What's New in Cybertalk?

by Jean Gora
March 2001

Note: CyberTalk is a column that appears monthly in LOMA's Resource, the magazine for insurance and financial services management. To see more contents of the magazine and to see how to subscribe, click on RESOURCE MAGAZINE.

Search Engines

Like everything else on the Internet, the world of search engines is the scene of a tremendous amount of innovation and competition. As anyone who has ever done an Internet search knows, the problem is not to find information; the search engines find more than a single human being could assimilate in a lifetime. Rather the problem is finding relevant information. For insurance company employees, the problem is finding information relevant to their jobs.

This article explores one small aspect of this problem—whether it is possible for such an employee to find useful, targeted information about underwriting preferred risks easily by running Internet searches. Assume that this employee simply types in the phrase "underwriting preferred risks" and does not enter quotation marks or any other Boolean operators. To find the solution, CyberTalk visited 34 leading Internet search sites and ran the same search through each of them. The table in this column shows the first five search results obtained from each site. (Click here to see the table).

Key Conclusions
There are several key conclusions from this query.

It is far from easy for this employee to find relevant, targeted information on underwriting preferred risks. A motivated employee with time to devote to searching and in-depth examinations of Web sites can find it. An employee who is pressed for time will not be able to find it.

Insurance agents and brokers run the majority of the insurance sites identified in the searches. Additionally, the information presented on their sites is targeted at either a prospect for insurance or another agent or broker. It is not targeted at insurance company employees who want to learn more about underwriting preferred risks.

It may take too much work. Even when the search results identify insurance sites that have some information about underwriting preferred risks, they often list links to pages that do not mention the topic and give no indication of where on the site information about the topic can be found. In other words, the information may be available on the site, but the employee would really have to work to find it. Usually the search engine delivers links to the welcome page of Web sites. Many welcome pages are low on content. In the case of sites run by insurance agents or brokers, there is one exception to this generalization. Links to the Americaquote Insurance Agency go directly to a page discussing preferred risks.

Some search sites deliver ridiculous results. For example, the fourth result delivered through the Altavista site is a link to the Turkestan section of the Central Asia discussion list. The first result delivered by Suite 101 is a link to a site on diversified learning; the site focuses on methods of teaching gifted students.

Some sites confuse queries with job searches. Many search sites appear to assume that a person who is looking for information on underwriting preferred risks is looking for an underwriting job. The search results deliver many links to insurance recruiters.

Many sites completely misinterpret the query. Even when some search sites correctly conclude that the searcher’s query concerns a financial topic, they misjudge the area of finance concerned because they appear to focus on one word in the phrase "underwriting preferred risks." For example, because the term "underwriting" can apply to investments, the third result from AOL Search is a link to a venture capital company. Altavista and Oingo both deliver links to articles about hedge funds. The term "risk" triggers a lot of links to various risks management issues that have nothing to do with underwriting preferred risks.

Pay-for-listing and algorithm juggling can result in bad matches. It is possible for site operators to structure their sites in ways that make them appear high on the list of results for searches on specific terms. Some pay the search engine for high placement. Others deduce the search engines’ algorithms and placement criteria and structure their sites accordingly. For example, some search engines consider the URL in placement. The URL of the Guide One Insurance Company is www.preferred-risk.com. Fidelity Investments, which runs the Insurance.com Web site, has successfully discovered how to be listed prominently by a number of the search engines reviewed here. It appears among the top five on the FindWhat, GoTo, Hotbot, iWon, Ixquick, LookSmart, NBCi, QBSearch, and Sprinks search results.

Some sites are better than others. Several search sites do better than the others at delivering links to sites that have information on preferred risk underwriting that would be relevant to insurance company employees. Google and Yahoo (which uses the Google search engine) deliver links to a SwissRe study on preferred risks in life insurance; the study concerns European data, but the discussion would be useful to all practitioners. They also deliver links to material on preferred risks by the Canadian Society of Actuaries and to material on the subject posted on the site of American United Life Insurance. The Northern Light site delivers a link to a Best’s Review article on underwriting preferred risks; however, you have to pay a fee to access the article. It also delivers a link to an insurance dictionary offering a definition of preferred risks.

PDF searching can result in better matches. One reason Google can successfully find relevant insurance content so effectively is that it has the ability to search PDF files. Many sites offering serious business content publish long documents on the Web in PDF format.

Natural language query can help. The sites such as Google and Northern Light, which deliver relevant information, clearly have some ability to handle natural language queries because they interpret the entire phrase "underwriting preferred risks" and not the individual words in the phrase by themselves.

Experts can be good if they know about insurance. Some of the search sites that claim to rely on human experts clearly do not yet have insurance experts. One exception is Ask.me, which delivered an accurate result (see the table) although one targeted at a prospective buyer of insurance.

In sum, an insurance company employee who wants to find out about preferred risk underwriting had better chose his search engine carefully. Or better yet, he needs to know the URLs of insurance sites that offer relevant information.

Some Background on Search Engines
Search engines are different from search sites, but it is easy to confuse the two. The search engine actually performs the search. The search site incorporates a search engine that allows one to perform searches. Thus, many well-known search sites use search engines developed by others. Thus, for example, the HotBot search site uses the Lycos search engine. The table below distinguishes between the two where that information is readily available. Because more than one search site uses the same search engine or the same back-end search service, many search sites deliver the same results.

The search engine market is evolving very rapidly as the developers of search engines attempt to add features that deliver more relevant results than are available through key word searches. At least four different Internet sites track developments in the search engine market. They are Search Engine Watch (www.searchenginewatch.com), Search Engine Guide (www.searchengineguide.com), Search Engine Showdown (www.searchengineshowdown.com), and Traffick – the Portal Portal (www.traffick.com). They provide much more useful information that can be provided in a short article. Those in the market for a search engine or those hoping to build traffic on their Web sites by getting better treatment from search engines will find these sites enormously useful.

Andrew Goodman, who writes for the Traffick site mentioned above, groups search engines in six categories on the basis of how they deliver relevant information.

Popularity engines show the most frequently requested results first. They assume that if people click through to a site frequently and remain on that site for a long time, they are finding that site relevant. This approach has several obvious flaws in that it may create self-fulfilling prophecies; a link that shows up first may attract visitors, whose visits cause the site to be ranked first. The fact that some one remains at a site long does not necessarily mean that person is finding what he wants. He may be spending time looking.

Meta search engines search multiple other search engines and generate site lists on the basis on an aggregate key word relevance score. Some consolidate the results; some do not. Thus, if the underlying search engines deliver useless information, so with the meta search engines. Using a meta search engine is certainly faster that going from one search engine to another in sequence.

Meaning-based search engines use their own lexicons and allow users to specify the intended meaning of the words. They may also analyze the relationships among words in a document.

Natural-language search engines allow the user to pose questions the way he would if he were talking to another person. It performs searches based on the phrases incorporated into the questions.

Expert guide sites use people to classify information. Some even do it in real time. If an expert is available in one’s area of interest, the information may be relevant.

Pay-per-click search engines allow marketers to pay for their search engine rankings. The end result may or may not be relevant information from the user’s point of view.