The theory of decision-making says that by knowing the possible actions and the probabilities of outcomes, optimal decisions can be calculated.
Of course, real life is rarely so accommodating, but it’s clear that fuller knowledge of actions and outcomes enables better decisions, which lead to more efficiency, productivity, and profits.
What are some removable obstacles to better information-based decision-making in the enterprise? I’ll use my experiences as academic researcher, software vendor, and search-engine user to offer some answers.
Obstacle: Users don’t know what information sources are available.
Enterprises subscribe to many information sources but users often don’t know about them. Even your public library card entitles you to many searchable electronic information services that you are probably unaware of.
Making users aware of these services is a real challenge.
A common approach is to create a Web page that lists all the sources or that lists the sources by categories, e.g., resources for marketing information. But even when the user is alerted to all the available sources, he or she is still burdened with memorizing their query syntax.
The better solution — advocated by Susan Feldman of IDC — is to create a federated (or meta) search of the available sources, perhaps grouping them by category of usage, for example, by tools for marketing professionals, for Linux software developers, or other interest group.
The result is one federated query can visit multiple sources, with syntax conversion as needed.
Obstacle: For a given need, users can identify a single right information source but are daunted by its user-unfriendliness.
For example, searches may return poor titles, only titles with no descriptors, static snippets, snippets that are contaminated by navigation links, or poor search-results ranking.
Intranet search managers often focus on relevancy ranking but ignore the quality of the description that is returned for each search result. With poor descriptions, users cannot easily judge whether to click on a result, leading to wasted time or missed opportunities.
Likewise, downstream text analysis tools, like clustering, are hobbled by poor descriptions.
Modern search engines can help solve these problems by, for example, configuring them to ignore navigation links and other irrelevancies during their indexing of the content.
For external-licensed Web-based information sources, customers of these providers must pressure them to modernize their search engines and interfaces to address these flaws, or to provide XML outputs which the enterprise can then overlay with its own more productive interfaces.
Obstacle: Users who must search multiple sources suffer from tedious multiple logins or from an inability to combine the separately returned search results.
Good implementations of federated search cure these ills because users can perform a single login to many sources, and then a single search query visits all the sources and combines their results.
Obstacle: Users who are overwhelmed by a glut of disorganized search results overlook significant information beyond the first few results.
Believe it or not, some installations try to solve information glut by rejecting queries that return too many search results. This forces users to formulate queries that return only a limited number.
The book world has a better solution: organize books into categories which lets users browse intelligently and find similar books in proximity.
Alternate Tech
Two alternative technologies, taxonomies and clustering, each aim to show search results in meaningful categories or folders so that users can:
In summary, enterprise users today make inferior business decisions because they overlook available information that is inconvenient to reach.
The reasons include the diversity of enterprise information sources with different syntax and logins, which are tedious to remember and access, and the long-list display of search results which lack the familiar organization found in the book world.
These obstacles to intelligent decision making can be removed by technology from a variety of software vendors, centered on the twin themes of “one query visits all” which characterizes federated/meta search, and “organized search results, not long lists,” which is the aim of both clustering and taxonomies.
Let’s recall the logician Alfred North Whitehead’s observation: “Civilization advances by extending the number of operations we can perform without thinking about them.”
Enterprises can apply this observation by relieving users of the excessive need to identify, memorize, type, scroll, discover, and collate when much of this can be done today by modern search and clustering technologies.
Raul Valdes-Perez is CEO and co-founder Vivisimo. Before starting Vivisimo, he was on the Carnegie Mellon computer science department faculty where he is currently an adjunct associate professor. Raul has published nearly 50 research articles and holds a Ph.D. in computer science from Carnegie Mellon.