ELT shares the same abbreviated words with ETL, but the swap of order makes all the difference. In the case of ELT, data is loaded before transformation takes place. Another way to view it is that data transformation is performed on-demand. ELT is the new generation of data utilization process which focuses on data availability, and can be more flexible and faster depending on the business case.
How does ELT work?
Traditional ETL method works well when the data sources are relational in nature and the final output is clearly defined and not set to change. In a visualization dashboard example, if the users know exactly what type of charts they want to see, they can utilize the ETL process to produce an optimized database and dashboard with set format. However, if different users like to see different visualizations based on the same set of data, the ETL process will need to repeat as each different chart will require separate development. Users will need to wait until data transformation is done and loaded to access the content. Under the more trendy ELT process which usually involves larger amounts and cloud infrastructure, data is loaded into a data warehouse or data lake right after extraction. So users can access the data immediately. This can be more efficient if the data is unstructured in nature and large in size. Because transformation of large data can take a lot of time and resources and perhaps only small proportions of the data is needed in the final consumption.
It is worth noting that ELT is not necessarily as reliable as ETL since the loaded data will be raw in form and will likely contain much more noises than the transformed data.
Get started for free right now!
Let’s build data apps to transform your business processes