Magazine article Computers in Libraries

Meaningful Web Statistics? Now You're Hitting on Something! (the View from the Top Left Corner)

Magazine article Computers in Libraries

Meaningful Web Statistics? Now You're Hitting on Something! (the View from the Top Left Corner)

Article excerpt

A million hits a week, that's what he said. I was talking with a colleague about his Web site statistics. That surely sounded impressive, particularly since I didn't even know how many hits our own site experienced.

Way back in the Dark Ages of the 20th century, I had had a statistical analyzer running called Webtrends (http://www.webtrends.com). It was one of the first releases of the product, and as I remember, it measured about 6 million hits a year for us. It had lots of fancy graphs and took a day to load the log files and analyze them. That might have been the CPU; I'm not blaming the software, but the combination of factors made it awkward to use.

Then we made some changes to the Web site. The first big change was to place the Web pages on two different servers and have them balance the load. In theory, they would trade hits back and forth. If one went down, the other would pick up the slack. I love redundancy in library systems, but this was a little more theoretical than practical. It didn't really work very well. One of the results of the change was that the statistical package would no longer work.

That's because the program worked by analyzing log files kept by Microsoft Internet Information Server (115). These are kept in a subdirectory above system-32 in your WinNT directory, and you'd better watch out. They cumulate forever unless you manually clean them out. Just 1 year's worth of log files can easily be a gigabyte. Those files keep track of every single hit on your Web server. Webtrends goes in and analyzes the files. Unfortunately, our version of the program doesn't analyze the files on two servers. Instead, we need an "Enterprise" version that can sift log files together. But that costs thousands of dollars, so, for the moment at least, we dropped the idea of gathering statistics.

The next thing we did was add yet another Web server and form a "cluster" for all three servers. This is really slick. It achieves our goal of redundancy with three servers pushing the same data. Further, if we make a change on one server, it automatically propagates to the other two. In fact, it does it within seconds. Suddenly editing became so much easier. Finally, we had a functional Web server that wouldn't easily go down. Wonderful. That's one less stress point for me.

In the meantime, from several quarters I was being asked for "statistics"--statistics I didn't have. Public relations wanted to use them for brochures. Administration wanted them to present to the board. Other groups wanted to know how successful their own pages were. How many hits did they have? I couldn't tell them. And, no, I wasn't going to put one of those ugly hit counters on individual pages.

What Exactly Does a Million Hits Mean, Anyway?

I found myself trying to slow all this down. It was kind of strange. After having been known as a technophile for a generation, I found myself increasingly in the position of pulling back on the reins and suggesting that people slow down in their rush toward new technology. I found myself cautioning people in their use of statistics. To say we have a million hits over "x" time period sounds pretty impressive, but what does it mean?

"What will you do with these statistics?" I would say. "What does it mean to say we have a million hits? How is that useful to you?" Well, sure. It's a million and a million is, like, a lot! So those people in Japan making several million yen a year are making a lot of money, right?

No one wants to hear that, of course. It became obvious that I would be doing something useful with Web statistics, and yesterday would be just fine, thanks. I came across an interesting and fairly inexpensive program called "LiveStats" by Deep Metrix Software (http://www.deepmetrix.com, where you can see a live demo). Capable of working in a clustered environment (thus cheaply solving my worst problem), LiveStats reads and consolidates all the log files and then produces statistics on the fly. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.