Case Study: Lyft achieves 30% productivity gains and faster data discovery with Neo4j

Lyft Speeds Up Data Discovery with Tool Using Neo4j

Lyft, the San Francisco–based ride‑hailing company, relied heavily on data to drive and evaluate decisions, but rapid growth—about 10 petabytes across thousands of tables—and an expanding user base made finding and understanding the right datasets slow and inefficient. Data discovery often ate up roughly a third of data scientists’ time as users navigated similar table names, asked colleagues, or pulled sample rows to understand contents.

To fix this, Lyft built Amundsen: a microservice metadata platform that combines search (Elasticsearch for relevance and popularity), lineage, and network‑based discovery with Neo4j as the editable metadata graph and source of truth. The tool achieved 90% weekly adoption among data scientists, boosted data science productivity by about 30%, earned an 8.5/10 CSAT score, extended use across engineers and product teams, and was open‑sourced for broader community contribution.

Lyft

Tamika Tannis

Software Engineer