ETL / ELT

ETL and ELT are the two patterns for moving data from your sources into a warehouse. ETL (extract, transform, load) cleans and reshapes the data before it lands. ELT (extract, load, transform) dumps the raw data in first and transforms it inside the warehouse afterwards, which is the modern default now that warehouses are cheap and fast.

For a founder the distinction matters less than the principle: you want one reliable pipeline that pulls every source, your app database, Stripe, your ad platforms, your CRM, into one place on a schedule, so analysis isn't a manual export-and-paste each week. ELT usually wins because keeping the raw data lets you re-model later without re-extracting. Off-the-shelf connectors (Fivetran, Airbyte) handle the extract-and-load so you only own the transforms. Get this plumbing right early and every metric downstream gets more trustworthy; get it wrong and you'll spend evenings reconciling spreadsheets.