Enterprise search is a tough nut to crack. You are never going to find a product that -- out of the box -- solves a significant portion of your organization’s search needs, according to consultants and analysts. As a result, implementing an effective enterprise search platform is an ongoing process.
The first step is to buy the right enterprise search engine, those experts agree. But that often is easier said than done.
Many Web content management systems come with built-in search engines or can be integrated with specific search software. Sometimes that’s the right tool for an organization, sometimes not, said Miles Kehoe, president of New Idea Engineering, a consultancy in Santa Clara, Calif., that specializes in search technology. “I’ll tell people that the best search engine for them is the one whose founders did a Ph.D. on data that’s like their data,” he said. In addition, enterprise search typically involves looking for more than intranet or other internal website content; the need to locate documents stored in shared drives and information kept in other repositories must be taken into account as well.
Learn more about effective implementation of enterprise search platforms
Read about why the enterprise search platform business case relies on numbers – and more
Learn how NASA and Google tackled a major enterprise search platform project
Before you buy an enterprise search platform, you need to know -- at least approximately -- what you want to get out of it, said Susan Feldman, research vice president for search and discovery technologies at Framingham, Mass.-based IDC. Do you want to give end users 200 answers to a query or one best answer -- or do you want to give them answers at all? Maybe they’d prefer to explore all the possibilities. “Companies won’t be happy unless they’ve figured that out,” Feldman said.
Limit your project scope
Be discerning about the scope of an enterprise search project, advise analysts. Do you really want to spend the time and money to include all those Lotus Notes archives and old file shares that are still in place throughout the organization? Are they necessary?
“Be picky about the criteria you use to merit when something is or isn’t part of your search initiative,” said Leslie Owens, a senior analyst for content and collaboration at Forrester Research Inc. in Cambridge, Mass. Too often, she added, companies end up including content that is not carefully curated or thought through, which can lead to performance roadblocks and other issues down the road.
“Use taxonomies to fuel automated approaches to enriching content with metadata,” Owens said. For the most part, that likely will consist of automatic tagging in bulk. While metadata can help improve search processes a great deal, she explained that it’s unrealistic to expect end users to tag content in an enterprise search system to benefit future search and analytics applications.
Enterprise search results: 60 guys named Sarah
Metadata can also be misleading. Kehoe related the story of a search for a woman named Sarah in one company. Surprisingly, most of the first 60 hits were for men. It turned out that everyone at the company had published their work profiles online in a standard template. And the name of the person who created the template was, of course, Sarah.
“If you’re not aware that kind of stuff is happening,” Kehoe said, “you’re really at a loss.”
His story illustrates not only the need to be able to tune the settings behind the search results, but also how essential it is to monitor the results produced by an enterprise search platform. For those reasons, it’s imperative to use search engine tools that provide monitoring and tuning capabilities, according to Kehoe and other analysts.
Monitoring is also an important aspect of building effective enterprise search taxonomies. “The most important taxonomy for your site, whether it’s inside the firewall or outside, is the taxonomy of words people use when they search your site,” Kehoe said.
If enterprise search is broken, fix it
Once you know what people are searching for, make sure they’re finding it. If they are not, find out why not. For example, take a close look at the top 50 searches and make sure they produce good results, Kehoe recommended. If not, it might require tweaking the search algorithm or your meta-tagging, or both. “Sometimes it’s as easy as adding a word to the document fields,” he said.
If there is no clear set of top searches -- in some enterprises, most searches are unique -- it then becomes necessary to monitor queries. The trick, Kehoe advised, is to watch their trails throughout the search system, try to build relevant search taxonomies and then attach accurate metadata to the documents and other content users are searching for.
No matter what search engine an organization picks, making it work comes down to managing the system effectively. Track results and keep making incremental improvements as needed, Owens suggested. “For example, people really like auto-complete,” she said. “That’s a little thing you can do that really helps.”
“Past a certain level of functionality,” Kehoe said, “almost any search engine will work if you watch it. It’s a cycle of monitor, tune, measure and repeat.”
Organizations that ignore those proscribed steps do so at their peril. Without such methodology, Kehoe said, an enterprise search platform will not operate the way it should. “Search is something people are never happy with,” he said. “But they’re never willing to spend the time to make it work.”
Mark Clarkson is the author of seven books and hundreds of articles on all things technical, from CAD/CAM to Photoshop. He lives and works in Wichita, Kan., where he rarely leaves his basement and maintains his website at markclarkson.com.