Follow Datanami:
April 11, 2019

From Big Beer to Big Data: Inside AB InBev’s Digital Transformation

(meunierd/Shutterstock)

With more than 500 beer brands and $55 billion in sales, Anheuser-Busch InBev is already the world’s biggest beer company. And if all goes as planned with its digital transformation project, it will be the best beer company in the world, too.

AB InBev is in the middle of a massive project to turn itself from an amalgamation of dozens of independent breweries — some of them more than 100 years old — into a unified powerhouse that uses data to enhance business processes, improve consumer relations, and make its beer better too.

The project involves consolidating not only its core transactional systems, but its global data and analytics infrastructure. On the ERP front, it’s in the midst of standardizing from 27 different ERP systems used at subsidiaries around the world to a single system, which is SAP‘s S/4.

The analytics side of the house is a bit further ahead. Beginning from a white board in March 2018, AB InBev has successfully designed, built, and rolled-out an enterprise data hub (EDH) atop a Hadoop distribution running in Microsoft Azure. The company is already powering various reporting, analytics, and data science endeavors in the cloud, and plans to expand those activities in the years to come.

One of the company’s core architectures principals was to be cloud first, Harinder Singh, director of data strategy for AB InBev, told Datanami at the recent Strata Data Conference, where Singh also presented a session on the company’s data journey.

“We have a three-to-five-year plan to be mostly on cloud,” Singh said. “As a company, we’ll never go 100% cloud. There are some crown jewels we have to protect and maintain on prem. But in the data journey, we are 100% on cloud, except for our source systems.”

Analytics in the Cloud

AB InBev’s EDH consists of a mix of Microsoft and third-party tools, all sitting in the Azure cloud. Azure Blob Storage serves as a landing zone for all the raw data that arrives in the cloud. The next layer of the EDH is based on Azure Data Lake Storage (ALDS), which provides the file-level security that AB InBev required.

Seven of the world’s 10 top-selling beers are brewed by AB InBev (Courtesy Harinder Singh)

It uses HDInsight, which a Hadoop and Spark environment that’s based on Hortonworks Distribution of Hadoop (now owned by Cloudera), to process less structured data. Relational database technologies, including SQL Server and SQL Data Warehouse, are used for crunching structured data.

The company’s EDH consists of several layers of data, including layers for raw data, clean data, and optimized data. A data catalog provides data scientists and analysts with a way to find and access the reverent pieces of data, whether it’s structured or unstructured, first-party or third-party data, Singh said.

The company has also carved out a spot in its EDH where data scientists can play using the technologies and languages of their choice. “We call it the analytics sandbox,” Singh says. “For example, if you’re trying to build a correlation between weather and sales, you’d pick the sales data from the EDW [enterprise data warehouse] layer, then pull the weather data from the lake, and combine the two together for your analytics.”

Simplified Data Pipelines

AB InBev used to have dozens of ETL and data integration tools, but it standardized on Talend to land terabytes worth of data into the EDH every day.

The company uses Talend Data Fabric to speed up with process of building data pipelines that connect to various data sources, including the various ERP systems, databases, file systems, and IoT sensors installed at breweries around the world.

Talend Data Fabric provides an abstraction layer atop data sources for AB InBev

AB InBev’s data engineers appreciate the breadth of connections that Talend brings, Singh said.

“It has 900-plus connectors, so in most cases, we’ll find a connector out of the box,” he said. “There have been instances – two that I can think of – where, even with all the connectors, there wasn’t one available, because a particular system was a homegrown system. But because Talend also generates code, we’re able to take the codebase out of an existing connector, customize it ourselves, and use it for the homegrown system.”

The reusability of the pipelines in Talend Data Fabric is another plus. Instead of building one-off pipelines to pull data from its existing data warehouses, like Teradata, Vertica, or Oracle data warehouse – or an ERP system, like SAP, ProMax, or SYSPRO – Talend’s software lets AB InBev’s data engineers reconfigure an existing pipeline for the new data source, which shortens development time.

“We have so many data systems. It will take us many, many years if we build one pipeline per system,” Singh said. “So the biggest part that we have done as a company, as a team, is we have built these reusable modules. We just plug and play and make minor adjustments along the way.”

Analyics Use Cases

The company’s data analytics infrastructure supports a variety of use cases, ranging from demand forecasting and fraud detection to social media listening and IoT analytics.  The EDH is maintained centrally via its global analytics center (GAC), which allows local teams of analysts to tap into the capability in a hub-and-spoke manner.

The supply chain behind AB InBev’s operations is massive and complex. It must source all the raw materials that go into the beer, and then distribute pallets of the finished products to customers in a timely manner. A disruption in one aspect of the supply chain can put a crimp in sales. Likewise, over-producing a product with limited shelf-life is not conducive to profitability, so AB InBev seeks to match supply with demand as best as it can.

AB InBev uses big data and machine learning to optimize demand forecasts, thereby optimizing its supply chain (Don Pablo/Shutterstock)

“If you can optimize demand forecasting, even by just 1 %, because we have a more accurate forecast, we can actually manage our supply chain accordingly,” Singh said. “For example, a year ago, when we had a big hurricane [Harvey], AB InBev stopped one of their breweries [from making beer] and used the same process, except we filled the cans with water.  That is possible because you’re able to make changes in real time.”

The company has an operation analytics project, where it uses to data to find ways to optimize and automate certain business processes. That includes monitoring data coming off brewery equipment. “The data generated by these machines, they all have the same indicators,” Singh said. “So if you can do it for one brewery, we can essentially scale it to all breweries. “

The company also does people analytics. “One of the [goals] of our company is to keep employees challenged and happy, and one of the best way is to look at the data, so that’s where people analytics comes in,” he said.

The data infrastructure also supports a compliance initiative to ensure it’s complying with laws. Even when you work for a beer company, fraud is possible, so the cloud infrastructure is used to hunt down fraudsters.

Another IoT use case involved tracking the life of a six-pack. It installed RFID labels into the containers, and tracked where it went, from the brewery to the distribution center.

AB InBev also tracks mentions of its products on social media for the purpose of connecting with customers, as well as detecting potential problems. “At the end of the day, people love our product and they recognize us with our brand, so brand reputation is very important to us,” Singh said.

Data-Based Journey

The company has made considerable progress on its digital transformation journey since Singh joined AB InBev from Walmart a year-and-a-half ago. Under Singh’s guidance, AB InBev’s team of 80 engineers have completed more than 30 data projects, and third-party vendors have completed a handful of additional projects under its marketplace.

Now that the EDH is sitting in the cloud and data silos are being effectively quashed, the company is well-positioned to tackle additional data projects and continue the company’s digital transformation journey.

“We want to carry the legacy and built the company for another 100-plus years and beyond, and In the process also increase revenue,” he said. “Right now we’re $55 billion. We want to get to $100 billion. All that requires digital transformation. To do that, you have to move to the cloud. You have to be able to unify your data. You have to be able to do better analytics.

“Our dream is to be the best beer company in the world,” Singh continued. “We are already the biggest, but we also want to continue to be the best.”

Related Items:

AI Hype Rockets, Hadoop Twins, and Other Learnings from Strata

How Walmart Uses Nvidia GPUs for Better Demand Forecasting

Big Data Fabrics Emerge to Ease Hadoop Pain

Datanami