There is a ton that can be done in SQL to transform data after data is bulk exported into staging tables. I really do like the ELT transition from ETL. Which is kind of nice as the DB is usually really beefy in terms of compute and if distributable like say BigQuery it just scales seamlessly (provided your credit card has a high enough limit) and you don't have to worry about all the distributed systems stuff you might have to deal with if you were running a Spark cluster on your own.