The challenge
A large North American beef producer with cattle head count in excess of 500,000 wanted to mine and potentially monetize their extensive data assets relating to genomic and expected progeny differences (EPD), feedlot operations, cattle and carcass performance, dairy and calf ranch operations, marketing and more.
The client wanted to uncover potential correlations between cattle phenotypical and production data for carcass performance improvements, particularly whether any additional factors have significant impact on average daily gain (ADG). They also wanted to identify disease resistance attributes in animals in order to increase production efficiency by decreasing mortality and morbidity incidents.
Finally, they were interested in identifying other possible statistical correlations to further monetize their data assets.
In summary, the client wanted to use AI to disrupt the conventional business model in order to position their business to be more proactive, productive, economical and competitive.
The TELUS Digital solution
To design, build and deliver this project, we used an AI technology stack from Google Cloud Platform (GCP). The platform comes with cutting-edge tools to handle complex challenges related to the AI pipeline. From data ingestion, preparation, storage and exploration, to training and generating insights, it supports the complete journey of the AI loop.
This project brought forth some unique challenges. First, cattle prices fluctuate significantly with supply, drought and feed costs, making the beef industry highly cyclical in nature. As this influences procurement and feeding strategies, TELUS Digital made certain assumptions for modeling engagement purposes. These included healthcare protocols and feedlot operations, assuming consistency across all feedlots dispersed over diverse geographies.
Further, factoring genomic, phenotypical, calf ranch, cattle and carcass datasets into system modeling with conventional technologies is not only tedious, but often unreliable. Using AI, we were able to provide techniques to simultaneously evaluate multiple factors and their interactions. AI algorithms helped to efficiently rank critical factors that impact profitability, as well as herd health and quality. The client also requested extensive handcrafting of their production data features, a process called feature engineering.
The results
By analyzing the datasets and employing various AI models, we identified the critical traits that define and differentiate outperformers (as well as laggers) of ADG. This profiling helped to correlate cattle phenotypical and production data to generate insights for performance improvements and procurement strategies.
Our AI models efficiently profiled attributes that directly influence the carcass quality grade. From an animal health perspective, we assimilated the common denominators for morbidity and mortality rates, pointing to frequent disease encounters and outliers. These data points can be further explored to effectively administer preventive interventions.
Based on these machine insights, TELUS Digital was able to make actionable recommendations to improve performance, productivity and economic factors, including animal breed preferences, calf ranch catalysts, cattle selection and intake criteria, vendor and feedlot outperformers, feedlot allocation and more.