Monday, August 01, 2005

The characteristics of a good Web site search

I'm writing this little post in hopes that someone who is charged with coding the search function for a Web site will read this and not cause me the undue pain that certain other sites have.

Starting with the basics, here's an example of a Web site that was coded by an IT monkey who didn't put much thought into how people might use it. Supply & Demand Chain Executive's search function, like their magazine, doesn't do very much for the average user. The search results get spit out in alphabetical order, by headline. And to find out the date of the article, you either have to click it, or hold your cursor over the link and look at the 4-digit ID number for the article, and deduce the date. Since I search this site for a few keywords every month, I have gotten pretty adept at judging the numbers, but it's still a pain in the neck.

And you would think that CNet news, of all news sites, would let a user put in a quoted string and churn up results on an exact phrase. Alas, no... if you put in the phrase "Sesame Street" you're going to get hits that have just one of the words.

Other sites, like the Pittsburgh Downtown Partnership's Web site, don't do fuzzy searches. This search is easily tricked by people who under- or over-punctuate words, when a little simple logic would return much better results that could account for someone not putting an apostrophe in the right place.

I won't pull out any more examples, but searches should have some way to disclude words, they should have AND and OR functionality, they should be sortable by date, headline, or key word relevance. I can't stand it when a Web site imposes its own arbitrary metric for evaluating the goodness of your results, and the measurement was designed by some half-baked intern with an attitude problem.

People who can't write good search functionality should just buy Google search for their site.

No comments: