The e-retailer spends at least 50% of its monthly display ad budget on the highly targeted, data-driven—and often cheap—ad placements using programmatic platforms.
(Page 2 of 4)
Endeca has created customized landing and guided search pages based on site search queries for such clients as Patagonia.com, NorthernTool.com, and CircuitCity.com. It also created the so-called store within a store for eBags.com, in which consumers can search by brand name.
Creating rules through testing
Determining what works and doesn’t is a matter of A/B testing. With enough testing, business rules can be put in place against customer profiles, search terms, merchandise categories, or any combination of those, according to Wolf.
The advantage of A/B testing of site search results and related landing pages is that the results empower merchandising executives to be more active in designing site search results. “A/B testing is a way to control the customer experience and to tweak it accordingly,” adds Wolf. “The more retailers can leverage that control, the more likely they are to create a point of differentiation with the customer and become the first place the customer stops to shop.”
Using test results to create dynamic landing pages or site search results makes it possible to quickly test merchandising strategies built around keywords or categories. At some point, most keywords and category searches lose their relevance as new phrases and keywords enter the site search lexicon, according to Bryan Surles, director for technical field operations for Mercado. “Tracking the amount of traffic associated with each keyword or category helps merchandisers determine the cost effectiveness of the word as it relates to the merchandising strategy,” he says. “Some clients manage more than a million keywords. Having insights into the traffic associated with those words can help retailers to create dynamic merchandising strategies around them and the points of entry to the site they open up.”
Mercado has developed a user console within its site search application that automatically suggests ways to optimize merchandising strategies around site search. The console, which was created with input from long-standing users, will generate an alert suggesting a synonym be created for a product that cannot be located based on customer queries. In most cases, the product that can’t be located is not cataloged under the term consumers are using to search for it.
Mercado corrected such a problem for OfficeMax.com when consumers looking for correction fluid repeatedly entered “Liquid Paper.” “The product cataloged was White Out, and entering ‘Liquid Paper’ was not recognized,” recalls Surles. “It was a simple fix and executed quickly because management was alerted to the problem.”
Identifying and correcting such problems also makes it more feasible for retailers to identify new keywords to which they can attach promotional content or other information that can build and deepen the customer relationships. “It is about leveraging the expertise of e-commerce managers to make it easier for them to fix, fine tune or tweak their site,” adds Mercado’s Leibow.
Retailers that take the time to think through the depth of their search and how best to display the results to the customer are more apt to deliver a wealth of information that can close sales and lead to upsell and cross-sell opportunities without overwhelming the customers, according to Thunderstone’s Howitt.
“As consumers crave more information on a product, you gain their trust by letting them search the product manuals or reviews, in addition to the catalog,” he says. “But if you’re going to do that, then organize the results too. It’s more useful to provide a choice of categories, instead of just one list of results.”
To illustrate his point, Howitt uses the example of a customer searching for slip-on shoes. Naturally, the search engine has no way of knowing whether the customer is looking for men’s or women’s shoes.
The solution is conducting a search using combined structured and unstructured logic. In this case, the phrase “slip-ons” is searched within the full text of the site, but separating the results into men’s and women’s shoes requires using the category structure inherent in the product database, which directs shoppers to the right product.
“It’s also important to do that efficiently, because it’s more computing-intensive,” says Howitt. “Providing such sophisticated functionality is one of Thunderstone’s specialties, and we have optimized our tools to perform such complex business logic quickly.”
Parsing the data
Parsing data so effectively requires a flexible application that can be easily integrated into the retailer’s operating platform so that data can flow between disparate sources and be stored in a common database. “It is important to generate reports on the data flowing into the site search catalog and how it is being used for retailers to know where the holes are in their knowledge base,” says SLI Systems’ Ryan, who holds a PhD in neural networks. “It is not uncommon for knowledge holes to lead to poor customer experiences.”
SLI Systems’ search application, which is a hosted ASP solution used by more than 100 web sites, learns as customers enter product or catalog queries. The more frequently a keyword is entered, the faster the application learns what results lead customers to close sales, upgrade their purchase, and buy related items. The information can even be used to enhance results on search engine sites to boost natural search results or increase sales from paid search.
“We’ve created related search pages for clients that Google can pick up and that lead the customer to where they need to go on the retailer’s site,” says Ryan. “We can also help determine keywords for paid search.”
The frequency rule
Another way to close site search data holes is to measure how often an action is taken by a customer in relation to the business rule put in place on the site search engine for a particular query.