Finding meaning in search

Search has been a mainstay of knowledge management since the discipline was first defined, but is not yet a perfect technology. Users want search to be simple and yet precise, a balance that can be difficult to attain. "The main issue is relevance," says Leslie Owens, an analyst at Forrester. "People may make the same general query, but don't all want the same information as output. It's hard for any technology to match up the users' information needs with all the possible results."

Features that allow user ratings, voting or comments add a new dimension to search. "The premise is that the past behavior of like-minded searchers—‘the crowd'—will help you find the right answer," says Owens. "It works well for the top queries, but less so for unique or obscure information needs."

Semantic search is oriented toward discerning meaning in content to produce relevant results, but requires groundwork to function well. Organizations that spend time analyzing their content and establishing metadata are able to leverage their intellectual capital more effectively.  "With semantic search, unified access is a primary objective, to be able to treat both structured and unstructured information across silos," Owens says. "The source of the information is transparent, and analytics are used to bring out key ideas on which taxonomies can be built."

Accessing Intellectual Capital

Mercer provides human resource consulting and related services in 40 countries throughout the world and is a wholly owned subsidiary of Marsh & McLennan Companies, a professional services firm specializing in risk, strategy and human capital management. To harness its collective expertise, Mercer began looking for an enterprise search solution that would have the precision and versatility that the company needed.

As of 2009, Mercer had several search solutions in place, but they were not meeting the needs of Mercer's professional staff. "We have exceptional intellectual capital," says Marcia Robinson, global leader for knowledge management at Mercer, "but our colleagues can't use it with our clients if they can't find it." The ultimate goal was to have a system in place that allowed users to quickly access relevant information through a single interface that provided targeted search and the ability to conduct discovery research, no matter where the information was located.

Mercer formed a project team that included its knowledge management group and the IT department to carry out a formal evaluation of enterprise search products. The team gathered user requirements from business groups throughout the company and, after reviewing the capabilities of available search solutions, narrowed the options to five contenders. From that group, the team selected two finalists, and provided the vendors with a standard configuration of servers and an identical set of documents to index. Mercer employees who were familiar with the content tested each system, which underwent further vetting by the IT department and a comparison of costs.

Mercer Implements Pinpoint

When the evaluation process was completed, Sinequa emerged as the best match for Mercer's requirements, because it provided excellent relevancy, utilizing keywords as well as faceted navigation. Sinequa was simpler to configure and customize than other products, and was competitively priced. Haroon Suleman, enterprise search architect at Mercer, says, "We liked the combination of Sinequa's ability to retrieve information precisely based on search terms, and to provide faceted navigation for in-depth research."

As the Sinequa enterprise search system, which Mercer is calling "Pinpoint," was implemented, Mercer's team worked closely with Sinequa to review the metadata contained in the initial group of documents to be incorporated, and to integrate it with its "People Directory," a database of employee profiles. When the initial implementation was completed, the system had indexed 40 million documents. "Our users can narrow a search from 40,000 documents to 20 in just a few clicks," Suleman says. "This has provided us the rapid retrieval of high-value content that we were hoping for."

The integration of the People Directory allows linking of people to content through metadata. "If the author field for a document maps against an employee, those are linked and presented in the search results," Suleman explains. "In addition, Pinpoint can detect the individual's online presence and launch our instant messaging application." The integration of knowledgeable employees is important because it allows Pinpoint to locate not just documents but also corporate expertise for project staffing, and to tie it to digital intellectual capital.

At present, Pinpoint searches documents on Mercer's intranet and file servers in Europe. They include Microsoft Office documents, a Livelink repository and SharePoint files. Future work will look to incorporate other high-value content sources such as client and project.

The Glue Across Knowledge Systems

Much of the success of the Sinequa system can be attributed to the effective teamwork between knowledge management and IT groups. "Our knowledge management group took on the challenge of understanding user needs and communicating them to the IT department," says Robinson. For the search project, knowledge management also provided communication, training, support and development of policies related to knowledge sharing, and collects feedback about system use for continuous improvement.

Another key element of the implementation was the understanding of Mercer's corporate content and how users wanted to access it. "In cooperation with our users, we developed a core taxonomy that covered major topics such as the geographical market or topical issue to which a document related," says Barbara Fiorillo, global content manager in the knowledge management function. "Without taxonomy, our users would have to struggle through a long list of results that was much less precise." A list of categories to the side of the results allows users to drill down into areas of interest.

Through careful consideration of its strategic goals and a systematic approach to implementation, Mercer has moved closer to realizing a comprehensive enterprise search system. Pinpoint has already achieved the hoped-for goal of integrating disparate repositories of information. "Search is the glue across our knowledge systems," Suleman explains, "allowing our users to access information quickly, easily and securely. As we expand the amount of searchable content, Pinpoint will become even more effective."

Sinequa was established in 1984 as a private research lab working in the area of linguistics, out of which grew an enterprise search solution focused on extracting meaning from a rich, semantic-based index. "Sinequa is designed to present information in a contextually significant manner," says Alexandre Bilger, CEO of Sinequa. "The goal is to produce a short list of highly relevant results for a given user rather than a big list that the user will have to sort through."

Part of relevancy comes from understanding who the user is. "Relevance for each group of people is different," Bilger says. "Presenting information in a contextual manner, depending on whether the individual is in the R&D department or sales, for example, makes it much more useful." R&D is a common entry point for Sinequa customers, particularly in large firms, because no one may have a complete view of all R&D projects and all of the subjects dealt with across the company. "Organizations can find synergies among different projects," Bilger adds, "and teams of people who have already worked on a problem, rather than starting from scratch."

Some of the traditional ways of organizing information are no longer useful, according to Bilger. "If people are spending a lot of time putting their e-mails into categories, it is a waste," he maintains. "It's much better to apply a search engine and generate the categories on the fly in particular, since these categories change over time." Because Sinequa provides a platform for content analytics, multiple applications can be built to generate different categorizations based on the same information. "With a custom classification system, multiple views are possible that are highly responsive to each user's needs," Bilger says.

Selection Dilemma

The plethora of search options makes selection difficult, especially for companies that have more than one type of search need. SharePoint users can stick with FAST, open source (see sidebar on next online page) is expanding as an option, and Google's search appliance provides yet another alternative. Several vendors have focused on developing embedded search applications within functional areas such as CRM.

"One technology cannot meet all types of search requirements," says Iain Fletcher, VP for marketing at Search Technologies. "Users should consider carefully what their needs are, and gear their choice toward the top priorities." In addition, they should learn enough to set their expectations realistically. "Some of the disappointment in search comes from not understanding the technology or their own requirements well enough."

Clean Content

Although customers are drawn to the "plug and play" approach for its speed and ease of implementation, product selection should be matched to the application. "In many cases, plug and play can be tweaked and the customer is all set within just a week or so," Fletcher notes. "These tend to be point applications for websites or relatively simple intranets." For complex enterprise search applications, more preparation is required.

Search Technologies implements search solutions, and spends a large proportion of its time helping customers prepare for the implementation. "When an organization has 20 different data sets, it needs a management system in place to normalize the data and obtain a consistent level of content cleanliness," Fletcher explains. "And a lot of our work is geared toward generating good metadata."  In some industries, such as pharma, the need for metadata is anticipated and provided for, while in others, the need is discovered somewhat after the fact. "Search navigators, one of the most powerful tools in search technology, rely on metadata," Fletcher emphasizes. "So without metadata, search precision is limited."