The adage “familiarity breeds contempt” captures a root cause of many enterprise search failures.
“Public search engines like Google, Bing and Baidu are part of the fabric of everyday life,” says Darin Stewart, research analyst, Gartner.
“They are simple to use, require no maintenance and generally perform well. Many executives and end users expect enterprise search to deliver similar results with similar investment and effort.
“This is an unrealistic expectation, and sets up the search team for failure before a search platform is even deployed.”
Stewart says web search and enterprise search share many concepts and depend on similar technology stacks.
Even so, he believes those elements must be applied and managed differently, depending on which side of the firewall they are expected to search.
“Too often, organisations approach search in a simplistic, one-size-fits-all, point-and-shoot manner,” he explains. “Effective search requires much more than simply crawling and indexing content.”
According to Stewart, end users contribute to poor search performance as well, by taking an overly simplistic approach to query formulation.
On average, an information seeker will provide 2.73 terms describing the information need, and will only attempt a search twice in a given session.
“If the right information isn’t located in those two attempts, the user assumes it does not exist or that it is, at best, inaccessible,” he observes.
For Stewart, standard textbox query interfaces encourage these bad queries.
“The average search box is 20 to 30 characters wide,” he adds. “While most will accept more characters, the limited display space is a visual cue that short queries are expected.
“Compounding the problem is the fact that most end users simply do not know how to write good queries.
“This does not necessarily mean they need a full understanding of advanced Boolean logic for general information retrieval.”
In most cases, Stewart thinks users simply need a few more judiciously selected query terms and an understanding of how they work together.
More advanced search scenarios — for example, looking for genetic markers in the life sciences — will require more advanced techniques, but these can often be handled by the application itself, rather than overburdening the information seeker.