Retailers look to make their site search engines smarter

As retailers deepen their understanding of site search, they—and more search technology vendors—are figuring out more ways to use it. Site search vendor Celebros now extends its offering beyond search and navigation to include help with on-site merchandising, based on how retailer customers are using its site search product. In the past six months, Celebros has added to its search engine, Qwiser, “machine learning” capability, which CEO Michael Crandell describes as “a smart salesman that watches certain metrics” in visitor behavior to guide merchandising.

Even if site search captures misspellings, rationalizes the use of different terms by searchers seeking the same product, and delivers results that are 100% accurate, retailers still face the issue of what to present first in those results. Celebros’s machine learning functionality monitors search behavior to feed it back into the search engine, which then adjusts its presentation of subsequent search results based on what it’s thus learned, in combination with business rules set by the retailer.

That lets a retailer who wants to prioritize best-sellers in search results ensure that best-sellers stay at the top, even as product status changes, for example. Retailers can also write business rules for the engine to adjust results presentation according to other monitored metrics or weighted combinations of metrics.

Celebros client and gift site PersonalCreations.com has increased its conversion rate by 15% since implementing the machine learning functionality on its site search, Crandell says. To facilitate the cycling of monitored metrics back into the search engine, Celebros has added a self-contained analytics package.

“It’s an analytics package that is focusing only on search and navigation behavior—what was the conversion rate at the cart for specific search terms, for example,” says Crandell. “That information is part of what feeds back into machine learning.”

Israel-based Celebros, which is focused primarily on clients in the small to mid-size range, has added 30 retail and manufacturer clients since launching in the U.S. in 2003, Crandell says.