ETL vs ELT

Loading data into a data warehouse is a task on its own entirely. There are two important paradigms to consider, which are ETL and ELT. The letters stand for the same word between the two, which are 'E' for extract, 'T' for transform and 'L' for load. The order of which they happen is what is important.

The comparison below outlines the main differences between ETL and ELT.

ETL	ELT
- Extract - Transform - Load	- Extract - Load - Transform
The “old school” method.	The modern method of loading in data.
Transformation happens outside of the data warehouse.	The data warehouse can do all the transformation work.
Raw data is not transferred into the data warehouse.	Raw data is transferred into the data warehouse.
Running different cloud services become expensive.	No additional job optimisation required.
Data loading into a data warehouse is slow due to prior transformation.	Requires more compute power than ETL due to the volume of data moved.
Good for compute intensive transformation.	Can better use the benefits of a cloud data warehouse.
Legacy/local systems with less compute power would handle ETL more easily than ELT.	Parallel processing.
	Cheap storage costs.

ETL is viewed as the older method, since it allows for a limited compute power due to the lack of cloud computing. With the advent of cloud computing, ELT has become more standard in the industry, since cloud power scales and working with large data sets becomes a question of price rather than availability.