Currently, Fivetran stores ingested files in a delta-style structure, which works well for replication purposes. However, it would be extremely valuable to have an optional setting across connectors to organize files using date-based partitioning.

For example, instead of only storing files in a delta format, allow customers to define a structure such as:

source/year/month/day/files.parquet

or

source=<source_name>/year=/month=/day=/files.parquet

This would provide major benefits for downstream processing and analytics platforms such as Spark, Athena, Redshift Spectrum, Snowflake, Databricks, and dbt workflows, improving:

• Query performance

• Partition pruning

• Lifecycle management

• Cost optimization

• Data lake organization standards

Ideally, this could be an optional configuration at the connector or destination level.