Lipton Teas and Infusions is the world's largest tea company, serving 250 million consumers daily in over 100 countries. The company has built many production use cases on Databricks, but one of the most demanding sits on the factory floor, where operators need live production data they can view and correct in real-time. Analytics and operations ran in separate systems, forcing data duplication and constant workarounds. With Lakebase, Lipton unified both on one platform, cutting storage costs 30%, accelerating changes from weeks to days and opening a path to replace costly SaaS tools.
Factory data needed to move at operator speed
Since becoming independent from Unilever, Lipton has built use cases on Databricks across finance, supply chain, and manufacturing business functions and chatbot solutions. On the factory floor, machine data streams into Databricks to track output, errors, downtime, overall equipment effectiveness and line losses. Operators needed more than a live view.
"An operator might know that a machine stoppage tagged as a breakdown was actually planned maintenance," said Dmitry Ratushnyak, Head of Data and AI Platform at Lipton Teas and Infusions. "They need to correct it and see the KPI impact instantly. Operators are busy and they don't want to wait."
That seemingly small human-in-the-loop change created a difficult technical problem. Lipton ran its live operational workflow in one database and its analytics, data science and predictive maintenance in Databricks. The same data had to be maintained in both places, raising costs and increasing fragility. The operational database was never designed to support application-style updates, so every correction required workarounds.
Even small changes could take weeks to deliver, and the architecture required specialized skills, making hiring and onboarding harder as the team grew. "The setup worked, but we were outgrowing it," Dmitry said. "We needed everything on one platform so we could move faster and build around skills the team already had." Dmitry partnered with Rahul Singh, Lipton’s Global Manufacturing D&T Transformation Manager, who initiated this project from the manufacturing and infrastructure perspective.
One platform for operations, analytics and governance
Lipton redesigned the workflow around the operator, not the workaround. Factory data flows from Kepware through Event Hub into Databricks, where:
- Spark Structured Streaming continuously calculates production metrics and KPIs.
- Lakebase serves as the transactional backend for the web app operators use on the line, delivering sub-second write performance so corrections register instantly.
- Operator updates flow into Lakebase, recalculate downstream KPIs and sync back to the data lake for historical analysis.
- Unity Catalog keeps governance unified, so the same policies protecting analytical data also apply to operational records.
Platform engineers use Declarative Automation Bundles to set up infrastructure, tables and security in one step, while developers work with familiar SQL, Python and standard Postgres patterns. Lakebase branching proved to be a major accelerator, letting teams develop application code and database changes together in isolation, test against production data and merge when ready, all without disrupting live workloads.
“One engineer built a fully working app on Lakebase in two evenings, complete with CI/CD and deployment,” Dmitry said. “When prototyping is that fast, you start asking different questions about what to build and what to buy.”
Lower costs, faster builds and a new approach to SaaS replacement
Once Lipton stopped duplicating the same streaming data across two systems, the architecture became both cheaper and simpler. The team saw measurable improvements across cost, speed and hiring:
- Storage costs dropped by at least 30 percent.
- Compute costs fell by roughly 15–20 percent.
- Application changes that once took about two weeks now ship in a couple of days.
- The team can hire for common SQL and Python skills rather than niche tooling, which makes onboarding faster and talent pipelines broader.
That speed is reshaping how Lipton evaluates build-versus-buy decisions. Faster prototyping gives the team a practical way to test whether focused internal tools should replace expensive SaaS products. For one application alone, the company sees a $500K+ annual cost avoidance opportunity.
Three more Databricks applications are already close to production, including a logistics control tower that consolidates manufacturing data, machine manuals and maintenance requests into a single operational portal. Genie-powered chatbots now handle questions dashboards never could, from payment statuses in accounts payable to purchase request policies.
Looking ahead, the team envisions Lakebase as the backbone for all data applications at Lipton, including a persistent memory layer for AI agents that can retain context across sessions.
“Our goal is to improve every business process in the company with data,” Dmitry said. “Lakebase makes it possible for a small team to build what once required enterprise-scale investment.”