Until a couple of months ago, a BellSouth residential customer using the company's Web site probably had a hard time finding information. On the search page there were three options – one for its product and services catalog, one for its business and people directory and one for its FAQs. With so many options, it was hard to figure out where to begin or what the results would yield.
That all changed on May 10, 2005, when e-Service senior manager Ed Dauginas and his team launched the Atlanta-based company's new natural language search engine for residential customers. It consists of a single search box powered by San Bruno, Calif-based InQuira.
BellSouth realized the need for an extreme "search makeover" when it started to receive negative customer feedback about the search tool. As a result, the company noticed an increase in e-mail and call center volume, and a decline in search traffic. Not only was BellSouth concerned about customer dissatisfaction, it worried about losing angry or frustrated customers to competitors with more user-friendly sites.
BellSouth worked nine months fine tuning and implementing its natural language search tool. Dauginas and his team started by researching those customers who had used the search engine and what terms they were looking up. Based on those results, Dauginas and his team decided on what type of search product would best suit BellSouth. It also helped them understand the scope of the questions that customers had --
"Our customers were asking natural language questions rather than [searching by] key word. Based on our customer insight, and the information and analysis about the kinds of searches that were conducted before, we opted for a natural language search engine," Dauginas said.
He also learned that customers' questions were often as detailed as, "Can I get DSL?," Who is BellSouth's CEO?," "How much do I owe?," and "Is BellSouth hiring?"
But in order for the search engine to return accurate information, Dauginas needed to tackle BellSouth's tens of thousands of consumer documents, most of it related to DSL support.
"Getting your arms around a search application -- especially for larger companies that have a lot of content publishing tools and content in general -- is the biggest challenge with any search deployment," Dauginas said.
So Dauginas and his team began the arduous task of indexing BellSouth's content. The team analyzed the content from each department and determined whether it needed updating, deleting and/or creating in order to present the most useful and relevant data to customers.
After it was indexed, Dauginas and his team ran test searches to gauge the accuracy of the tool before and during the official testing phase, which enabled them to make continuous adjustments. For example, they performed a search on "call forwarding," which brought up results ranging from product information to public speeches where the phrase "call forwarding" was mentioned.
"We saw [a wide range of results] in some cases, and adjusted the tool and the indexing to ensure that the most relevant data appears at the top of our results," Dauginas said. Older data was either not indexed at all, or it was given a lower accuracy rating.
Although it's been a month since the new search feature has been launched, there is still a lot of work to be done. In fact, Dauginas and his team remain just as focused on BellSouth's post-launch content issues as they were on pre-launch issues.
"Content is king with a search solution," he said. "[In] any kind of large enterprise, you're creating and updating data daily – if not hourly. So, since launch we're continually refining our results to make it a better experience for our customers."
Dauginas said it was still too early tell if the project is a complete success, but early positive customer feedback indicates that BellSouth is making good progress.