BI Testing Complexity
Businesses rely on visual BI dashboards for insights into key data trends, identifying opportunities and potential improvements. Though simple in appearance, these dashboards and reports are built on complex data aggregation and analysis. This must be tested rigorously to avoid basing critical decisions on inaccurate Business Intelligence.
Dashboards and reports are built on data mined from live systems, often stored in fact tables that refer to dimensions tables. These tables are analysed for interrelations between dimensions across the columns and tables. The mined data is often “folded” to reveal low-level transactions, and is frequently denormalise up at stages to aggregate, summarise, and spot temporal trends in it.
Testing these processes is complex, and typically slow. It requires every data combination reflected in the dashboards: every time a given dimension or set of dimensions join, and every time they intersect but do not join. This data is often created in SQL, but this can be complex, especially where there is no existing data in live systems, or if data is protected by data privacy legislation. Test data often lacks the negative scenarios and unlikely combinations needed for rigorous testing, while the data must be slowly and manually inputted into systems. Expected results are likewise hard-to-define, and must be manually compared to the vast actual results produced by testing. Manually created test cases and data are furthermore brittle, and must be recreated manually each time business requirements change.
Rigorously Test Dimension and Fact Tables
Test Data Automation and Test Modeller combine to provide a complete BI testing solution for rigorously validating fast-changing data transforms. Accurate models map the complex interrelations between data, before using a powerful test data management engine to systematically find or create data for every possible combination. Expected results are defined systematically at the same time as test data, while a high-speed workflow automatically inputs the test data and compares actual to expected results.
BI testing teams can use powerful coverage techniques to test business rules exhaustively, or can easily filter data noise to test exact scenarios. As requirements change, they only have to update the central models, automatically re-creating and re-running the rigorous test set.