Great question! When you enable a source, we begin running a job on an interval for you that pulls the data from the source, applies some light normalizations and transformations, and sends the data to our Object API (which is in charge of upserting the data into Segment and flushing it to your warehouse).

In Salesforce's case, we issue bulk queries to pull the complete collection on the first run, then modify the queries thereafter to request only data that's changed since the last run.

We don't do any aggregation of the data. We load it into a data warehouse (redshift or postgres) in its complete, raw form so that you can use SQL to aggregate/join to your heart's content. Here's an example: https://help.segment.com/hc/en-us/articles/208215583-Salesfo...