Catching up with EasyAsk

By John Dix
Network World, 03/11/02

In the year that has slipped by since last we looked at e-comm search engine company EasyAsk, it has landed big name accounts such as Lands' End, Talbots and Coach, released a Unix version of its software and added a new high-end product.

Not bad for a little company with 40 people.

To refresh your memory, EasyAsk is out to solve a problem best described with this real world example: Go to Ritzcamera.com and search for "Canon digital cameras" and the site returns one page listing two accessories and not one $400+ camera. The company should be weeping.

Now go to Coldwatercreek.com, a customer using EasyAsk's Precision Search product, and enter, "women's red sweaters, size small." This detailed search returns a bunch that fit the bill, including some that aren't even described as sweaters, like the "V-neck cardigan".

How does EasyAsk do it? Instead of relying on text relevancy like most search engines, EasyAsk grooms search queries and presents detailed requests to your product database.

Grossly simplified, EasyAsk uses tools to crawl your database looking at structure and nomenclature, and builds a custom dictionary that spells out how your products are categorized and what attributes are used.

The dictionary is loaded on your servers and used to translate the language of queries into the language of your database, for example, ladies to women's, pleated to pleat, cordaroy to corduroy (notice the spelling correction) and slacks to pants.

The next step is triangulation, associating query components with categories (women's, pants) and attributes (corduroy). Anything the system doesn't recognize will be dealt with in a text search.

The optimized output passed to your database may look like this: categoryID=102,149; Material=Corduroy; Description contains "pleats", says President and CEO Robert Alperin.

The company says its software can make your site searches 90% accurate. And the system learns and gets better by itself. Alperin says after six weeks 98% of Coldwater Creek inquiries were being answered correctly.

That should result in increased sales. Alperin says EasyAsk did a four week test at Coach in which half of all search queries were routed to Precision Search while the rest were handled the traditional way. After four weeks, for every $1 of sales generated by the old system Precision Search generated $4 in sales.

In beta test now is Search Advisor, a high-end product designed to help users realize they can use more than one to three words to search. With Search Advisor, if you search for "women's sweaters" the tool returns a screen showing some sweaters, but it also brings up a list of subheads for: Categories (Cardigans, Pullover), Brand (Anne Klein, Carole Little), Fabric Name (Blend, Cashmere), etc.

That enables the customer to focus while also letting the retailer show the range of its products, which should lead to increased sales.

Although EasyAsk's focus is still on Web retail, Alperin says the technology could help large business-to-business sites as well.

For those of us who have been frustrated by Web site searches, all we can say is bring it on.