FNB Corporate, the corporate banking division of First Rand Bank, has bought Sybase's Adaptive Server IQ Multiplex 12.0 analytical database to power its data warehouse, enabling the financial institution to identify its most profitable customers.
FNB Corporate has over 25 banking products, each with its own system, and more than 6 000 clients and 100 000 client profiles. The data warehouse solution was implemented to provide the bank with a 360-degree view of its clients so that it can ascertain cross-selling opportunities, and detect potential customer defection, says Malcolm Stevens, responsible for data warehousing at FNB Corporate.
"We have a variety of disparate systems from various vendors," says Stevens. "Before the implementation of the data warehouse certain queries were impossible to answer. We can now determine total revenue per client, total profit per client, our 30 top and bottom clients, and perform various other profitability and customer relationship management analyses. We can also pre-empt client defection. The retention of valuable clients more than justifies the expense of a data warehouse."
FNB Corporate first implemented a client-centric data warehouse in 1999, adopting Sybase's Adaptive Server Enterprise 11.5. When the system was no longer able to cope with the analyses required to determine which customers were profitable, and which had the potential to be profitable, the bank approached Sybase.
"Our challenge was to find a data warehouse solution that would meet our current business requirements, and that would cope with future business growth," says Stevens. "Sybase showed us there was a way to maintain our existing database architecture while bolstering its performance with Adaptive Server IQ Multiplex 12.0."
The data warehouse stores historical client data; links client profiles from various systems; provides a unified view of the client to Essbase, Hyperion's analytic solution; runs pre-defined Web-based reports for around 150 users; runs ad hoc queries by about 20 'power' users; and will provide data mining functionality to a team of statisticians for in-depth analysis.
The bank's warehouse now stores 40GB of compressed data, 80 to 100GB of raw data. The largest table on the warehouse has 261 million rows of transactional data, in which every column is indexed, often more than once. Where traditional databases store rows of data with indexes on top - which means fast row retrieval but slow vertical selection, IQ Multiplex provides rapid vertical selection. Stevens says IQ is also easily horizontally scaleable, requiring only the addition of reader nodes.
"Our response times are now sub-second for simple queries. More complex queries take slightly longer but are far faster than a traditional relational database. This has had a major impact on our users; with faster response times more people are accessing the data warehouse, and we already know that the company's future growth is taken care of."
Stevens adds that the bank now has a unique warehouse architecture consisting of Sybase's ASE and IQ Multiplex, and Ascential's data integration suite, DataStage, on a Sun E3501 server. For large data retrievals Essbase is loaded from ASE.
"The data warehouse provides us with the ability to produce critical business information when we need it," says Stevens. "Data loads are vastly simpler and faster, tables can be loaded while users are online, indexes do not need to be dropped and recreated, and load windows can be met more easily."
FNB Corporate has also found IQ Multiplex to be robust and stable, requiring minimal maintenance. Database consistency checks are run periodically, but no index maintenance is required. "The scripts are all automated, so less than half an hour a day is spent on maintenance," says Stevens. "This frees up the database administrator's valuable time for warehouse design and construction, and for helping new users."
Share
Editorial contacts