Keeping the shelves fully stocked at high street retailer Debenhams is no mean feat, with 16 million customers purchasing thousands of products, ranging from clothing to beauty products, jewellery to flowers, and electrical appliances to furniture.
Analysts at the retailer use business intelligence tools to track purchasing patterns and optimise stock levels. For that analysis to be at its most powerful, populating the data warehouse with timely sales information is critical, says Jagpal Jheeta, business development manager at Debenhams. “This key data [from the data warehouse] is then used to assess the previous week’s performance and make crucial business decisions relating to sales and merchandising.”
Until recently, the weekly process of moving its trading information from its 124 Debenhams stores across the UK and Ireland, along with online transactions, into its data warehouse system, took an entire weekend. Frequently the process was not complete when business analysts arrived at work on Monday morning.
One reason for the delay was the bottleneck created in the extract, transform, load (ETL) procedure, where data was extracted from the source system into a third-party server using a row-by-row operation, before being loaded into the target data warehouse system.
In order to meet the 7am Monday morning deadline set down in its service level agreement, Steve Kircher, IT director at Debenhams, began a data integration initiative in August 2005 to bypass the transformation process between the source and target systems, and instead use the power of both Debenhams’ transaction databases and its data warehouse on a Netezza Performance Server to do the integration process instead.
To facilitate that, in February 2006, Kircher and his team implemented a new form of ETL tool from data integration vendor Sunopsis, which reduces the complexity by transforming the data after it has been extracted and loaded.
This technique reduced the time taken to populate the data warehouse to 12 hours. In future, Kircher plans to move towards a daily upload of transactions, and eventually into real-time processing.
Already the project is paying dividends, increasing the availability of popular product lines and making its supply chain more efficient.
