A major, internationally active bank needed to disclose its risk-weighted assets (RWA) to the respective supervisory authorities pursuant to the regulatory provisions of Basel II. The calculation of results was conducted with Fermat, a leading product for detecting key data of Basel II.

With the objective of conducting the calculation with the support of Fermat, all individual transactions had to be loaded into different data tables depending on the financial instrument. Since the bank had chosen a decentralized architecture (each larger organizational unit has its own Fermat installation), various ETL tools were employed to transform and upload the data inventory subject to requirements of Fermat. Next to the approx. 30 purely transactional business tables for loans, netting, counter-parties, etc., approx. 60 parameter tables needed to be filled to calculate all business transactions pursuant to the provisions of the federal supervisory authority as well as those of the parent group (so-called home/host provision).

High project effort

The total project effort within the banking group was very high. Next to the provision, transformation and uploading of data (ETL) conforming to Fermat, the technical and business side tests were particularly costly.

Before RapidRep came into operation, testing was almost completely done with a solution based on Excel. The team consisted of approx. 20 testers, which utilized different Excel workbooks depending on each test cluster. By these means test data and expected results were stored into Excel. They were then manually harmonized with the calculated results.

Challenges

The following factors document the high testing requirements for Basel II results data:

• The data requirements were high and weren´t easy to fulfill.
• Problems with data quality often emerged. The team was generally only slightly acquainted with Basel II and the Fermat software. The question whether a test case result was correct or not could not be sufficiently answered in several cases.
• The defect management was only inadequately implemented. Errors were often captured incompletely with Mercury TestDirector, so that it was hard for business analysts to reproduce and analyze the defects.
• For years, the provisions of Basel II were in flux, just as the bank-internal guidelines regarding the Advanced Approach. The testing was subject to great structural changes. The implemented calculation engine of Fermat went through many versions and patches.
• There was a high amount of test cases (several thousand test cases).
• A strict time schedule induced a preference of quick adhoc solutions over structural improvements.

Consequences

• Long testing cycles and cigh costs.
• All testing cycles (system-/components-/acceptance- tests) took very long and no improvements could be observed in the process, neither on a qualitative level nor regarding the schedule.
• Due to the large amount and complexity of test cases (asset-backed securities, cross border limits etc.) the very mostly manually conducted testing cycles could no longer be managed.
• For cost reasons the bank simultaneously decided to outsource the testing to India. This was a convenient moment to consider new testing strategies.   Solutions The key for solving those problems was to adjust the testing process with the support of RapidRep.

Solution

The following steps were implemented:  

• SQL experts created a report definition with the help of RapidRep Designer that combinded the technical and business  know-how of the calculation engine’s respective version with the laws of Basel II. This succeeded with a single report definition that could cover all test clusters for retail and non-retail (incl. credit-risk migration and ABS).
• Thereupon, the testers concentrated on defining the test cases and loading the test data. RapidRep automated the elaborate manual interpretation of the results.
• The testers (in the transition phase these were resources in Europe and India, later exclusively in India) used this report definition in RapidRep Runner for the validation of results. In addition, a color code in the Excel sheets allowed for the quick recognition of regression errors.
• Failed test cases were uploaded together with the RapidRep Excel report into the Mercury TestDirector. This Excel Workbook included all information that a business analyst would need to reproduce errors and detect error causes.  

The success

All actions have proven as highly effective. The duration of an entire testing cycle - Loading the data -> Calculation -> Validation of the results - has been cut in half. The time needed validate the results has been reduced to a tenth!  

This time saving measure does not yet take into account that the downstream defect management also benefited from the implementation of RapidRep. Due to the detailed information collected in the Excel Workbooks analysts were able to identify error causes more quickly and thereby to reduce the problem solving cycle considerably.  

The separation of the RapidRep Designer and the RapidRep Runner has ideally supported the outsourcing of testing activities to India. The business /technical expertise in form of the report definition remained in Europe during the first months, while the bulk tests were conducted in India with the RapidRep Runner.