BMW Group Brings Extra Horsepower to its Operational Data Workloads With Snowflake’s AI Data Cloud

BMW Group is renowned for innovation and delivering quality. That doesn’t just apply to the cars coming off the production line; it’s part of the organization’s DNA. Since 2018, BMW’s data team has embarked on a radical program, defined by its Cloud Data Hub (CDH), to bring together data-driven insights from across the business.

The guiding principle behind the CDH has always been to deliver a decentralized set of best-of-breed data tools, all supported by a central data governance model that maintains consistency and quality.

Given the vast volume and variety of data the company collects — some 6,000 data sets and products across 15 BMW domains (including service, manufacturing and post-sales processes) — the CDH needed the performance and scalability to serve the more than 10,000 users that access this data every month.. 

To empower those users with faster compute across large operational analytics workloads — while ensuring seamless integration across the CDH ecosystem — BMW Group needed a cost-effective platform that could handle large data sets and complex workloads within a strict governance framework. It found all that and more with Snowflake’s AI Data Cloud.

For many of the use cases in CDH, BMW relies on a mix of AWS infrastructure and Apache Iceberg tables to deliver high performance and flexibility. “We have a strong partnership with Amazon Web Services, and AWS tools meet the needs of the majority of our use cases,” explains Ruben Simon, Head of Product Management, Cloud Data Hub at BMW Group. 

However, the CDH team found some use cases, particularly those involving large operational datasets, could benefit from a different data approach. “For operational use cases, we needed very fast compute, and Snowflake came out ahead of the competition. We decided to integrate Snowflake into the CDH for this reason.”

Now with the AI Data Cloud as a key component of its data foundation, BMW Group can deliver high levels of performance for large data workloads. For example, the company’s service data — housed in a massive data lake containing vehicle service and repair transaction data from over the last decade — now runs on Snowflake. With Snowflake’s compute power behind it, dealerships and garages can almost instantly bring up service histories for customers. This means dealership staff can get work more efficiently and get the information they need faster, and customers get speedier services and repairs. 

While Iceberg offers a better cost-benefit balance for simple workloads on single tables, the CDH team has quickly found that Snowflake can deliver high performance at a lower cost on more complex workloads. “As it’s a premium compute resource, we assumed Snowflake would be more expensive, but we’re actually saving costs on heavy-compute workloads,” says Ruben Simon. On its service data workloads, BMW Group has reduced average costs by 25% with Snowflake compared to its legacy platform.

Snowflake also offers a simple interface and easy configuration for users. While AWS tools require users to configure roles, Snowflake instances can be deployed in just a few clicks. The CDH team worked closely with Snowflake to further lower the barrier to entry in the AI Data Cloud and make it easy for users to get started. This may explain why the uptake has been so rapid, with more than 60 CDH use cases — spanning sales and service to supply chain — up and running on Snowflake in just 18 months.

Any tool integrated into the CDH must meet BMW’s governance framework and its ethos around how data is stored and managed. Snowflake was no exception.

“Within our data governance framework, any data on our platform must be transparent,” explains Khalid Al-Khalili. “If a data steward allows access to a data set on one system, that permission must also carry over to other platforms — whether that’s AWS or Snowflake.”

Practically speaking, this meant that Snowflake had to be able to read and write data across the CDH and see identities across BMW’s internal systems and AWS. This level of integration across the entire data estate might seem daunting, but BMW was able to bring Snowflake into its complex, multi-vendor ecosystem with relative ease over 18 months, using a team of just six dedicated Snowflake specialists and counterparts in the CDH team. 

Thanks to their efforts, and Snowflake’s own history of integrating with other technologies, Snowflake supports several other tools that are regularly used across BMW. “Everything has to feel seamless for our users,” Ruben Simon says. “Snowflake’s ability to bring compute to us has made that a reality. It’s what made us realize Snowflake would be a great partner, as we can store data on AWS, Iceberg or anywhere else and still get the compute benefits of Snowflake.”

BMW Group has plans to continue evolving the CDH. It’s looking to combine structured and unstructured data, find suitable generative AI use cases and transition to an open lineage structure.

It’s also planning to bring multicloud functionality to the core of the CDH, while ensuring everything operates within its data governance framework. Snowflake’s seamless integration capabilities will play a part in supporting the car manufacturer’s most intense workloads — no matter where the data is stored.

“We’ve always wanted to avoid vendor lock-in,” says Ruben Simon. “Snowflake’s ability to integrate with even our open source tools really impressed us. Even with our existing infrastructure commitments, Snowflake has proven to be worthwhile due to its flexible integrations.”