DATABERG
IT managers are faced with challenges because the ETL process (Extract, Transform, Load) takes up 70% of the time in a data-centric project. The ETL process is part of the data integration process and is of high importance in data warehousing.
Innovations to speed up this process, as well as databases moving to the cloud, have led to a new order, ELT. To decide if your business should make the switch from ETL to ELT, it is essential to understand the differences between the two processes, and which advantages ELT has in today's world.
Functions of ETL
The extraction phase consists of reading data from one or more databases or data sources. For example, the extracted data can come from ERP (Enterprise Resource Planning) applications, sensors, industrial networks or manually maintained spreadsheets.
Transformation is converting the extracted data from its previous form into the form in which it needs to be in order to be placed into a data warehouse or another database. Transformation is made possible by using rules or lookup tables, or by joining data from multiple sources.
The loading phase injects the transformed data into the data warehouse. Data warehouses contain many rules on how the data will be used, like summarization rules, standardization of encoded attributes and calculation rules.
ETL tools also transport data between source and target, keep track of how data elements change during this transition, exchange metadata with other applications as needed, and administer all runtime processes and operations.
The purpose of ETL is to load the warehouse with integrated and cleansed data. Redundant data is eliminated, and less time-critical data is consolidated. What is left is the essential business asset for a company, which can be used to aid in decision making.
How is ELT different?
The ELT process includes the same steps but in different order; Extract, Load, Transform. Therefore, after extracting the data from its sources, it gets loaded in its untouched raw state into a data warehouse or data lake. The transformation is done in the data warehouse as needed.
Due to the massive amounts of data that results in being stored, this process can in most cases be done only in the cloud, not on-premises. It is predicted that in three years, at least 75% of all businesses will load their data into a cloud platform. Therefore the importance of ELT will grow. The ELT process follows a push-down optimization mode, where the target database will contain the transformation logic.
As the world is shifting from on-premise data warehouses to data warehouses and data lakes in the cloud, it seems like ELT is a better option than ETL regarding data integration. Let's look at ELT's advantages.
Advantages of ELT
Speed
Data loading happens faster since you now don't have to wait for the transformation step to complete, which usually takes up most of the processing time. However, since the transformation step has not taken place yet, the eventual analyzing of the data will take more time, as you first need to transform before doing analytics.
Quantity
Since you can only put transformed and structured data into a warehouse, you might lose data and, therefore, valuable insights. Thanks to the ELT process, you can load data in its raw and unstructured state before having to transform it. This allows you to store all your big data, which will offer an opportunity to gain more insights, even after a long period of data storage prior to analysis, as new ways to dig into the existing data may be found later on. It will also make sure that no data you extract gets lost in the process.
Accessibility
The good thing about a cloud data lake is that they can contain both the raw unstructured data, as well as the warehoused, transformed data. Because of this, different areas of the organization can access the data lake for their own, particular purposes: data scientists can make use of the raw data, while business users can dive into the data warehouse with structured data.
Scalability
Since ELT is a process that is only available when loading data into cloud storage, it comes with the traditional benefits of flexibility and scalability of the cloud model. You can adjust the amount of storage space you need, matching the business requirements. This eliminates overhead costs that would previously be created by keeping data storage space readily available onsite to match the worst-case scenario that may be used only for a part of the year.
Cost reduction
As is widely known, investing in an onsite warehouse is expensive, especially since the amount of data is growing every day, and you'd need to invest in expanding that warehouse. Today, with cloud storage, better solutions are offered. Especially if you are a small-to-medium enterprise, opting for the ELT strategy might be better suited.
Companies like AWS, Google or Microsoft offer affordable in-cloud options by using SAAS solutions for data storage, seamlessly adjusting to companies' needs.
Disadvantages of ELT
Governance
What can be tricky about ELT is that it requires you to upload all of your data before removing sensitive information. Doing this wrong could violate GDPR, HIPAA, and CCPA standards. Confidential information will be more vulnerable to hacks and inadvertent exposure. The additional risk is having cloud-servers in other countries, with different data governance rules.
Conclusion
Ultimately, your company has to choose in the battle that is ETL vs. ELT. With its many advantages regarding access, costs, flexibility, and more, it is safe to say that ELT is increasingly becoming the favoured approach. 
