An Amazon Data Firehose stream (named Amazon Kinesis Data Firehose in older material) receives JSON sensor readings and delivers them to Amazon S3, where analysts query them with Amazon Athena. Because the objects are stored as JSON, every Athena query reads every field of every record and the scan costs keep rising. The team wants the data to land in S3 as Apache Parquet instead, and it does not want to write or maintain any transformation code.