Case Study: Large Retailer achieves faster, more reliable pollution forecasting from nested JSON with Rockset

Large Retailer - Customer Case Study

Large Retailer data engineer Doug Balog set out to combine hourly National Weather Service forecasts (scraped for about 100 points in Allegheny County) with crowdsourced pollution reports to forecast inversion-driven pollution events and warn vulnerable residents. He faced nested JSON, schema drift and unexpected field values (e.g., “20G30” for gusting winds) that made traditional relational databases and ETL impractical.

Using Rockset, Doug ingested the raw nested JSON with no schema or data preparation, created Rockset collections and accessed the data via a reliable SQL API, and handled unanticipated field types without errors. Rockset dramatically reduced setup and maintenance effort, saved significant development time compared with alternative approaches, and enabled Doug to provide query access for researchers and to train ML models to predict local pollution levels.