How to Evaluate Enterprise Search Options
Google and other Internet search engines virtually never fail to deliver relevant results nearly instantly. That creates a problem for IT in terms of setting employee expectations around the search capabilities they use at work.
"They think an awesome search engine is a straightforward, must-have tool, and they wonder why the company doesn't have one," says Leslie Owens, a senior Forrester Research analyst. As companies seek to address that issue, they enter the world of enterprise search, where they'll find more than a dozen products available. Choosing the one that will work for your enterprise involves evaluating the types of products, coming up with a requirements list and performing a proof of concept test, among other tasks. To be sure, it's a challenging task. "Users' needs can be unique," Owens say, "and finding one system that serves a diversity of queries and users can be tough."
Tough but worth it, says John Gillies, director of practice support for Cassels Brock & Blackwell LLP, a Toronto law firm with more than 200 attorneys. The firm recently replaced its existing enterprise search platform with Recommind's Decisiv enterprise search product. Having a single search engine that integrates results across multiple repositories-also known as "federated search"-enables you to create reports "with a much greater informational value," Gillies says. "Four separate reports from four different repositories just don't have the same impact."
Enterprise search can also be a powerful tool for boosting productivity, Owens says. "Knowledge worker efficiency is looming as the next great competitive differentiator," she notes, which is why many organizations are investing heavily in search, social networking and collaboration/communication tools to speed the flow of information. A Q4 2010 Forrester Research survey of Information and knowledge management (I&KM) professionals in North America and Europe found that 47 percent are implementing or planning to implement information access software, such as enterprise search tools.
Look at What You Have
Before diving in to an enterprise search project, you first need to determine if such a product is even necessary. Some CIOs believe desktop search tools and the search capabilities inherent in the organization's information repositories, such as email and content management systems (CMS), are sufficient, Owens notes.
Make an inventory of where crucial content lives and which vendors you're already using to search that content, Owens advises. "Think about what you need to search across your various information repositories, and if you can stretch the native search of your repositories in any way," she suggests.
Get to Know the Players
If you decide to forge ahread, you'll need to determine which type of product is the best fit. In general, enterprise search products fall into three categories, Owens says:
Specialized search vendors address specific user information needs (such as customer service) or industries. Vendors include Attivio, Coveo, Endeca, Exalead, Sinequa and Vivisimo.
Integrated search vendors such as Autonomy, IBM and Microsoft merge robust search capabilities with other information management functions, such as web content management. They also sell search technology independently.
Detached search vendors, including Google, ISYS and Fabasoft, focus on ease of deployment and flexibility.
Google, Autonomy, and Microsoft dominate enterprise search but other players have capable products worth examining. Notes Owens: "Coveo and Vivisimo specialize in customer service; Attivio, Exalead, and Endeca in custom applications to merge structured and unstructured information; Sinequa and IBM in semantics; ISYS in OEM; and Fabasoft in eGovernment."
Detail Your User Requirements
The next step is to start a wish list itemizing everything your users want from a search engine, says Gillies. Group the items under related topics and prioritize them. Cassels Brock & Blackwell came up with five categories: "Essential," "Very Important," "Important", "Nice to have" and "Useful but not critical." The firm also drew up a list of the top 10 essential items, which "proved very useful in doing a focused comparison between the two search engines we compared."
Gillies says the four key aspects of an enterprise search engine to evaluate are relevance, responsiveness, consistency of results and proper working of key functions.
In Forrester's Sept. 2011 evaluation of 12 enterprise search vendors, the research firm considered 10 criteria, including the following:
- Secure mobile support.
- Social and collaborative features, such as support for social tags, ratings, and recommendations.
- Security, including integration across multiple directories for authentication and the flexibility to map and modify security.
- Interface flexibility, including WSIWYG tools for customizing the interface.
- Relevance model, such as the ability to tune and bias search results based on user behavior.
- Platform readiness, including support for language-specific APIs, an SDK for custom development and support for different operating systems.
- Gillies recommends additional factors to consider when weighing enterprise search products:
- If you use other applications from the vendor, how responsive have they been to issues you've raised?
- What's on the vendor's development roadmap?
- How well will the search technology integrate with your existing applications?
- What repositories will you be indexing? How will the search technology integrate your different repositories?
- What will be the internal support requirements?
Perform a Proof of Concept
Once you've decided on a winner, it's time to perform a proof of concept (PoC). "The goal is to determine whether the search engine not only does what the vendor has promised, but whether it also does what you need it to do, in the way you need it done, and does so properly in your technical environment," Gillies says.
Typically, you'll load the search software in a test environment, index a small percentage of the documents it will search, and run initial tests to see how it responds, he explains.
Include "a representative mix of document types, sizes, and security profiles" for the search engine to index, Gillies says. Doing so provides the raw data you'll use for the rest of your PoC testing as well as for pilot testing.
After your data repositories have been crawled and indexed, set up the user interface and security modules. It's essential to confirm that the search engine respects the permission walls you've erected around sensitive information. "This should be straightforward to confirm for DMS documents, but if you're including other repositories, particularly accounting information, pay special attention to this issue," Gillies says. "Nothing will sink acceptance of your search engine faster than the discovery that users are suddenly able to access documents that should be hidden from them."
Create a List of 'Dirty Words'
Dealing with sensitive documents (such as confidential memos and performance reviews) is often among the most difficult challenges you'll face on the road to deploying an enterprise search engine, Gillies says.
Companies sometimes discover sensitive content filed in a publicly accessible part of the document management system, according to Gillies. The content was effectively hidden because previous search tools weren't strong enough to surface it. With a powerful, federated search tool, however, you may be helping users find more information than you want to.
Before rolling out the search platform, actively look for sensitive content and ensure it's firewalled from unauthorized users. "One way to find and secure this content is to draw up a list of 'dirty words,'" Gillies says.
Among the phrases he recommends searching for are: promotion, bonus decision, evaluation, resignation, termination letter, direct deposit, and performance review. "Check with your finance and HR departments to find out which terms they would search for," he adds. "Also, seek suggestions from your pilot group, since they may well come up with terms that your implementation team will not have thought of."
Getting More Like Google
By using backlinks, social media, content freshness, data compiled from billions of searches around the world, and hundreds of other signals, Google has continually raised the bar for search engines. Enterprise search engines don't have such luxuries, as they can't typically incorporate such signals as backlinks to deliver the most relevant results.
"It takes coding on the IT side to give someone in HR a different answer to the same query from someone who is an insurance agent in the field," she says.
Nonetheless, over the next few years, enterprise search platforms will strive to become more Google-like by delivering highly personalized, more contextually relevant results. "This is what's really going to be exciting about enterprise search in the future," Owens says.