Water Cooler: Web log files aren't enough

But while such tools can provide valuable data, they're not enough. These days, when information is as likely to be stored in a database as on a static Web site, if you're not paying attention to how people are interacting with those databases, you could be asking for trouble.

It's a lesson we learned recently with our DocFinder database:

If you get the paper version of Network World, then chances are you've seen DocFinder numbers with many articles. Come to the Web site, type the number that goes with the article and presto, you get a copy of the article and links to additional resources. We originally developed the system back when we had a bizarro encrypted-URL system (back before cookies). But we kept it when we switched to bookmarkable URLs because some users told us they liked the convenience.

But what about people who don't read Network World in print? What would they do with the DocFinder box we put under the Search box (which hits against our Ultraseek search engine) on every page?

Ignore it, we hoped. More out of curiosity than anything else, I recently started trapping the numbers people were typing into the DocFinder box - which articles were most likely to generate requests?

Helpful hint: If people outside your immediate development team have access to your Web stats (the marketing department, say), write up a short FAQ or cheat sheet on what the stats mean. Otherwise, count on getting lots of messages asking why odd directories are getting lots of hits (because that's where you store all your nav-bar graphics) or even why the number of "page views" is so much lower than the number of "hits."

The log file was quite an eye opener, alas. Turns out that more than 40% of the "numbers" typed into the box were actually words -- people were mistaking it for our Ultraseek search engine.

Obviously, we were oblivious to a design flaw. Now that we know, though, we're going to fix it. We'll be replacing the two input boxes with a single "Find" box -- and a JavaScript that can tell if a user is typing in a four-digit DocFinder number or a search phrase and then route the query to the right database.

Speaking of Ultraseek, though, when's the last time you looked at the query log from your search engine? It can provide some clues to help you with your site design and content.

If you see a lot of searches on a particular keyword or phrase, could it be that people can't find how to get to that information easily from your home page or toolbar? Or maybe there's a trend happening -- something you offer is really hot and people want it right away. Or maybe they're looking for something you don't even have on your site -- yet.

So enjoy the holidays, then come back and dive into those database files.