There’s a bit variety of tools available that might be being used at your company to prepare the data. It might be conventional code e.g. C# or Java, in which case it might not have an obvious pipeline structure, but hopefully normal testing stuff is happening on it.
If it’s more pipeline-y, then it could be stored procedures in a database, a specialist on premises bit of software like Microsoft SQL Server Integration Services (SSIS) or Talend, or something more exotic like Apache Spark. Or it could be something more cloud-based, like Azure Data Factory.
I’m saying all this not because I want to confuse you, but to help you identify what tools might be in use at your company, and so the appropriate ways to test them.
At my current place we use a combination of SSIS and stored procedures. The bulk of the business logic is in stored procedures, and SSIS is the control logic that strings them together. This means that the business logic can be unit tested using tSQLt, leaving less to do at system test time (where it’s harder).
Also, it might be referred to as ELT rather than ETL (the T and L swapped around). ETL means Extract, Transform, Load (i.e. read, process, write). ELT is Extract, Load, Transform. (Again, just to help you work out what your organisation is using.)
ETL is where data is read from somewhere, processed, and only then written to where it would be used in e.g. a dashboard. The data is in a good form before the dashboard reads it, which means that there’s less work for the dashboard to do. However, it means that the dashboard can’t change until the ETL has changed to prepare data in a new way. In ELT the data is made available e.g. to dashboards in an earlier state, so the dashboard has more work to do, but it’s more able to change without worrying about what upstream processing / teams are doing.