Revamping your search capabilities means more than buying the latest tool from a search engine vendor.
In fact, it requires carefully matching the search engine to your needs and mapping out your data sources.
"A technology doesn't solve all problems," said Kenny Bunnell, program manager of technical services at Waltham, Mass.-based Novell Inc.
Two years ago, Novell rolled out search engine technology from Cupertino, Calif.-based Kanisa Inc. The networking software vendor deployed Kanisa's technology both for internal customer service engineers and customers who go online to answer questions. Because it had information spread across myriad sites and forms, Novell needed to make all .pdf files, product documentation, white papers and research notes available from a single search engine.
"It took a little more pain than we anticipated, but that was more on the business side," Bunnell said.
Looking for a technology solution to a business optimization problem is a recipe for failure, said Tim Hickernell, vice president of technology research services with Stamford, Conn.-based Meta Group. In fact, businesses often give project ownership of a search engine implementation to the wrong people.
This is especially true of companies using the latest technology.
Search software vendors are now allowing customization that enables managers to do things like promote certain Web pages ahead of others based on priorities such as overstock, marketing campaigns or frequent service problems. That has left organizations in a dilemma over who is making the search decisions.
"I do see some projects where only one area, like marketing or customer service, makes [search engine] purchases for content," Hickernell said. "But unless that's the case, it's usually IT that's left holding the bag making decisions and determining the business needs for everyone else."
Better to take a phased approach and determine the business unit with the greatest need for the technology, Hickernell said. Businesses can then add features as needed rather than redesign a site and relaunch a search engine across it.
Getting search right also involves a time commitment. Since Novell got its system up and running, it hasn't required more than an occasional tuning, Bunnell said. The hardest part was identifying which data to use and where it resides.
"Our biggest challenge has been consuming data from all repository types and pulling them into a contribution engine," Bunnell said. "You really have to have a clear definition as a company what the purpose of those data stores are."
Where service reps once had to run multiple searches across multiple repositories at Novell, they can now run one search. Novell has cut in half the time spent helping the customer, Bunnell said. The results are similar on the customer-facing side and are improving monthly, as Novell fine-tunes its knowledge map.
Despite improved self-service, don't expect a search implementation to suddenly stop calls to the contact center, though.
"The trend for call volume is up," said Mark Angel, founder and chief technology officer of Kanisa. "As an enterprise, you can accept that or you can do things to mitigate it. If you expect some sort of magical thing to happen where you can shut your 800 number off, that's fantasy land."
Search also requires vigilance. Search maintenance is an ongoing requirement for all organizations, Hickernell warned.
"To do search right requires people, people dedicated on an ongoing basis," he said. "And not IT people. This is not a 'keep the server running' issue. It's a business process issue to use reporting and analytics."
With some exceptions for external-facing processes, monitoring search is not a full-time job in most organizations, Hickernell said. Maintaining search still falls in the lap of IT, partly because the vendors have not made interfaces for the business user, he said.
Optimized search can also alert an organization to potential product problems faster, Bunnell said. Customer complaints and queries can point to issues like design flaws early.
"The real gold mine is knowledge management to really understand what's going wrong with your product," Kanisa's Angel said. "The problem is it's hard to mine because it's all in natural language."