Case Study: Zip Co achieves a 65% reduction in Spark cluster costs and faster cluster startup with Databricks

Zip Co - Customer Case Study

Zip Co, an Australian payment and credit solutions provider, needed to model large volumes of financial data (bank statements, transactions, etc.) to predict credit risk, but complex ETL architectures and heavy DevOps requirements slowed time-to-insight and raised costs. Their EMR-based pipelines were costly, error-prone during auto-scaling, and data preparation for machine learning was resource- and time-intensive.

By fully migrating interactive and automated workloads to Databricks, Zip Co. automated cluster management and enabled autoscaling, Spot Instances, and auto-shutdown, giving data scientists faster self-serve access to compute. Within four months they cut total Spark cluster costs by 65%, reduced cluster spin-up times to under three minutes (vs. 12–20 minutes on EMR), and maximized compute utilization while minimizing spend on idle resources.

Zip Co

Yiting Shan

Big Data Engineer