The ASB Bank will save thousands of dollars by swapping its Oracle data warehouse environment trunning on Unix to Microsoft SQL Server 7.0 running on Windows NT, says Garry Fissenden, ASB's general manager technology and operations. ASB is one of the first SQL Server sites in the country.
The eight-year-old data warehouse forms the basis of the bank's entire marketing segmentation strategy. Every transaction and bit of customer data is captured there. There are nightly updates for transactions and monthly feeds from other systems.
Fissenden says the cost of replacing the Hewlett-Packard Unix machine that the data warehouse runs on with a present day equivalent would have been $400,000. The bank has instead bought a Dell PowerEdge server with four 450MHz Xeon processors running Windows NT for $150,000.
He says maintenance costs of the Windows NT operating system will be cheaper - $20,000 annually as opposed to $60,000 for HP Unix, and database maintenance for SQL Server is included in the bank's overall Microsoft enterprise maintenance agreement, compared to $40,000 a year for the current Oracle version 7.3 database.
Oracle national sales manager Robert Gosling says the performance difference between Oracle 7.3 and Oracle8i in a data warehouse environment is "very significant".
"We could also have run on the Dell/Windows NT platform so it really comes down to a difference of $40,000 for database maintenance. I would have thought that given the benefits of a data warehouse, $40,000 is not a lot to pay for peace of mind that it will be up and running 24 hours a day."
However, Fissenden says another reason for the switch is that the bank has more SQL Server skills in house than Oracle skills. "Every branch has a smaller SQL Server database so we already know it well and all our skills are in standard Microsoft tools," says Fissenden.
The data warehouse is expected to be running on the new platform by the end of the year. Testing so far has dispelled concern about whether SQL Server can deliver the performance and scalability of existing heavyweight databases. "We have stressed it and thrashed the living daylights out of it and it just kept asking for more," says Fissenden. "We're finding that performance is phenomenal [compared to Oracle 7.3] but to be fair we haven't done any work with Oracle8 so it might be an unfair comparison."
In a build-up to the conversion from Oracle to SQL, the bank has used data warehouse data and SQL Server's online analytical processing tools to build two 3D analysis cubes, containing 50 million transactions collected over the last 14 months. Using the cubes the bank can predict branch usage (based on how many transactions go through each channel) to then forecast staff requirements.
Another model calculates the profitability of every customer and is based on a matrix of customer balances, what services they buy, what fees they are charged and what transactions they make. "We use that to work out what each transactions costs per channel and then debit the customer's profitability with that transaction cost," says Fissenden.
"We also break customers into categories because the absolute dollar value doesn't mean a lot. They range from very profitable to unprofitable and we track customer movements in and out of those categories."
This information is used for targeted marketing campaigns and determining high-value customers. "It has also allowed us to cost out profitability of our customers and each product. It's used for everything from monthly reporting to parents through to reporting on transaction statistics."
The cubes are used by personal bankers and branch managers who access the system through the company's branch wide intranet.
Eventually the data warehouse will also integrate with a customer relationship management (CRM) system the bank has chosen by Seattle-based Onyx. The CRM will be rolled out across the bank over the next year.