Search Examples - Page 26

Sites are an important unit of hierarchy on the Web, so clustering search hits by the site they belong to is a great way of making users see the forest for the trees of pages in their search results. In this example, I would probably have sorted www.world-ready.com above www.useit.com because it would seem to be more important to have 50 percent more pages about the query than to have whether the highest scoring page on the site scores 83 or 82.

Clustering also works for searches within a single website, although the structuring mechanism obviously needs to be different. For example, results could be structured by subsite or by category as done in this example from ZDNet. Also note the use of shortcut links to the most important areas of the site of interest to the user's query. I am less enamored with the advertising search box: Even though it is clearly marked as an ad, it intrudes too much on the user.

Note how pages that provide human-written descriptions make it much easier for the user to determine whether the page is of interest. Compare, for example, the first and second hits on this page: The first hit is represented by Infoseek's automatically generated summary of the page, and the second page is represented by a real abstract. Also note on this page how the search engine tries to identify a few alternative phrases that might be used for query reformulation. I searched on "usability," so the suggestion to search for "user interface" would most likely be helpful and would find pages that were missed by the original search. Obviously, the system's capability to find relevant alternatives is limited, as shown by its second suggestion, "Computer science institutes in Georgia." (Even though there is one such institute that does usability research, there are much better alternative queries.

See What People Search For

Several of the major Internet search engines have a service where you can view random samples of the queries entered by other users. It is quite interesting to spend a few moments to look at the ways people phrase queries and try to estimate what they might be trying to find. Searches provide first-hand insights into users' wants.

(The Webcrawler service I used to collect the statistics discussed in this chapter is at http://webcrawler.com/cgi-bin/SearchTicker.)

In addition to looking at people who search the open Internet, you should also study the search logs from your own site search. Any terms that occur frequently in your search logs obviously represent information that many users want to get but have trouble finding on your site.

The Search Results Page - Page 25
Designing Web Usability
Search Destination Design - Page 27