
Every website has it. That small box in the header with the picture of a magnifying glass to the right. We call it the search box, and it's kind of the unsung hero of your e-commerce site. And maybe it isn't the sexiest topic in the world, but it doesn't have to be because site search has a return on investment story that speaks for itself. Retailers typically report two or three times the amount of conversions for site search users.
However, one question still remains: Which approach to site search best connects shoppers with the products they're most likely to buy, making shopping easier and retailers more profitable? Do online shoppers use natural language search, which interprets subjective terms to serve up search results? Or do the most relevant search results come from learning search, which "learns" what specific search terms resonate most with consumers and re-ranks the order of search results based on the latest activity of users?
Testing ‘Natural Language’ Search
In e-commerce, there's some uncertainty around the demand for natural language search. For instance, Charles Caison, e-commerce manager at North Face, says, "most shoppers search using terms that describe the product, not ambiguous phrases that require natural language processing to decode."
Lakeshore Learning, an education supplies retailer, also finds less use of natural language search from its shoppers. Lakeshore recently analyzed its top 1,000 searches and found that consumers use an average of 1.8 words to search. This report indicates that consumers today search via keyword search and that natural language-type searches are still infrequent.
A new SLI Systems study supports Lakeshore's findings. To demonstrate site search user behavior today, SLI evaluated natural language terms, focusing on subjective search terms including "cheap," "nice" and "cute" for a Fortune 100 retailer. Out of 67,000 searches (see chart below), the word "quality" was only used three times while "cheap" and "nice" had similar results. The findings reveal that subjective search terms aren't yet commonly used among online shoppers.
