Agri Feed Mix Optimization – TrueNorth Group

Project Information

Client Name: HealthCare

Techniques: Woe LR

Technology: Power Bi
Azure Databricks
Synapse

Business Benefits Improved Patient Care
Increaed Productivity
Reduction in cost associated with unnecessary staff

Industry:

Agriculture / CPG

Techniques:

GMM, K-Means, Monte Carlo Slim, Optimization, Champion–Challenger Testing

Technology Stack:

Power BI, Azure Databricks

Business Benefits:

Decrease in Feed Conversion Ratio

Reduced feed input cost

Increased egg production

Ready to unlock the same results?

Talk to Our Experts

Overview

TrueNorth partnered with an agricultural CPG client to develop a data-driven feed optimization model aimed at improving the Feed Conversion Ratio (FCR) – a critical performance metric for livestock efficiency.

By leveraging unsupervised machine learning and optimization algorithms, the team identified the most effective feed formulations for different breeds, regions, and environmental conditions.

The Challenge

The client’s feed formulations were based largely on historical averages and manual experimentation.

This approach made it difficult to:

Accurately determine optimal feed composition per region or breed

Adjust for variations in weather and life stage

Minimize feed input costs without compromising productivity

To overcome this, TrueNorth needed to design a solution that could automatically cluster behavioral patterns and recommend the most cost-effective feed mix under varying conditions.

Our Approach

TrueNorth implemented a machine learning optimization pipeline that combined clustering analysis, simulation modeling, and feedback-driven testing.

Cluster Identification via Unsupervised Learning

Used GMM and K-Means clustering to segment livestock based on breed, region, life stage, and weather - identifying distinct performance groups.

Feed Mix Optimization

Developed multiple feed composition scenarios (e.g., different corn-soy-mineral-moisture ratios) and evaluated them using Monte Carlo Slim and optimization algorithms to determine the feasible region for maximum FCR efficiency.

Champion–Challenger Testing Loop

Ran controlled experiments to compare baseline feed mixes (“Champion”) against optimized formulations (“Challengers”), creating a feedback loop that continuously improved accuracy and yield predictions.

Collect & Correlate

Collected and structured key datasets, including:

Demographics and population health data

Epidemiology and diagnosis statistics

Historical patient volumes and level-of-care requirements