizer working with metabolizable energy evaluates energy ingredients more precisely and can identify less intuitive combinations that produce lower cost with greater actual delivered energy value.

Composition inaccuracy as a hidden cost source

When the formulator uses reference table values for ingredient composition, the model is mathematically precise for an ingredient that doesn't exist: the table's average ingredient. The actual soybean meal going into the mixer has a composition that can differ from the tabulated average depending on the supplier, processing, storage, and grain origin. This gap between the tabulated value and the actual value creates a deviation between the calculated cost and the actual production cost, and between the formulated nutritional level and the level actually delivered to the animal.

If the formulator uses 46% crude protein for soybean meal and the supplier consistently delivers 44%, the optimizer calculates a soybean meal inclusion that isn't enough to reach the actual minimum protein requirement. To compensate for this inaccuracy without having the data, formulators usually add safety margins to the nutritional minimum. That margin has a cost: it requires more protein ingredient per ton of feed to meet a level that exceeds what is biologically necessary. With the supplier's actual analytical data, the necessary margin is much smaller, since it's calibrated to the observed standard deviation, not to an unquantified uncertainty.

The way to get this data is systematic analytical monitoring of received raw materials. With ten to fifteen analyses of the same ingredient from the same supplier, the formulator can calculate the actual average and standard deviation of that supplier's composition for critical nutrients. When this data feeds the local composition in the formulation software, the model starts working with the actual profile of the ingredient the plant uses, not with the average of a population sample far removed from its operational reality.

Calibrating safety margins with analytical data

A safety margin of two standard deviations added to the nutritional minimum guarantees, under normal distribution conditions, that the formula produced meets the required nutritional level in more than 97.5% of batches. For an ingredient whose crude protein standard deviation is 0.7 percentage points per supplier, a safety margin of 1.4 percentage points is statistically robust. A margin of 4 percentage points applied to the same ingredient represents almost three times more than necessary, and that excess has a real cost in every ton produced. Historical analytical data is what allows the margin to be reduced on solid grounds, without compromising compliance.

Shadow price: the map of the formula's pressure points

Shadow price, also called the marginal cost or constraint's optimal price, is calculated by the optimizer for every constraint active in the solution — that is, for every nutritional requirement being met exactly at its minimum or maximum limit. It represents how much the formula's total cost would decrease if that constraint were relaxed by one unit.

A constraint with a high shadow price is a point of economic pressure: the formula is paying dearly to maintain it. If the digestible lysine minimum is active with a shadow price of R$8.50 per percentage point, that means for every 0.01 percentage point reduction in that minimum the formulator could technically justify, the formula's cost would drop by R$0.085 per ton. Multiplied by monthly production volume, that number becomes concrete and justifies the time investment of revisiting the requirement with updated animal performance data.

Systematically reading the shadow price of all active constraints ranks, by economic importance, where it's worth spending analytical effort. In a formula with twenty active constraints, three or four typically account for 80% of the total shadow price. Revisiting those three or four with current nutritional requirement and animal performance data has more impact than reviewing all twenty indiscriminately.

Shadow price of ingredients outside the optimal solution

The same principle applies to ingredients that didn't make it into the formula because they're too expensive relative to the nutritional contribution they offer. For each of these ingredients, the optimizer calculates the optimal inclusion price: the maximum price per kilogram below which that ingredient would become part of the optimal solution. This data is directly usable by the purchasing department as a negotiation reference. If the calculated optimal price for sunflower meal is R$1.15/kg and the supplier is quoting R$1.30/kg, the buyer has an objective reference for where the gap is and how much negotiating room is needed to make the ingredient viable in the formulation. Without this data, negotiation is driven by perception, not evidence.

Parametric sensitivity analysis: from reaction to anticipation

Parametric sensitivity analysis extends point optimization into a range analysis. Instead of calculating the optimal formula only for an ingredient's current price, it calculates the optimal formula for every point in a price-variation range defined by the formulator, generating a curve that shows how total cost and formula composition behave across the entire relevant range of variation.

