MarsTQM: Our AI Infused Tool Addresses the Biggest Challenges in ETL Testing

pointing-at-data

ETL Testing has come a long way from the early days of testers using sample data sets and writing code as second set of eyes and hoping to catch all defects; to understanding the need to identify all boundary test scenarios, setting up synthetic test data to simulate these scenarios for defect free testing. There are a lot of tools available, including homegrown and out-of-the-box SQL-based, that improve data quality and reduce testing time but don’t address the biggest challenges.
These tools either don’t support or are inefficient in 20% of situations that have transformations involved especially complex transformations. Even though these situations are less frequent, they are the biggest causes of unidentified defects.
So there is heavy reliance on skilled testers with or without tools to achieve good test quality, but the quality still varies from tester to tester and time to time because of human factors and experienced testers are expensive adding significantly to the testing costs

Missed test scenarios lead to gaps in test coverage:

  • Test scenario identification requires experience, and which is especially true for the more complex transformations.
  • It’s a proven fact that automation removes variability that is inherent in manual process.

Data setup is not automated:

  • A significant amount of testing time is taken up by this process.
  • It’s very hard to keep the test data from getting corrupted when there is an overlap of source data usage across multiple project teams and testers.
  • Test data errors produce false positives, adding significant time to debugging processes.
analyzing-data
entering-data

SQL Coding is still required for execution and validation:

  • Even some simple transformations require significant SQL coding skills to generate the execution and validation query.
  • It’s always a challenge to hire and retain highly skilled testing resources.
  • Errors in SQL code will lead to false positives and increased debugging time.

Need of multiple artifacts and channels for Test management:

  • Validation queries need to be executed in a SQL editor and the results have to be saved in a repository
  • Status reports have to be compiled, saved and sent to stakeholders through one of the communication channels