Maximizing the Market Basket

OfficeMax is using a database management system to analyze shopping patterns.

If last year’s holiday shopping season proved anything, it is that promotions alone will not entice customers to shop. Worse, the wrong promotions tend to attract those dreaded shoppers who reserve their dollars exclusively for sale items. Since this practice takes a toll on retailer margins, chains need to be cautious when creating promotions.

Industry experts say that campaigns should revolve around items shoppers want, as well as those that will spur consumers to purchase additional items that will increase shopping basket sizes. That is why it is paramount for retailers to supplement all promotions with the ideal assortments and optimal inventory levels storewide.

Chains can achieve this goal by digging deep inside data warehouses and analyzing consumer-shopping patterns. These patterns run the gamut, from understanding product affinities and the impact of customer loyalty during each visit, to merchandise returns and potential fraud instances. The placement and availability of merchandise in store aisles also impacts each of these variables.

For OfficeMax, compiling this information was not a problem.

“Our challenge was gathering the information in a timely manner and quickly interpreting the effect of a promoted item on related merchandise,” explained Charlie Baugh, senior VP information technology, OfficeMax, Naperville, Ill., which operates more than 1,000 stores.

Historically, the chain’s high-volume data processing workload required extensive and frequent tuning, and specific complex queries often took hours to complete. “For example, it could take an average of almost three hours to extract one week of retail sales data from our database,” Baugh noted.

These delays caused other issues. Without optimal insight, OfficeMax often didn’t have enough information to accurately understand correct inventory levels.

“It was difficult for us to be prepared, from an inventory perspective, to meet consumer demand,” Baugh said. “Worse, a lack of insight would force us to make adjustments to safety stock at the wrong levels.”

”In 2008, the retailer began its search for an analytics tool that could access information faster, help the chain create optimal promotions and make decisions more efficiently.

From a business perspective, the chain needed a solution that could uphold a high performance level and seamlessly integrate into OfficeMax’s existing analytics environment, which consisted of Sybase database management and business intelligence software, and Oracle databases.

The chain evaluated 12 database products before embarking on an internal proof-of-concept test based on two vendors’ massively parallel (MPP) columnar database solutions. The retailer chose the ParAccel Analytical Database, an MPP columnar database management system from San Diego-based ParAccel. The system is helping OfficeMax understand purchase affinities, predict consumer-shopping patterns and maintain proper inventory levels—three factors that will encourage bigger market-basket sizes and faster-turning merchandise.

The retailer’s existing data warehouse is connected to the ParAccel appliance, and OfficeMax filters data required for market-basket analysis into ParAccel. The chain is now able to complete comprehensive market-basket analysis based on detailed sales history from multiple storage locations and files.

OfficeMax added the ParAccel platform earlier this year but is still testing the technology.

Early results are promising. OfficeMax now loads data, on average, 20 times faster. More specifically, in the past it took OfficeMax almost three hours to load one week’s worth of data; now it takes less than one minute.

“The level of performance on the query side is greatly enhancing our market-basket analytical applications as well,” Baugh reported, although he would not share specific details.

Baugh hopes to use the platform to replace existing data marts and “streamline existing ETL (extract, transform and load) processes that pull data from outside sources.”

Login or Register to post a comment.