Mass Data, Migrations and Scenario Analyses
IT landscapes in the financial services and public administration sectors are marked by large amounts of data having multiple relations. Because the data has evolved over time, data quality can be problematic. Analyzing and testing mass data during migration, for example, requires extra effort (and cost). The same holds true for scenario analyses: e.g., proposed changes to tariffs (premiums) or other policy changes frequently present particular challenges when it comes to testing the scenarios, since business-critical cases must be pinpointed and processed on an individual basis, and under severe time constraints.
Conventional testing methods focus on program quality, and not on the quality of the data itself. As a result, their effectiveness is limited in cases where relatively simple programs, like those used for data migrations, must be applied to very complex data pools.
The OBJENTIS Approach: A single picture indeed paints more than 1,000 words!
Visual data exploration offers a highly effective alternative approach: The use of statistical techniques permits data structures and patterns to be graphically displayed, thus enabling existing deviations from reference values to be identified within a fraction of the time that it takes using conventional means. Visual data exploration goes beyond mere random sampling or testing against reference values, however, as it actually permits data pools to be analyzed fully: The tester is able to identify all deviations down to the individual dataset.
Tools capable of graphically representing large data pools via explorative-descriptive statistical methods are available to support the task. The key to success is a well trained eye for reviewing the data and selecting the proper means of displaying it graphically.
Visualization of data permits errors to be identified and localized with less fuss, thus significantly saving time. The visualization process needs to be defined and set up only once, thereafter it runs automatically. As a result, data quality can be determined at “one quick glance” each time changes are made – either to the data itself or to the software module that handles it.