How Grammarly Serves 5 Billion Daily Events in 15 Minutes with Databricks

Grammarly's mission is to improve lives by improving communication—and its writing assistance platform now serves 30 million people and 50,000 teams worldwide. Every suggestion accepted, rejected, or ignored generates an event, totaling roughly 5 billion events per day. Managing and analyzing that data at scale became the company's defining infrastructure challenge.

Grammarly had built its own analytics platform, but it required a custom SQL-like language that was time-intensive to learn and slowed new hire onboarding. As the company grew, the platform couldn't ingest external data sources, couldn't connect to Tableau, and ran large Amazon EMR clusters around the clock. Each team began solving its own analytics needs independently, creating data silos and making it hard to know which dataset was correct. "Every team decided to solve their analytics needs in the best way they saw fit," said Chris Locklin, Engineering Manager for Data Platforms. "That created challenges in consistency and knowing which data set was correct."

Grammarly evaluated multiple vendors—including Snowflake, which fell short on data science and ML capabilities, unpredictable costs at scale, and crucially, didn't allow complete ownership of company data. Databricks won on all three dimensions. The company migrated to the lakehouse architecture with Delta Lake at the core, Databricks SQL for querying and Tableau integration, and Unity Catalog for fine-grained access control and end-to-end data lineage. Security was non-negotiable: just as Grammarly ensures users' data always remains theirs, the company required the same ownership over its own analytical data.

The results were immediate. Querying is now 110% faster at 10% of the previous ingestion cost. The 5 billion daily events that previously took four hours to make available for analysis are now ready in under 15 minutes, allowing teams to get rapid feedback on new feature rollouts and understand adoption as it happens. Marketing can now consolidate data from advertising platforms, extrapolate lifetime value, and compare it against acquisition costs in a single environment. Unity Catalog solved the distributed access control problem at scale—giving analysts the flexibility they needed while enforcing usage standards and data lineage.

The broader shift is cultural. "Doing everything from ETL and engineering to analytics and ML under the same umbrella removes barriers and makes it easy for everyone to work with the data and each other," said Sergey Blanket, Head of Business Intelligence. Migrating off rigid legacy infrastructure gave Grammarly the adaptability to evolve alongside its data needs—and the confidence that the platform will scale with the company.