Data connectors are ready-made pieces of code, each of which knows how to pull data out of one particular app or database and load it somewhere else, usually a data warehouse. With a connector, nobody has to write that part of a pipeline by hand. They are the extract and load steps of ETL and ELT.
Why is pulling data out hard? Every app exposes its data differently. Most do it through an API, a set of requests that the app accepts from other programs. Each app's API has its own way of logging in. It returns data a page at a time. It limits how many requests a program may send each second. And it changes from time to time, when the app adds or renames a field. A data team that copies data from 30 apps would have to write and maintain 30 pieces of code. A connector seller writes each one once, keeps it working, and sells it to every customer.
A connector handles five jobs. It logs in, usually with a token that the app's owner approves once. It reads the data page by page, and waits when the app says too many requests have been sent. It turns the app's records into rows and columns. It remembers when it last ran, and next time asks only for records changed since then. This is called an incremental sync. And when the app adds a field, it adds a matching column to the table in the warehouse.
Say a skincare brand in Delhi sells through its own online store, and wants its orders in the warehouse. The first time, the connector copies the store's whole history of 1,20,000 orders. The store's API returns 250 orders per page, so the connector sends 480 requests. The API allows two requests a second, so the first copy takes about four minutes. After that, the connector runs every 15 minutes. At 11:00 a.m., it asks only for orders changed since its last run at 10:45 a.m. It receives 85 new orders and 12 changed ones, which are refunds. A week later, the store adds a gift-wrap field to its orders, and the connector adds a gift-wrap column on its own.
Connectors are used beyond ELT tools. Business intelligence (BI) dashboards use them to read an app's data directly. Analytics automation tools use them as the first step of an analyst's workflow. And software companies use them through embedded iPaaS, to connect their product to their customers' apps.
Some sellers go one step further. Their connectors make every source look like a database, so that a program can read an app with ordinary SQL, as if the app were a table in the data warehouse.