In full, the ELTP model encompasses the following steps:
- Extract data from raw data sources.
- Load data into durable storage, such as a data warehouse or data lake.
- Transform raw data into new datasets.
- Publish data to downstream users and business applications.
This design preserves the benefits of ELT while also giving a framework for ensuring that data is efficiently published to downstream consumers and applications.
# Publish Destinations are more than Reverse ETL
All Reverse ETL destinations are Publish-type destinations, but not all Publish destinations are Reverse ETL. The term “Reverse ETL” can still be useful when communicating the unique challenges of publishing to a SaaS system like Salesforce and Hubspot, but the term falls down when describing a publish flow to a partner’s SFTP site or to an external database. Whether publishing to Salesforce, SFTP, S3, or SharePoint, the Publish framing offers a more robust and holistic architectural model that is not tied to the “pickiness” of API-type destinations.
ELTP: Extending ELT for Modern AI and Analytics | Airbyte