Web self-service: searching for answers
Few people who visit self-service Web sites have escaped unscathed from the frustrations that all too often accompany their use or attempted use. Simple transactions such as checking a bank balance can usually be accomplished efficiently, but more complex needs such as finding information about a health insurance policy or how to obtain replacement parts for a product can throw the visitor into an inescapable loop.
Zachary McGeary, principal analyst at Jupiter Research, explains, "Customers often are not sure how to phrase their questions or search requests. In addition, over half the customers who report usability problems say that their searches offer too many results to be helpful."
Ideally, the Web site will offer flexibility in how the user seeks information, including the ability to process natural language queries phrased in a variety of ways, and will tune the search engine to produce a relevant and specific answer.
Water Pik manufactures a line of oral health and showerhead products for the consumer market. Several years ago, the company began providing the option for customers to contact the company by e-mail as well as through its existing call center. The number of e-mails rapidly ballooned to the point where more than two full-time employees were required to answer customers’ questions.
Water Pik needed to develop a more efficient method of responding to its customers. "We wanted to get customers the information they needed," says Joan Ritchey, consumer services manager, "in a way that made best use of our resources." The company looked at a variety of options, including automated e-mail responses. At a customer service association seminar, one of Water Pik’s managers heard about Astute Solutions’ RealDialog and considered it the best match for the company’s needs.
The application developed using RealDialog is called "Ask Waterpik," and resides at the top of Water Pik’s customer service page. Other options on the page include links to instruction manuals, to information about replacement parts and to a set of FAQs. In the Ask Waterpik dialog box, customers can type in their inquiry, using natural language. Rather than presenting users with a series of documents that may or may not answer the question, the RealDialog software provides a precise answer in natural language form.
To produce the answer, RealDialog searches through an information base that was developed through conversations with Water Pik’s customer service representatives (CSRs). The knowledge was easy to acquire because the CSRs knew all the front-line questions. One of the most frequently asked questions was how to get warranty support. Waterpik developed a "conversation" in which RealDialog walks the customer through the process of obtaining warranty support.
The Ask Waterpik function began with 75 questions and now contains more than 500 that are linked to appropriate answers. And because RealDialog understands natural language and intent, the customer can state a particular question in any number of forms and still get the right answer. If a customer asks a question three times, the system assumes they are not getting the necessary answer and, at that point, provides a contact form for the question, which is sent to the CSR. If the question proves to be one for which no information is available in the knowledgebase, a new answer can be added without IT involvement, after CSRs develop appropriate content.
When RealDialog was first deployed, 97 percent of customers’ questions were answered successfully. That rate has now increased to more than 99 percent. Staff who had been handling e-mail inquiries were able to focus on the more complex questions directly with customers.
"Customer service is a top priority for Water Pik, and this tool has enabled us to meet our consumers’ needs," Ritchey says.
Poor search capability is a major reason that users become frustrated with self-service Web sites, according to Joe Sanda, president and CEO of Astute Solutions. "Many sites have a search tool that produces a list of results, which does not necessarily help consumers who are looking for answers," he says.
Another key factor is being able to escalate out of self-service if the answer is not forthcoming. "Web self-service should be integrated with the contact center," Sanda continues, "so that users don’t get stuck if the answer is not already available."
RealDialog allows for continuous enhancement of the knowledgebase. "New questions can be analyzed by RealDialog via machine learning, and clustered into topics," Sanda says. "Our linguistic analysis identifies gaps in the knowledgebase."
Searching for technical answers
Mentor Graphics produces software that helps engineers develop electronic products. Its sophisticated design software is supported by a call center staffed by engineers. After developing a knowledgebase that could be used by the call center staff, Mentor Graphics began to consider providing its customers with access to the information as well. However, two problems emerged. First, the knowledgebase was limited. CSRs sought information in technical documentation on an ongoing basis and pasted excerpts into the knowledgebase, but it still did not cover the topics comprehensively.
The second problem centered on search functionality. "We thought about making all the technical documentation available online to broaden our information resources," says Tom Floodeen, VP and general manager of customer support at Mentor, "but the search engine for the knowledgebase only worked for the content in that repository, and our keyword-based Google appliance did not successfully find the answers our staff needed." Bringing up a document that the user then had to search further was not helpful.