Intelligent Enterprise

"If your Web site sucks, it's your own fault." That's the tough love Avinash Kaushik shared today here at O'Reilly's Web 2.0 Conference in New York in a presentation entiled "Web Analytics 2.0: Rethinking Decision Making in a '2.0' World." Kaushik offered a bunch of great advice on how to better measure site performance and he also listed a handful of free tools. "So there can be no more excuses" like not having enough data, not having the right data or not having enough money for Web analytics, he concluded.

A Google executive and author of "Web Analytics: An Hour a Day," Kaushik said the biggest problem with conventional Web analytics tools is that they tell you the "what"(as in what pages are drawing traffic on your site), but they don't tell you why people visit. What's more, they don't tell you what else visitors would like to do/see on your site.

Kaushik declared path analysis as "a complete waste of time," and he said one of the most important metrics site owners should look at is bounce rates. "A bounce is a visitor's way of saying, 'I came, I puked, I left," Kaushik declared. "It's hard to get under 25 percent but if your bounce rate is up around 70 percent, it's really clear that your site sucks." To make things simpler, Kaushik suggested checking bounce rates for the top 20 pages on your site as a starting point.

All too often the folks who call themselves Web traffic "analysts" are actually "reporting monkeys," Kaushik complained, citing pages-long keyword lists and top-ten lists as examples of useless reports that are often ignored by users. You can't possibly do anything with massive data dumps, and top-ten lists are almost always show the same results.

"Go beyond the top ten and look below the surface at what's changing," he advised. "Look for places that are sending you traffic that have never sent you traffic before" and look for parts of the site that are gaining or loosing ground.

Ajax, Flash, streaming video and other Web 2.0 phenomenon are challenging conventional Web analytics tools that try to create what Kaushik called "fake page views." He welcomed the advent of event-stream analysis and event logging as the best way to track rich media. As an example, every auto manufacturer has a "car configurator" on its site, but it wasn't until Toyota turned to event logging (rather than gross page-view metrics) that it could mine the configurator for priceless data on what colors and options people want on which vehicles. That data can even be broken down by region and city to aid marketing and manufacturing planning.

Event analysis is also being used by movie studios and streaming video sites to see exactly what people are watching and for how long.

Lots of sites try to measure stickyness, but average number of visits per X is not meaningful, says Kaushik. Instead you should study visitor loyalty, particularly those who are returning 10, 20 or 30 times per month. Visitor recency (people who came back less than a day or two ago) is also a hot measure. These deeper loyalty measures give a better idea why people are visiting your site. For more explicit "voice-of-the-customer" feedback, try this three-question exit survey:

1. Why are you here? – Many won't cite the purpose you had in mind. Mine the feedback for fresh new ideas.
2. Where you able to complete your task? – Force them to answer "yes" or "no." The results will be revealing.
3. If you were not able to complete task, why not? – Make this an open-ended question and look for what Kaushik calls "pockets of discontent."

So what are those free tools that Kaushik recommends? Of course Google tools were frequently mentioned, including Google Analytics, Google Website Optimizer and Google Insights for Search, but he also suggested third-party tools including Compete.com and Hitwise.com (though the last is not free).

So there you have it. Free advice and free resources, but now you have no excuses!