ETL Testing: How to Extract, Transform, and Load Data

Pohan Lin
Published 03/16/2023
Share this on:

How to extract and transform data with ETL testingThis article will cover the main aspects of ETL testing you might need to know.

 

What is ETL testing?


ETL stands for Extract-Transform-Load. ETL testing is conducted after a business transformation to ensure that data loaded from the source to the destination is accurate. For example, tools like the Apache Hive components software are available for this purpose.

Data is extracted from an online transaction processing database and often non-online transaction processing systems such as spreadsheets or text files. This is transformed to match the existing data type in the data warehouse and then loaded into the data warehouse database.

 

An example of ETL


diagram of ETL Testing
Source

If you have a charitable organization with different departments such as fundraising, marketing, and accounting, each department will be handling the information of volunteers and donors in their own ways and also store that data differently. This is even more likely with hybrid working scenarios.

If they had to manually cross-reference every individual on their systems, it would take a long time. A data warehouse for storing data from different sources in a single, unified structure with ETL massively speeds up this process.

 

What does ETL testing look for?


ETL testing verifies whether, for example, your inventory forecasting data has been moved as expected. It checks whether the counts in both the source and the target match, and is used to see if the transformed data matches expectations.

It also checks that the foreign primary key relations have been preserved during the process, and for any duplications in the loaded data.

 

ETL testing method


The first part of ETL testing consists of preparation – designing the most appropriate test strategy based on the ETL process logic and transformation rules. The test strategy consists of a document with a list of information about the test length, the reasons, and methods for testing, and the tools and people needed. This strategy is then organized into test cases – numbered instructions for running the test.

Testers need to look at the ETL process documentation, such as technical specifications, mapping specs, and business requirements, to decide which data should be integrated, including their sources, formats, and destinations.

Testers also need to ascertain the transformation rules for transforming and cleaning the data, and where and how the data will be loaded. Then test data is generated. Real-time synthetic data is needed against which testers can run tests.

After this, test cases and scripts are created. These test cases are used to describe the things to be checked and the method for testing. ETL test cases are normally written in SQL. They also tend to have two queries – one which extracts data from its source, and another which extracts it from the target storage.

 


 

Want More Tech News? Subscribe to ComputingEdge Newsletter Today!

 


 

Three types of ETL tests


There are essentially three types of ETL tests, which we will describe here.

 

ETL performance tests

During the data integration testing, the entire application is tested in an environment that imitates production. Various functions, such as interfaces, databases, etc., are all tested against the data. This process allows you to find any obstructions and ensure that the process is ready to scale up with the increasing amount of data.

This method tests whether data loads and queries are carried out within the expected timeframe, and load times for various data volumes.

 

Data quality tests

This test confirms that data has been extracted properly, without any losses or duplicates.

Duplicate data is checked, to make sure there are no rows with the same unique key column or column combinations. Data is validated in accordance with validation rules to check for any incorrect dates, input errors, or inaccurate symbols that could render the data invalid.

Metadata testing checks that definitions conform to the data model and application design, and reference testing checks data against required attributes.

 

Data transformation success tests

Data transformation testing is used to check that data is transformed as required. The two approaches to this are white box testing and black box testing.

White box transformation testing entails testing application code. A tester reviews how the source-to-target transformation happens in the documentation, and then uses this logic in code to compare the transformed data to the documentation.

Black box transformation testing allows you to check how the system functions. The tester prepares data to mirror various transformation scenarios listed in the documentation. You don’t need coding for this method.

 

To sum up…


Whether you’re in the business of manufacturing ERP software, running a small coffee business, or anything in between, ETL testing can help you to align your data in one single, unified data warehouse.

 

About the Writer


Pohan LinPohan Lin is the Senior Web Marketing and Localizations Manager at Databricks, a global Data and AI provider connecting the features of data warehouses and data lakes to create lakehouse architecture. With over 18 years of experience in web marketing, online SaaS business, and e-commerce growth. Pohan is passionate about innovation and is dedicated to communicating the significant impact data has in marketing. Pohan Lin also published articles for domains such as Landbot.

 

Disclaimer: The author is completely responsible for the content of this article. The opinions expressed are their own and do not represent IEEE’s position nor that of the Computer Society nor its Leadership.