Skip to content
BENIBIX Learn

Extract, Load, Transform (ELT)

Data🧭 FoundationsData Pipelines

Also known as ELT

Load the raw data first. Then clean it in the warehouse.

Think of it like

It is like putting all the food in the fridge first. You wash each thing when you need it.

Example

A startup loads raw app events into BigQuery each hour. Then analysts write SQL. It turns them into a count of users per day.

Why it matters

The raw data is kept. So when you fix a bug, you do not need to go back to the source.

Definition

ELT uses the same three steps as ETL, in a new order. Extract reads the data from its source. Load writes it, raw, straight into the warehouse. Transform then reshapes it inside the warehouse, using SQL. This became common once cloud warehouses got cheap and fast. The raw copy stays, so you can transform it again later.

Example code

sql
CREATE TABLE clean_orders AS
SELECT id, LOWER(email) AS email, total
FROM raw_orders
WHERE total > 0;
The transform step. It runs as SQL in the warehouse, on rows that came in raw.
Same three steps, different order. The order decides where the cleaning happens.

Common mistake

Loading raw data and never cleaning it. ELT moves the transform step, it does not remove it.

Key takeaways

  1. 1
    Load first. Raw data lands just as it is.
  2. 2
    Transform in place. SQL runs inside the warehouse.
  3. 3
    Raw copy kept. You can rebuild any table later.

Compare

Learn it properly