Allied Irish Bank’s (AIB) enterprise data warehouse project manager Michael McMorrow is a great proponent of Teradata’s functionally neutral and self-managing approach to data storage. The bank rolled out the vendor’s Teradata Warehouse solution a few years ago and is now focused on keeping up with the data management changes required as a result of new source system inputs, such as accounting system changes, he explains.
“If you are rolling out a data warehouse solution it needs to be a genuinely neutral model; don’t model the solution too closely to the idiosyncrasies of your bank. Otherwise you will constantly be reacting to new requirements. If the data is consistently managed and robust, a new report for regulatory purposes shouldn’t be daunting,” says McMorrow.
Before its rollout of the Teradata warehouse, AIB collected customer data via an internally developed customer information file solution, which provided a single view of the customer. However, this solution was not robust enough to suit the needs of the end users in terms of data analytics, as it was a lengthy process to develop new functionality, and thus it opted for the Teradata offering. The bank’s existing customer data was then migrated onto the new solution and augmented with customer history and transaction history data.
McMorrow explains that there is a semantic layer between the warehouse and the user outputs, including analytics, management information system (MIS) and reporting systems, which means end users are unable to affect the centrally stored data sets. The source systems that feed the data into the warehouse are also responsible for data cleansing, so that the warehouse itself is focused on keeping the golden copy that is produced clean. “The interface model means that AIB can change the source systems but we are able to shield the rest of the system from these changes,” he explains.
To this end, the bank is currently reengineering its core accounting systems and it is up to the data warehouse to rework its internal data storage systems to take account of these changes. “The changes mean that fundamental bits of data are changing in structure and it is a real challenge to both map the new data to the old and alter the systems where required,” says McMorrow.
At the start of the data management process, McMorrow indicates that the challenge of senior management sponsorship was significant: “We had five changes in CEO during the implementation process.” Although the rollout has now been completed, the data warehousing team must also remain wary of any budget cutting activity that may negatively impact the maintenance of the system. “We are also aware of the problems around a legacy of usage when staff move on. We need to make sure that strong governance and strategy is maintained by retaining specific data stewards for each data unit. These stewards sign assurance forms to ensure a strong process for the ownership of that data,” he elaborates.
Each phase of development also needs to provide tangible rewards in terms of either cost savings or benefits, adds McMorrow. “However, something that may be harder to prove is the inherited benefit of previous implementations,” he says. “For example, our customer data warehousing project meant we were later able to kick off our Basel project and this is hard to quantify directly.”
McMorrow’s philosophy for a data warehouse is therefore for it to be treated akin to a utility that is charged back to the end user for service provision. Although the warehouse is responsible for data verification, it is not responsible for data cleansing and is therefore similar to a system of record. “End users have to sponsor any changes that need to be made and if they wish to receive new data sets,” he concludes.