IBM, which is more forthcoming about its algorithms, uses a blend of weighting factors for relevance in its enterprise search. These include: user click patterns, the format and position of an entry in a document (headings have higher relevance than in-text entries), metadata (so that text in a link will be ranked differently than similar text in the body of a document), and so on.
Most products today (see "Enterprise search vendors bet heavily on BI") provide a way of increasing relevance of certain documents or URLs so that they occupy first place in a given search. (For example, a query on "sexual harassment" can be tweaked so that the company's policy is always the first item returned.) In addition, many products enable customization for company-specific lingo. This permits search engines to know, for example, that a query regarding "Region 1" refers to the Eastern seaboard.
Proceed with security in mind
Access control is a central aspect of BI search. This problem occurs in two directions: how does an employee access all the needed data for a report and how is an employee blocked from seeing confidential data? In a perfect world, single-sign-on would address the first issue, and access to a directory LDAP server would resolve the second. The problem is in the implementation: much of the data is located on systems whose access control is not tightly neatly defined by a corporate-wide access mechanism.
The problem is actually worse than it appears. Says Maxime Tiran, an engineer in IBM's Data Management division, "When company IT departments set up enterprise-wide searching tools, they are frequently horrified by the kinds of confidential data that is widely accessible and completely unprotected on their intranets."
Security schemes vary, and sites contemplating adding search to their BI need to determine how access control is handled by the products they're considering. Many products simply pass the user credentials to the BI package or other back-end software and rely on those applications to limit the returned results according to their built-in access mechanisms. This aspect is a particular strength of Oracle's Secure Enterprise Search product.
On the horizon
Michael Corcoran, who heads up Corporate Strategy for Information Builders predicts that the integration of BI and search will only become tighter. Search engines will gain better access to BI data and the BI companies will facilitate this process. For example, Information Builders today can take data from transactions in process and make it available to Google's Enterprise search engine.
Information Builders has a division that provides some 300 connectors to data sources and it is actively using them to broaden the reach of search capabilities and its own BI products. Says Corcoran: "This greater integration will really help users. Today, BI still requires users to know where their data is. For example, they still must specify 'call center data.' However, the needed data could be anywhere, and the user should not need to know its origins to be able to locate it."
The next step, says IBM's Andrews, involves integrating analytics functions with search and being able to query the data in a variety of ways to probe for market opportunities that equate to increased sales and greater efficiencies. For the time being, however, most enterprises will be content just to have better access to the business intelligence they currently generate.