Most of the buzz about online search focuses on search engine marketing. But onsite search — the effectiveness and user-friendliness of your Website’s search function — is also an essential part of doing business online.
According to “The State of Retailing Online 7.0,” a study by Cambridge, MA-based Forrester Research, 48% of visitors use the search function when visiting an e-commerce site, as opposed to finding merchandise by drilling down or using other navigational tools. And given that use of onsite search has risen 50% during the past few years, according to David Schatsky, senior vice president of New York-based Jupiter Research, it’s safe to say that online merchants would be wise to invest in improving their onsite search functions with better dictionaries and thesauri and additional options.
The first step in creating a successful search function is to determine if the current search is meeting the needs of site visitors. This entails comparing conversion rates for your overall site and for your site’s search function to those of other online retailers, says Rich Stendardo, retail solutions manager at Cambridge, MA-based search technology provider Endeca. Looking at the average conversion rate of specific search terms used by consumers on your site, Stendardo says, can also help you assess if proper search priorities are in place.
According to DoubleClick’s November 2004 “E-Commerce Site Trend Report,” the average conversion rate for search-driven purchases was 2.1% in the third quarter of 2004, up from 1.5% the previous third quarter. The overall site conversion rate for the third quarter of 2004 was 4.6%. During the same period, 9.3% of all sales came via search, compared with 6.6% a year earlier.
If you find that your onsite search is falling short, chances are it’s because of one or more rather common problems. The good news: These problems are easy to fix.
PROBLEM 1 The search function doesn’t retrieve all the applicable matches — or it retrieves too many.
A recent search on Amazon.com for “evening dress” retrieved a whopping 6,000 results, including watches, books about dresses, and a My Little Pony toy as well as actual dresses. Another recent search, on Sundance Catalog’s site, for “blue jeans” retrieved two results — neither of which was a pair of pants. Interestingly, both problems most likely have the same cause: a weak dictionary and thesaurus. Overly broad match algorithms (in which a term is deemed a match regardless of context or without regard for restrictive criteria) can make matters worse, as can a search function that draws results based on only a handful of product attributes or data fields.
When online toys merchant eToys launched in 1999, it used a basic out-of-the-box search function. Brian Bass, director of product development for Denver-based eToys, says that the search function’s dictionary included no more than 600 words and was updated only a few times each year to include commonly misspelled words. In addition, the search function drew results based only on four attributes (product name, manufacturer, and short and long descriptions). As a result, eToys’ search function had difficulty retrieving product results because of its limited dictionary — though at times it retrieved too many because of its broad keyword recognition.
In August 2003, eToys went to Endeca for help with its onsite search and came away with improved basic keyword and phrase search, giving customers more-relevant results. The new search also allows eToys to establish manual search redirects, which were programmed to override Endeca’s results and instead bring consumers to one of eToys’ more than 400 product-specific boutiques. What’s more, it draws results based on 17 product attributes including product name, toy character name, brand, and doll ethnicity, providing much more relevant results.
PROBLEM 2 Search function is not specific to your industry. Restaurant supplier Ace Mart’s onsite search function lacked an effective industry-specific dictionary and thesaurus. Written inhouse in 2002, Ace Mart’s first onsite search function offered only a manually updated dictionary and a rudimentary thesaurus, says Ryan Rodkey, webmaster for the San Antonio, TX-based merchant, and neither took into account spelling and vernacular variants.
Ace Mart turned to Santa Barbara, CA-based search technology provider Celebros to upgrade its onsite search last year. It sent Celebros a database of product information, which the provider mapped into its system to build industry-specific category groupings and subcategories. The current search database is updated daily with information Ace Mart supplies to Celebros regarding product offerings, promotions, and holidays. Celebros also built the company a dictionary and a thesaurus that recognize restaurant industry terms that Ace Mart provided to Celebros.
