Sears eyes big data for dynamic pricing, cost savings

Keith Sherwell, CIO of Sears Holdings, recently outlined the retailer's technology overhaul, which depends on Hadoop, open source and cutting maintenance costs.
Written by Larry Dignan, Contributor

Sears is revamping its technology infrastructure with a focus on open source and big data and the cost savings expected to go with it.

Keith Sherwell, CIO of Sears Holdings, recently outlined the retailer's technology overhaul. Sherwell joined Sears in 2012 to re-architect the company's technology infrastructure.

Sherwell's key points were delivered at a big data powwow hosted by Cowen & Co. Peter Goldmacher, an analyst at Cowen, relayed the rough sketch of the big data moves at Sears.

We'll be examining similar use cases for big data at our invite-only TechLines panel in New York on Oct. 4.

Also: TechLines panelist profile: Ford's Michael Cavaretta on internal big data | TechLines panelist profile: NASA's Nicholas Skytland on big data literacy

Among the key points on Sears' big data moves:

  • Sears claims to have one of the largest Hadoop clusters in retail and the company is betting heavily on open source.
  • Sherwell is conscious of technology costs given that affects Sears profit margins. Sears realized it was using old technology and upgrading would cut half of the costs.





  • Sears has a dynamic pricing system and that's where big data comes in. Sears has to compare its products to the pricing at other retailers.
  • Big data could also play a role in analyzing weather patterns and how they impact sales.
  • Sears is consolidating 10 data warehouses to two. One of those data warehouses will be entirely open source.
  • Sherwell is moving away from Oracle and Netezza software. Biggest move for Sears is moving its loyalty program off of Exadata to go open source. The savings is estimated to be "millions in fees per year."
  • Dell and HP storage is overpriced relative to Hadoop, said Sherwell.
  • A Teradata-based data warehouse will remain in Sears' plans because Hadoop doesn't join two sets of data together well.
  • The most relevant big data vendors will be the ones that are able to present data well.
Editorial standards