How a Nuclear Waste Company Cleaned Up Its Enterprise Search Problem

EnergySolutions, a leader among companies that process nuclear waste, seemed a prime candidate for new enterprise search technology. After all, the company, which reported $502 million in revenues for the first quarter of 2008, had formed from the combining of nine companies in two years, and each company had different document management systems and file types.

RELATED LINKS

Study: Companies Need Better Enterprise Search

Finding Value in Your Enterprise Search Options

The Future of Enterprise Search

The files contained some of the company’s most precious intellectual property, including engineering papers, drawings and safety documents. And the amount of storage these documents encompassed was astonishing, says EnergySolutions CIO, Carol Fineagan. “It’s not a technical term, but it was a bazillion terabytes,” she says.

Picking one document management systemand ditching the others wasn’t an option, Fineagan says. “The structure of these document management systems meant a lot to the engineers who developed the data on each of them,” she says. “They know what [electronic] filing cabinets to go to when they want to find something. Plus I’m almost 49, and I would not live long enough to convert all the data out of their native document systems over to one, whichever the winner would be.”

Two years ago, Fineagan began looking at the enterprise search vendor landscape. Her search was two-pronged: she wanted a search tool to work on users’ individual desktops, but another to serve the enterprise needs of culling data from the various document management systems.

For the enterprise-wide search, she looked at what she described as the major vendors at the time: Fast (now a Microsoft subsidiary), Endeca, Google and its search appliance, and Autonomy. She settled on Autonomy. According to Fineagan, Automony was best suited to recognize and pull information from the company’s plethora of document types.

“JPEGs, e-mails, MS Office documents, video—you name it and we had it,” Fineagan says. “Some of this stuff went back to the early 1980’s. We needed something that could make sense of older file types, manage the indexes.”

When the Expensive Choice Makes Sense

According to a report by Forrester Research in May, Autonomy is a top choice in its category for large enterprises, largely based on the fact it can read old documents and file types.

“It’s not that others (such as Google and its search appliance) can’t hook into well-known places,” says Leslie Owens, the Forrester analyst who authored the report. “But large enterprises with complex needs and certain file types usually go with a higher end search product [like Autonomy] because of the scalability.”

What are the drawbacks of Autonomy?

“It’s more expensive,” Owens says. “They have a vague and complicated pricing matrix. These higher end search products look at how many servers you have, users, and how many connectors you may need.”

Google’s search appliance, meanwhile, publishes its pricing on its website, a move Owens says enterprise customers have liked.

But Fineagan settled on Autonomy for its scalability, security and intuitiveness, she says. According to Fineagan, Autonomy is sensitive to the fact that people use different search terms to look for the same information. As an example, there were subtle differences in the English search terms that Fineagan’s users in the U.K. (Queen’s English) and the U.S. (colonial English) would use to search for the same document.

“With as much data as we have, we can’t rely upon on someone’s head for search,” Fineagan says. “We need the tool to be smart enough to understand what the person said and apply it in a broader base and bring back different results.”

For desktop search, Fineagan says that she looked at options from Google and Yahoo!, but settled on Autonomy here too, because she believed it was important to give users the same intuitive features that Autonomy’s enterprise search had.

“Google and Yahoo is what people used at home and were familiar with,” Fineagan says. “The problem with those tools is it’s a lot like a keyword type search without the intelligence. They turn back way too many results to be useful.”

Making Your Deployment Work

For the enterprise search implementation, Fineagan says it was important to determine what servers she wanted culled and to make sure that the data on those servers was reasonably organized.

This is not to say they have to be perfect, she says, but it will help the search indexing process. “It just speeds up the indexing and from a bandwidth standpoint you don’t want your indexing tool and your pooling of documents to go hither and yon,” she says.

After that, she says, you point the Autonomy IDOL server in the proper direction (towards databases) and it begins indexing.

Meanwhile, she had a few administrators who specialized in databases such as SQL and Oracle trained by Autonomy on how to monitor the system. All told, she says, the implementation, which started in November, 2007, took a mere 30 days, a timeframe Fineagan describes as “phenomenal.”

One factor that has set the Autonomy search apart from the crowd for Fineagan is security. Whatever security exists on the application layer, she says, Autonomy acknowledges it. For instance, if someone does not have access to salary data (and most people don’t,) Autonomy will not return any results because it reads the security prefixes from the HR database.

“It respects the security we’ve built in,” Fineagan says. “We didn’t want to have to have duplicate security layers.”