ETL or ELT: choose when to transform new data

The pipeline needs to run and I don't know if the cleanup happens before or after it lands in the warehouse.

Beginner to intermediateRead 4 minDoing: an hour or two, mostly reading real sample records and writing the note5 steps

▶ Start

What you'll have at the end

A written decision, ETL or ELT, with the one compliance or cost reason that settled it

Before you start

You need

A new data source you're about to connect (an export, an API feed, or a database extract), plus a handful of real sample records pulled from it.

Not covered

Writing the transform code itself once you know which side of the load it runs on, and building out a full PII-detection pipeline beyond a first redaction pass.

Leans on

5 steps

Where to go next

Choose full reload or incremental for a new source

for deciding how much of the source to pull each run, once you know which pattern loads it

Customer emails and other PII sit unmasked in a shared table

if personal data already landed raw before anyone caught it, for masking it after the fact instead of before

Checked 25 Aug 2026

Part of the Data Warehousing cookbook