The most valuable data this analysis produces is the inflection point: the exact price above which the optimizer starts replacing one ingredient with another, and the rate at which that substitution occurs as the price keeps rising. For corn, for example, the analysis might reveal that substitution by sorghum begins when the corn price exceeds R$0.82/kg, and that the substitution is complete within sorghum's inclusion limits when corn reaches R$0.98/kg. These two points define the transition window that the purchasing department should monitor.

When the buyer knows that R$0.82/kg is corn's inflection point, they can set an alert for when the corn quote approaches that value and prepare sorghum quotes in advance. The response window stops being the time to notice the price changed, plus the time to reformulate, plus the time to quote the alternative ingredient. It becomes just the time to get a quote, because the parametric analysis has already been done and the formulator knows exactly what will change in the composition once the price crosses the threshold.

Parametric analysis for inventory planning and MRP

The parametric curve also informs production planning (PCP) about how ingredient consumption will evolve under different price scenarios. If corn prices rise and the formula progressively shifts to sorghum and DDGS, corn consumption falls while sorghum and DDGS consumption rises. Production planning can use the curve's points as planning references: for each corn price level, what is the expected consumption of each ingredient in the production batch? This view reduces the risk of excess stock of an ingredient the formula is about to deprioritize, and of running out of an ingredient the formula will demand in greater volume.

Multi-formulation: cost reduction at the portfolio level

Optimizing each product individually only considers that product's own constraints and finds the lowest cost for it in isolation. Multi-formulation treats the entire portfolio as a single optimization problem and finds the lowest total cost for the production batch, which is often lower than the sum of the individually calculated costs.

The difference occurs because ingredients shared across multiple formulas have an optimized distribution opportunity that individual formulation can't see. If a co-product is available in limited volume and could be included in several formulas in the portfolio, formulating each product separately doesn't distribute that ingredient optimally among the formulas: each individual optimization includes the ingredient without knowing what the other optimizations will do with it. The result can be that some formulas demand more than what's available and others demand less, and the final distribution ends up being done manually with efficiency losses.

In multi-formulation, the co-product availability constraint is global: the total available volume is distributed among all formulas by the optimizer in a way that minimizes the batch's total cost. The solution considers the interactions between formulas and finds the distribution that is optimal for the set, not for each product in isolation. The resulting consolidated consumption is the precise data that production planning uses to plan purchases and that MRP uses to generate production orders.

Stock constraints in multi-formulation

The ability to insert ingredient availability constraints directly into the multi-formulation model has a direct impact on inventory management. Ingredients bought opportunistically on price, which need to be consumed within a specific cycle to avoid excess stock, can be treated as minimum-consumption constraints in the model: the optimizer distributes that ingredient among the portfolio's formulas so as to consume the available volume, while respecting each product's nutritional limits. Production planning doesn't need to manually adjust plans to work through the stock; the optimizer handles that distribution as part of the solution.

Calculated cost versus actual production cost: closing the loop

The cost calculated by the optimizer is an estimate based on the data the model receives. It is precise to the extent the data is precise. Systematically comparing the model's calculated cost with the actual cost recorded in production, cycle by cycle, is the most honest indicator of the quality of the data feeding the software and the reliability of the decisions based on it.

When the deviation between calculated cost and actual cost is consistently small, the formulator can trust the model's indicators to make reformulation, purchasing negotiation, and inventory planning decisions. When the deviation is large and variable, there is inaccuracy in some of the input data: outdated prices, compositions based on tables without local data, constraints that don't reflect the reality of the production process. Identifying and correcting that inaccuracy is the work of calibrating the model, and it is just as important as choosing the optimization algorithm.

Formulation software that integrates lab data through a quality management system, like the flow between Labinfy and Formulamix, and that receives prices directly from purchasing without manual intermediation, structurally reduces sources of inaccuracy. Input data reaches the model by the shortest possible path, with the smallest lag window between the actual event and the update in the model. The smaller that window, the closer the calculated cost is to the actual cost, and the more reliable the whole chain of decisions the model informs.