Lund Seeks to Improve Forecasting With New Technology

JAMES TENSER

LAS VEGAS — What does a data warehouse project have to do with boosting profits on Halloween?

Executives at Lund Food Holdings, Edina, Minn., are confident that data warehouse and analytics capabilities installed this year will power smarter forecasts and buying decisions across its 21 supermarkets in Minnesota, which operate as Lunds and Byerly's banners.

“We get one shot a year at Halloween candy,” said Curtis Funk, senior director of non-perishables at Lund, in a presentation at the Teradata Partners conference here this month.

“This year, three weeks out, we have two stores — one is already 98% sold through, another 75%. But we know our biggest sales come during the three days immediately prior to the holiday, so those opportunities are already lost; there's no chance to reorder or fill in those last few days,” he said.

Better data visibility and decision support from Lund's newly installed data warehouse system are expected to change this circumstance in future seasons. Debbie Briggs, manager of information systems at Lund, said the company's previous IT reporting methods required merchandisers to pull data from multiple locations in a time-consuming process that delivered static reports. “This made it difficult to analyze data, and it limited our ‘train of thought’ approach to discovering opportunities.”

Briggs said that Lund received its Teradata enterprise data warehouse and Retail-Decisions software in November 2007 and deployed it 15 weeks later. That wasn't soon enough to impact this season's Halloween orders, but it's already having an impact in terms of delivering operational intelligence and faster decision-making.

“Our CIO vision was to create one version of the truth and provide visibility into that data,” said Briggs. Using a Teradata Appliance and software tools from Teradata and MicroStrategy, that goal is on its way to being met.

“Our category managers are more productive than before,” Funk said. “They know more about the business today. They know more than the suppliers. And the enthusiasm levels of the group are much higher.”

Funk said he is already looking ahead to how the new tools can support smarter buy-in decisions for next year's Halloween season. “In the future, we will be able to order more to forecast and take a reasonable amount of mark-downs and still do better,” he said. “I bet we will increase Halloween sales by 50% in the coming year due to better data and record-keeping. It will help prevent us from under-ordering.”

Brent Cobb, senior programmer analyst at Lund, said a total of 72 users, including 20 “power users,” had to be rapidly trained on the new system as it was brought on board. “We used an accelerated testing strategy that included unit and system testing and user acceptance testing. The latter also served as an initial training opportunity.”

The team employed a train-the-trainer approach to training and support. It was regarded as a great success, as it helped reduce user reliance on IT resources.

Asked about the cost of the data warehouse, Funk responded, “Ten years ago it would not have even been a consideration. Today a midsized retailer like ourselves can not only afford this, we have to afford this.”

“We started small and worked in increments,” Briggs said. “That made it more affordable.”