MAY 15, 2006
Executive Guide to Search
By Marc Strohlein
Enterprise Search Gets Lost
The model for standard Web searches just doesn't fit many corporate needs. Instead, vendors need to think modular and customized
What could be simpler? Type in one or two words into a Web search box, hit enter, and voilà, you receive screen after screen of mostly relevant links to documents and Web sites.
Through this elegant simplicity, Google (GOOG) has transformed the way many of us live and work. One would be tempted to think that a similar approach could be in used an enterprise with similar results, but that assumption turns out to be surprisingly faulty. In fact, despite continuing advances in search technology, Outsell's research shows that the amount of time that employees who work with information such as market research, financial reports, and technical documentation spend on searching for what they need is actually on the rise.
SEARCH TIME UP. Surveys conducted at the end of 2004 and 2005 show notable increases in overall information-related task time for workers in academic, corporate, government, and health-care enterprises. Across all enterprises, the average time spent increased 1.1 hours per week per employee, growing from 10.9 hours to 12.0 hours.
Worse yet, the overall portion of their task time spent on information gathering increased on average by almost 1 hour per week, making up the bulk of the overall increase in task time.
Many very good enterprise search engines are available on the market, so why are knowledge workers spending ever more time looking for information? The simple answer is that enterprises are different from the open Web and present a more difficult environment for search engines. Challenges include:
• Enterprise information includes not only the documents, images, and HTML of the Web but also complex data from enterprise applications such as Enterprise Resource Planning or Customer Relationship Management (CRM) systems. Search engines are only beginning to really grapple with the latter category of information.
• Information in most enterprises is scattered among numerous servers, mainframes, and most challenging of all, the desktop and laptop systems used by employees. Search engines can't find that which they don't know exists.
• Much information in enterprises is proprietary in nature, with access restricted to only those employees who need to see it. Enterprise search engines need to obey security strictures.
• Enterprise searches tend to be high-stakes: Missing a site in a search for information about an actor or musician is far less worrisome than missing a document pertaining to clinical trials of a pharmaceutical drug.
• Knowledge workers employ a variety of applications to do their jobs, and if information isn't returned in the appropriate format for a given task they have to go through gyrations to manipulate or format that information.
• Link-popularity ranking algorithms, such as those used by Google, don't work as well behind corporate firewalls because the population of links is relatively small.
COMPLICATED USAGE. Even Google, the king of open Web search, has an enterprise search engine that's only beginning to address the issue of data in enterprise applications and outright flunks several of the challenges noted above.
Another big reason that enterprise search is so difficult is that search in and of itself isn't enough. Knowledge workers discover and use information in increasingly sophisticated and complex ways, and enterprise search products currently on the market don't meet all of those needs, contributing to the rise in time spent looking for information.
Moreover, search-engine buyers and deployers often don't know their users' information-seeking behaviors or their information needs, so they aren't able to develop truly effective information discovery and retrieval solutions. Dropping a search engine or appliance into an enterprise without understanding and targeting unique needs is bound to lead to unsatisfactory results.
WHY IT'S NOT WORKING. Finally, the activity called "search" is really a subset of a family of information-seeking behaviors. Search isn't a singular activity aimed at finding information. For most people, searching is part of a continuum of activities including browsing, searching, analyzing, and categorizing, that result in task completion, e.g., finding an answer, making a decision, or learning something.
Consequently, focusing solely on "search" is counterproductive, since most workers aren't interested in searching: They want to find information. While this may sound trite, focusing on search to the exclusion of user behaviors and needs leads to some decisions and behaviors that may run counter to helping workers find information
Key reasons why pure "search" approaches haven't been totally effective include:
• Many workers prefer browsing, akin to browsing library stacks or bookstore shelves.
• Search engines generally don't do a very good job at facilitating serendipitous information discovery for different, but related, topics.
• Most search engines require a level of skill and focus that many users don't have.
• Even the best search engines will return poor results when confronted with a jumble of disorganized and poorly structured content. Many of the companies that Outsell talks to aren't satisfied with the state of their enterprise search, but most also admit that they haven't put enough effort into organizing their content.
GOING MODULAR. Enterprise search needs to be thought of as supporting one of many different information-seeking behaviors and integrating a variety of external and internal information sources. Moreover, enterprise search needs to evolve from being a "go-to" function to being a set of enabling technologies that are integrated into various tasks and computing work environments. While many of these components are available from a variety of vendors, it can require a lot of "technology sewing" to stitch them together.
What's needed is a modular, service-oriented approach to search technology that enables the creation of search technologies that are configurable to the varied needs of enterprises and users. Fast Search & Transfer and Endeca are examples of two vendors that take this sort of modular approach to search, offering a range of information handling and analysis capabilities on a "pick and choose" basis.
In fairness to search vendors, enterprise search isn't broken, it just hasn't evolved to where it needs to be. This evolution isn't just a product/vendor phenomenon, it's also endemic to enterprise buyers. Most don't have sufficiently established information architectures to support the deployment of highly sophisticated technologies to aid in information discovery.
Discovery and retrieval of information through search and other mechanisms is a key component of modern enterprises. While many organizations blithely state that information is the "lifeblood" of their businesses, most haven't invested the time or energy needed to use that lifeblood effectively.
[an error occurred while processing this directive]
