Boosting site search relevancy with the right classification

Manual entry into a database is the last mile in getting product information from vendors, manufacturers and internal sources into a form that site search technology can pull from in response to customers’ queries. And as with any manual function the execution may not always be perfect. Version 3 of Celebros` Qwiser search product, to be announced within the next few weeks, helps catch those errors with a classification function built right into the product. That`s functionality previously outside the realm of most site search products, according to CEO Michael Crandell. Frequently, online stores seeking a classification engine to aid search have had to obtain it from a third party. The benefit is in fewer inaccurate or irrelevant search results, he says.

Databases pull data from sources that in turn provide varying qualities of product description, Crandell says. “That puts a lot of pressure on the manual data entry process to be completely accurate and properly categorized to begin with.” Glitches can include typos, misspellings, and product not correctly categorized.

For example, a searcher looking for “Merlot” as a refinement of “red wine” won’t see Merlot in search results if it’s been incorrectly labeled in the database as a white wine. Crandell says Qwiser’s classification engine basically “understands” text descriptions in relation to a defined hierarchy to see if a product has been classified correctly, and flags for adjustment items that may not have been. The automated classification function backs up manual entries to the database and thus increases the reliability of the search function’s relevancy by another notch, he says.

That’s one usage of one of several new features upcoming in Qwiser’s next iteration, Crandell says. Driving most of those enhancements are two underlying trends he sees in site search: the desire of site operators to automate as much as possible to save time and resources, and the desire to still have the flexibility to easily make changes manually, when needed, as product databases grow and change.

“We are seeing over and over from online stores and directors of e-commerce that as e- commerce matures, and the demands get higher, they have more and more specialized requests, tweaks and modifications they want to make to the system,” Crandell says.