Mentor Graphics took a stab at resolving the problem in 2002, ultimately deploying what Egli calls, in retrospect, "a Band-Aid" solution. "The InQuira salesperson was fantastic and really educated us" about ontology, NLP, and the power of search, she says of the meetings the support team had with various search providers at the time, but Mentor "couldn't get the budget to implement InQuira right away." So instead, the team settled on a "lower-end search appliance that was cheaper but not as comprehensive."
What they really needed, Egli says now, was "the power of the language capabilities that InQuira had. We needed something that could search to find meaningful answers"—something powerful enough to parse and understand Mentor's complex product hierarchy, which exceeds 500 product areas. "We have so many different product segments that it becomes hard to troubleshoot online," she explains. The company also needed a solution that supported its use of PDF manuals.
By the following year, Mentor support team members had educated themselves about what they needed and why they needed it. This time, their pitch for the "hundreds of thousands of dollars" Egli says they ultimately needed to implement InQuira Intelligent Search was successful. By mid-November 2003, Mentor was working with John Baum, senior director of InQuira's Professional Services organization, to create an enterprise-wide knowledge base for SupportNet. That process involved collecting and storing more than 300,000 pages of PDF documentation on the SupportNet site, leveraging the knowledge of Mentor employees to enhance existing support documentation, and developing a more sophisticated search function that could understand the intent of users asking the same questions in different ways.
The resulting KnowledgeBase, as it's known, gives registered users 24-7 access to thousands of up-to-the-minute solutions via TechNotes, AppNotes, and links to detailed technical information. The searchable database crawls structured and unstructured content from web pages, databases, applications, and other repositories and then indexes that content for its meaning. "The secret to InQuira is our ontology and our dictionary," Baum says of the company's core Intelligent Search technology, which powers NetSupport's KnowledgeBase. "Part of what we do is take our out-of-the-box understanding of language for a particular industry and then fine-tune it for a company's vernacular." The other element in the dictionary tuning has to do with word patterns. "If you take certain concepts and put them together in a question, you can trigger certain responses," Baum says. "Instead of looking at individual keywords expressed by users, InQuira's languageunderstanding model drills down to the concept itself."
When searches fail to return the information necessary to resolve a question, users can click on the "open service request" button at the top of the results page to submit a request for further assistance. The three-step process asks users to describe their problem; returns five possible solutions; and, if none of the suggested solutions addresses the problem, submits the service request. After following up on such requests, Mentor invites users to complete an online survey to determine why they visited the site and to identify what they couldn't find; support staff then use that information to improve the KnowledgeBase and refine the ways in which that particular piece of information is delivered.
All told, Baum says about 50 people—document writers and support engineers among them—"touched the project in one way or another" by the time the new and improved SupportNet went live in April 2004. Refinements continue, of course, but the results achieved thus far have been impressive. On a seven-point scale, customers have given the site an average rating of six. Tracking reports suggest that the site is meeting customer needs in 62% of all technical troubleshooting-related interactions. For other types of interactions, the "success on the first try" rate is as high as 89%. Getting customers to first visit SupportNet for technical assistance has paid off in other ways, too, reducing Mentor Graphics' overall support costs by roughly $8 million annually and helping improve overall customer support margins by 10%.
At some point, Egli says, Mentor will refine the site even further by deploying Information Manager, an InQuira module for identifying, acquiring, organizing, and sharing enterprise information with customers or call center agents who work with customers. Leveraging Intelligent Search's ability to automatically index any information asset by the semantic meaning of its content, Information Manager eliminates the burden of manually tagging information assets and ensures that they're catalogued consistently for retrieval. "We spend a lot of time thinking about knowledge creation and knowledge management," she says of the company's future plans for the site. "We have to, because our products are so complex."
"That's one of the pleasures of working with Mentor Graphics," Baum notes. "They already had a knowledge-centered philosophy when we arrived."