You are wasting time. Find out why

The cost of ineffective search

A company that employs 1,000 information workers can expect more than $5 million in annual salary costs to go down the drain because of the time wasted looking for information and not finding it, IDC research found last year.

Think that’s bad? A survey this month of 1,000 middle managers found that more than half of the information they find during searches is useless.

There seem to be no shortage of enterprise search applications that help companies find information hidden within their networks. So why are searches so ineffective?

It turns out, analysts say, that most enterprises are not using the most up-to-date search applications. Not only that, enterprises aren’t using the applications they have as effectively as they should.

“They’ve never invested a whole lot in it,” says Matthew Brown, a senior analyst at Forrester Research. “Companies will spend lots and lots of money on architecting portal systems, intranets, dashboards and databases and everything else. Search, typically, for internal applications, companies don’t spend a lot of time on it.”

As much as 10% of a company’s salary costs are wasted on ineffective searches, the Butler Group, an IT research and analysis organization reported last October. Richard Edwards, who co-authored the 240-page report, says a lack of metadata is one of the key problems.

Suppose you create a Microsoft Word document. If the program is set up to index metadata, you will be prompted to fill in fields recording information like author, title, subject matter or the expiration date of the information contained within the document, Edwards says. These metadata fields are like “outer markings” that make it easier for search engines to determine whether a document should be returned on a hit list, and reduce their dependency on full text searches.

A decade ago, when enterprise search programs were less widely used, it was “horrendously difficult” to get employees to enter this kind of information, according to Edwards. And it remains hard today, he says, even though the widespread use of enterprise search provides a clear incentive to enter metadata.

“Ninety percent of the documents that are created have no useful metadata,” Edwards says. “Until we get more of that metadata it is going to be an uphill struggle to get better results out of these very capable search technologies.”

Some high-end enterprise search applications, such as Autonomy, do a “modest job” of determining what a document is about on its own, he says.

“They do more than just pick out the main words and index them. They can look at parts of the document, the document title and headings and work out what the document is about, and use that in the future to help you retrieve information that is of a similar nature,” Edwards says.

Susan Feldman, vice president for content technologies at IDC, has been studying the cost of ineffective enterprise search for five years. Her latest research paper, released in April, found that information workers waste 3.5 hours each week on searches that don’t turn up the right information. Assuming an average salary of $60,000, including benefits – a figure based on 2004 Bureau of Labor Statistics data – the cost of ineffective search is $5,251 per worker per year. For 1,000 workers, the cost is $5.25 million.

“People spend about 9 to 10 hours a week, on average, looking for information. Of that time, they don’t find the information they’re looking for a third to half the time,” Feldman says.

Those figures apply only to the use of enterprise search applications, rather than Web searches such as those on Google, which are less effective, she says.

Sometimes, searches fail because the answer someone is looking for simply isn’t there. Or, the right document might exist, but it hasn’t been “spidered,” the term Feldman uses for applications that crawl a repository and create an index of words and documents. If, for example, the server for the CRM department hasn’t been spidered, lots of useful information could be excluded from searches, she says.

When setting up an enterprise search application, company managers need to figure out exactly what they need the search tool for.

“Any of the major enterprise search vendors have service groups that do exactly this for you,” Feldman says. “They’ll sit you down and say ‘why do you want a search engine? Why do you want to find your documents? Are you going to need it for customer service? For e-discovery?’”

New federal rules requiring companies to maintain electronic documents potentially needed in litigation is one factor driving companies to upgrade search engines. The best enterprise search applications widely available today use concept searches, which look for documents and files tied to specific concepts, she says.

If you did a keyword search on Google for “high blood pressure,” you might not get useful information about hypertension, the technical term for having excessive blood pressure, she notes. But an application using concept search would know that high blood pressure relates to hypertension, and turn up more useful information.

Searching based on concepts is “the generation of search that is just being adopted now.” Most enterprises have not yet upgraded to enterprise search platforms that use this technique, Feldman says.

Modern search applications also detect trends, pulling out information like stock prices and presenting it in a chart format, she says.

Feldman and some other analysts are optimistic that semantic technology will fuel the next generation of searches. The word semantic “means meaning,” she says, so an application using semantic technology understands not just keywords but the relationships between subjects, verbs and modifiers.

“This means you can type in a question and it will understand it,” she says. “More and more applications are able to understand who, what, when, where and why questions, and differentiate among them.”

The so-called “Semantic Web” has been a hot topic of discussion in technology circles for several years. The Semantic Web has been defined as “an extension of the current Web in which information is given well-defined meaning, better enabling computers and people to work in cooperation.”

Building a Semantic Web requires providing a layer of meaning and understanding of content, which is a major challenge because the content on the Internet is owned by so many people, Edwards says.

It’s simpler to do that within an enterprise because the content generated by employees would be owned by the corporation, he notes.

A search using semantic technology will turn up the most-desired result about 80% of the time, while searches used by most companies today turn up the right answer less than half the time, according to Benjamin Grosof, assistant professor of IT at the MIT Sloan School of Management.

“Semantic technologies will advance the state of the art within the next few years by a significant amount,” Grosof says.

A hurdle today’s search engines have trouble overcoming is that information within a company is spread out over many databases and comes in many different formats, he says.

Still, Grosof notes that search technology has improved dramatically over the last decade, fueled by the Web and artificial intelligence.

The U.S. Army used artificial intelligence to create a virtual officer known as “SGT STAR,” who is capable of answering a range of questions posed by potential recruits visiting the Army Web site.

SGT STAR is about as sophisticated as it gets these days, Feldman says. The technology is sure to become more widespread in commercial use because it can replicate call centers, which are expensive to maintain, she notes.

Enterprise search applications today don’t answer questions as well as SGT STAR, but some can extract sentiment and opinion, she says. A company might ask its search engine how people feel about a new brand. The application will then search blogs and the media to figure out what people are saying, and represent their opinions in a graph.

Unfortunately, Feldman says many companies are missing opportunities by not staying up-to-date with the latest search technologies. While poor information leads to poor decisions, wrong information leads to wrong decisions, she says.

“Those wasted search hours,” she says, “add up to millions of dollars over the course of a year, even for just a midsize company.”

Learn more about this topic

IT managers waste a lot of time searching

We ask the U.S. Army's new Web tool about girls, guns, gays and invading Iran

1/04/07

New product helps companies find documents needed in litigation

1/03/07

Time spent searching cuts into company productivity

10/20/06

Join the discussion
Be the first to comment on this article. Our Commenting Policies