Case Study: Inneractive achieves on-demand scale and cost-efficient ML for ad analytics with Databricks

Data-driven Ad Buying at Scale with Databricks

Inneractive is a global mobile ad exchange (INNEX) that processes massive real-time traffic—3–5 million requests per minute and billions of rows daily—serving native and video ads. Their relational/BI systems and traditional data warehouses could not scale to handle ~240+ GB of raw data per day, causing slow or stalled queries that delayed reporting and risked impacting real-time ad serving.

Inneractive migrated to Databricks on AWS, using Spark with Parquet on S3 and Databricks’ cluster manager to decouple compute and storage and scale on-demand. The platform sped up ingestion, querying and feature extraction, enabled distributed model training with MLlib (e.g., logistic regression), freed engineers from infrastructure work, and delivered better cost-efficiency and scalability—allowing timely reports and improved ad-buying optimization.

Inneractive

Richard Grossman

System Architect