Bayer, the life-sciences company active in more than 100 markets across some 83 countries, has described how its Consumer Health division rebuilt a fragmented analytics estate into a single, governed data platform using Databricks and its Unity Catalog governance layer. The goal, according to the company, was to give a large, globally distributed workforce secure, searchable access to data and to enable self-service analytics without the silos that had slowed earlier decision-making.
The starting problem: fragmentation and “data tourism”
As a globally distributed organisation, Bayer Consumer Health had previously run analytics across markets that each used a different technology stack, often in different data centres. When data needed to be shared, it was frequently copied — sometimes several times over — a pattern the company calls “data tourism.” The result was higher data-management costs, duplicated copies of uncertain provenance, and slow delivery of new solutions. André Wuthenow, Principal Cloud Platform Architect at Bayer, has said the earlier setup made it difficult to apply machine learning effectively because each component sat on a different stack in a different environment.
The approach
The division’s data and analytics group set out to build a global, scalable platform. By its own account it serves more than 2,000 business users across some 25 zones in three global regions, supported by a team of more than 250 machine-learning and data engineers. The brief was a cloud-based system that would use serverless technology where possible, scale to any data volume and concurrent user count, and remain cost-disciplined by charging only for what was used — while allowing new services to be trialled at small scale before a wider rollout.
To deliver that, Bayer reports building template-based environments with dedicated service instances for resource isolation and lifecycle management, with Unity Catalog supplying a centralised governance and metadata layer across them. That layer lets teams govern core data assets once and then share and reuse them securely across projects and regions.
From push to pull
Replacing an older Hive metastore with Unity Catalog allowed Bayer to shift from a push-based to a pull-based model of data sharing. Rather than copying datasets between environments, each data-domain team can decide what to share and with whom, and consumers simply request permission to access governed, trusted core assets. Combining serverless compute with Unity Catalog also let engineers build against production-grade data from their development environment, which the company links to faster delivery of analytics solutions.
The broader principle is that governance is built into the platform rather than bolted on afterward, with shared standards for access, naming and security applied while individual teams retain flexibility for their own markets. In Wuthenow’s words: “With Databricks, we are building reusable core assets, enabling self-service analytics and fostering a data-driven organization that provides insights for all and data silos for none.”
How to read a vendor case study
It is worth keeping the source in mind. This account originates from Databricks’ own blog and reflects a customer that adopted its platform, so the figures and outcomes are self-reported rather than independently audited, and the narrative naturally favours the chosen tools. The underlying lessons, however, are vendor-neutral: consolidating duplicated data, governing assets centrally, and moving from copying data to granting access to a single trusted source are widely recognised goals in enterprise data management. Organisations evaluating a similar move are best served by mapping these principles to their own constraints — existing systems, regulatory requirements and cost — rather than assuming one company’s results will transfer directly. The same enterprise-data foundations increasingly underpin newer AI tools, such as the retrieval agents covered in this look at enterprise knowledge agents.
The full case study is available on the Databricks blog.