Ace Mart’s new search function allows customers to refine their search by searching for specific terms within the results. Before it offered the refinement option, Rodkey says, half of its users used to sift through the first two pages of results and drop off before making a purchase.
The onsite search overhaul produced favorable results, says Rodkey. From Sept. 14 through Oct. 13, 10,000 visitors used Ace Mart’s search function, an increase from 7,000 searches between June 13 and July 13. More important, the number of products sold that could be directly linked to search more than doubled during the same time period, from 2,800 items to 6,000.
PROBLEM 3 Misspellings stymie the search function. Part of understanding natural language is deciphering common misspellings. “Misspellings are huge,” says Amy Africa, president of Williston, VT-based Web marketing consultancy Creative Results. Endeca’s Stendardo says that although the number of poor searches resulting from misspellings varies greatly, “crummy” searches happen more often than good ones.
A quality dictionary and thesaurus that “learn” from consumers can help overcome these differences, says Stendardo. A technologically dynamic dictionary will update itself by recognizing common logical permutations.
Not every merchant is willing or able to invest in a dynamic dictionary, of course. The function generally isn’t available from off-the-shelf packages, and costs for customized search functions range from more than $50,000 to hundreds of thousands of dollars. The other option is to manually update your dictionary, after periodically analyzing search logs to determine what terms are the most commonly misspelled or misdirected.
PROBLEM 4 The search function doesn’t allow visitors to narrow down the number of results. According to Jupiter Research, the number-one function desired by online consumers making purchases of more than $100 is the ability to refine results by attribute; 34% of online shoppers surveyed said they sought this sort of improvement.
“The more that users are allowed to be in control of their destiny, the more successful a search is going to be,” Africa says. By helping shoppers refine their search results, merchants are helping to improve their conversion levels, by ensuring that the products the searchers want appear on the first page of the results. As Celebros CEO Michael Crandell says, “Whatever’s not on the first page doesn’t exist” — or might as well not exist.
The easiest way to determine which refinement options to offer is to “know your customers,” says Stendardo. Are they buying items for themselves or gifts? If gifts, are they for boys or girls? Answering such questions should enable you to determine why shoppers visit your site and how they search, so that you can provide them with their most useful refinement options.
Before upgrading its search function, Bass says, eToys offered only basic refinement options for customers to narrow down product results by attributes such as product type. Now the engine allows for refining searches by the same 17 categories it searches with. All details for eToys search function were determined during strategic requirement meetings with Endeca. The changes, Bass says, have resulted in higher conversion rates and a 20%-25% increase in page views.
PROBLEM 5 The search function lacks imagination, which means you’re missing out on cross-selling and upselling opportunities. You may not think that suggesting additional products to visitors is a service to them, but it is. Thirteen percent of respondents to the Jupiter Research survey of consumers making purchases of more than $100 want Websites to make product recommendations — for instance, if a consumer searches for black pants, the results should call up all available black pants with links to complementary products such as coordinating belts in a navigational column near the desired results.“These are relevant alternatives that are welcomed by the customer rather than off-the-wall or not logical results,” Crandell says.
EToys is using a dynamic merchandising function, which delivers additional relevant content. Acting as part of the search engine and maintained by the retailer instead of a search provider, the dynamic merchandising function retrieves promotions and related products based on rules of association established by the retailer. When Endeca created eToys’ dynamic merchandising function it allowed only for 50 rules; eToys decided to expand those to 1,000 rules to offer a wider breadth of suggested product feedback.
Crandell says searchers like to know what others are buying online. This additional information to help in decision-making is exactly what Schatsky recommends retailers provide on the search results pages. Crandell agrees, adding that customers also want comparison functions and product reviews to help them decide what to purchase.
When launching or upgrading an onsite search function, you essentially have four choices: an out-of-the-box program, a custom-altered out-of-the-box solution, a solution built by a third party, and a solution built inhouse.
Providers of off-the-shelf search software include:
Suppliers that will create and customize a search function include:
Fast Search & Transfer (FAST)