BRF transforms data, democratizes AI to improve business outcomes with Azure Machine Learning

Nourishing a population

BRF helps keep a lot of people fed. With 100,000 employees and 20,000 suppliers, BRF is one of the largest food companies in the world. It sells many food products, such as pasta, sauce, and frozen vegetables, with a focus on raising poultry and pork. Approximately 10,000 integrated producers work with the company, which has 350,000 clients in 117 countries around the world.

The company’s supply chain is unpredictable because BRF primarily works with commodities. Prices fluctuate and weather is erratic. For example, during a drought, corn prices might go up, affecting the price of meat production. The company needed to better predict prices and forecast the amount of saleable meat that various animals would produce. BRF was on a path to becoming a more data-driven company and wanted to use AI to tightly monitor its supply chain, reduce food waste in manufacturing, produce food more sustainably, and personalize recommendations for customers.

Hungry for change

BRF started a Center of Excellence (COE) in advanced analytics within its IT team. The COE spearheaded the adoption of Microsoft Azure Machine Learning to develop and train machine learning models to improve business outcomes and sales forecasts. “We selected Azure as our best option because of the quality of the algorithms and how flexibly it works with other solutions that we have and because it’s friendly to our analysts,” says Wellington Monteiro, Global Data Science Chapter Lead at BRF.

Part of the reason BRF chose Azure was because it is easy for a range of business users to work with. The company has a team of analysts who look for purchasing patterns of customers such as grocery stores, mini-marts, and bakeries. Then the analysts generate and share with the sales team personalized recommendations for each customer. In Brazil, the company offers more than 650 SKUs, so this can be a complex process. The analysts previously spent a lot of time and manual effort to get the insights they needed.

“Before, getting insights took an average of 10 days, and we couldn’t deliver business information to our sales staff in a timely manner,” says Alexandre Biazin, Technology Executive Manager at BRF. “With Azure Machine Learning, we’re bridging different user groups, and skill sets, so they can gain quick access to valuable data.”

Serving up automation

To automate the company’s machine learning capabilities and build reproducible processes, BRF data science and data engineering teams implemented three existing models using automated machine learning capabilities in Azure in just three months. The COE also uses machine learning operations (MLOps) capabilities in Azure to automate all stages of its model development, from training to deployment.

In early 2021, BRF began running a pilot test of a recommendation system for the analysts with seven of its sales organizations, representing around 70 percent of the company’s annual revenue. The data science team performed calculations considering the city data, purchasing history, and similarities across several customers, then created the recommendation engine. “We launched the first version of our recommendation system with Azure Machine Learning in just a week or two,” says Monteiro. “We really liked the drag-and-drop design because we could very quickly tweak several datasets, create algorithms, and make the recommendation system available to our sales staff.”

Adds Biazin, “We’re scaling with automated machine learning in Azure and MLOps capabilities in Azure Machine Learning so that our 15 analysts can focus on more strategic tasks instead of the mechanics of merging spreadsheets and running analyses.”

Trusting the ingredients

BRF transformed swiftly and deepened the data science practice across the COE and various business teams, improving collaboration across the teams. The COE ensured that the insights were as explainable and transparent as possible instead of coming from an opaque environment. That way, business users had more visibility into why the models produced certain outcomes.

“We can customize the scoring model with MLOps and include explainable AI features if needed,” says Biazin. “We think that ability to explain is very important and can help us make sure that the business stakeholders trust our models. We have a complex technical architecture, but we seamlessly connect several components and get very good, fast results.”

Sweet taste of success

As a result of agile, trustworthy machine learning accessible across business units, BRF democratized and broadened the reach of AI across the company. It reduced the time needed to deploy new models, which has increased productivity. BRF is accomplishing its aim to monitor its supply chain, forecast products, and gain insights into customer purchasing habits.

With the recommendation engine, BRF is already seeing revenue growth. “The more our customers adhere to our recommendations from Azure Machine Learning, the more we all profit,” says Biazin. “When customers have followed at least 70 percent of our recommendations, we’ve gained a statistically significant improvement in sales.”

Concludes Monteiro, “For many of our customers, BRF is synonymous with quality. To increase that quality, we must embrace technical innovation to grow while also keeping in mind environmental responsibility and resource consumption in our complex supply chain. Adopting Azure was part of our commitment to deliver even better quality to consumers.”

Find out more about BRF on Twitter, Facebook, YouTube, Instagram, and LinkedIn.