Case Study: Ibotta achieves a petabyte-scale, self-service datalake and faster ML & analytics with Qubole

Ibotta Builds A Self-Service Data Lake To Enable Business Growth

Ibotta is a mobile cashback app that had grown from tens of terabytes to nearly a petabyte of data as it added first‑party features, ML models and richer partner analytics. That rapid growth—over 70x since 2017 with more than 20 TB arriving daily—exposed limits in its cloud data warehouse (Redshift), where compute was tied to storage, costs became prohibitive, and teams (data science, analytics, engineering) lacked self‑service access to the data needed to build product features.

Ibotta built a cloud data lake on S3 using Qubole as the compute and operations layer, running Spark, Hive and Presto with Airflow orchestration, autoscaling and heavy use of EC2 Spot instances to separate compute from storage and enable self‑service. The platform tripled processed data within four months, handled ~30k queries/week, enabled rapid delivery of ML features (recommendations, A/B testing, item classification) and produced large cost savings—an estimated $1.2M saved with ~$270k spent over the measured period, and 70–80% reductions in big‑data EC2 costs.

Ibotta

David McGarry

Director of Data Science