Connector

A connector is a prebuilt piece of software that knows how to extract data from, or load data into, a specific source or destination, a Salesforce connector, a Postgres connector, a Google Ads connector, so an engineer doesn't have to write authentication, pagination and error-handling logic from scratch for every integration. Platforms like Fivetran, Airbyte and Stitch are largely libraries of connectors managed through a common interface.

A connector typically wraps an API integration or database driver behind a standard interface: configure credentials and a schedule, and the connector handles extracting records, detecting schema changes, and tracking what has already been synced so the next run is incremental rather than starting over. This differs from writing a one-off custom script for a single source, which offers more control but has to be maintained, monitored and updated by hand as the source's API evolves.

Connectors matter because most organizations need data from dozens of different tools, and building and maintaining bespoke integrations for each one doesn't scale. They are the standard building block of data ingestion in modern ETL and Reverse ETL platforms. The main pitfall is assuming a generic connector perfectly matches a specific use case: subtle differences in how a connector paginates, handles deletes, or represents nested fields can silently produce incomplete or duplicated data if not checked.