Search Traffic and Web Analytics: Easy Answers, Hard Lessons

A popular speaker, pressed for time, asked her assistant for a quick statistical rundown of her audience for an upcoming presentation. She uses this data to help shape her speeches to appeal to the average attendee.

She asked for the average age, income, and educational level of audience members, which her assistant (who was also overworked) promptly provided.

The average age of the attendees was about 36. Average income was about $30,000, much less than she expected and much less than the typical audience for her speeches. The educational level was also lower expected, with the average attendee having slightly lower than a high school diploma.

So she worked diligently to retool her speech to accommodate this audience. When the day came, she was horrified to read the fine print: she would be speaking to a group of retired CEOs and their young grandchildren.

The retirees, many of whom had advanced degrees, were mostly living on investments and nest eggs. Most of their grandchildren were in the second or third grade.

The moral isn’t hard to grasp. Be wary of search traffic statistics, and be ready to dig deeper to get a more comprehensive picture. The time spent adjusting to data you believe to be true is sometimes better spent confirming or disproving your instincts. And nowhere is this more important than in Web analytics.

Be Wary of Averages

As the anecdote above suggests, don’t assume averages tell the whole story. Search traffic data is tricky; some data charts (e.g., search engine referrals vs. time of day) appear as a traditional bell curve, while others (such as frequency of referring keyword) have a large head and long tail. On some sites I’ve worked on, length of referring keyword phrase appears as a U-shaped valley, in others as a Bactrian camel hump.

In two of these three cases, you’d be foolish to assume the average value speaks for a majority of data points. In general, your time is well spent plotting some data before you assume you know how the data behave.

Will the Real Referrer Please Stand Up?

Because each Web analytics package measures traffic differently, some suites are good at discerning between the various services offered by major engines and some aren’t. Most packages can divide organic and PPC (define) traffic, but within the organic referrals you’ll sometimes see a mishmash of traffic from Google, Yahoo, and MSN services like these:

• Regular search
• Blog search
• Mail
• News search
• Groups

At a minimum, read through your analytics suite’s documentation to ensure search traffic numbers are assigned to the correct campaign.

Living and Dying by the Page View

It’s no one’s fault, but site owners are conditioned to rely on certain Web metrics as an indicator of search campaigns’ success or failure.

Ultimately, conversion tracking is the single most important metric in your analytics suite. But along the conversion path, we watch other numbers as well. Page views have long been one of the Web’s most reliable measurements. But because of technologies like AJAX (define), Flash, and RSS, a site can perform at engines better than ever and users can spend as much (or more) time on your site than ever before, but the page view count won’t reflect it. Page views rely on Web 1.0’s click-and-wait model. In last fall’s “Pageviews are Obsolete,” Evan Williams articulated it perfectly:

So what’s a better measurement? Good question. Like many good questions, the answer is “it depends.” If you’re talking about what’s important to pay attention to on your own site, you have to determine what your primary success criteria are and measure that as best you can. For some sites, that could be subscribers, or paying users, or revenue, or widgets deployed, or files uploaded, or what have you. It may even be pageviews.

Sites with an income model that relies on excellent search engine positioning and subsequent page views must be especially diligent in showing potential advertisers a true picture of the site’s user experience. Whether it’s shifting the influence of time spent on a site, adding script-based click tracking to internal AJAX applications, or something entirely different, a multifaceted approach to Web measurement is becoming more and more important for Web monetization.

Exit Points and Bounce Rates

Whether you choose to believe it or not, your site has a 100 percent bounce rate. Eventually, every user who visits — whether from organic search, PPC, or even a “Favorites” folder — will leave it. So when you view an exit page report and see large numbers next to various URLs, don’t despair right away. The numbers must add up to 100 percent!

Important questions to ask are:

• Which users leave certain pages, and why?
• Did they find what they needed?
• Did they complete a desired action?
• Did they leave in frustration, abandoned on some navigational or content-based peninsula?

If different types of referrers tend to send traffic that leaves the site at different points, contrast what the organic visitor sees with what the PPC and e-mail visitor see. Though most of us are getting used to plotting the conversion rates of organic, PPC, and other types of traffic, few people are accustomed to watching exit points by referral type.

If too many visitors leave from a page that’s only halfway through your ideal conversion funnel, it’s time to explore why that happened and change it. But if their exit page answered the question posed by their search query string, chances are they left satisfied. The remaining question is whether you’re satisfied with them leaving so soon. Depending on your revenue model, you may be completely satisfied.

Film director Billy Wilder had this to say about evaluating feedback: “An audience is never wrong. An individual member of it may be an imbecile, but a thousand imbeciles together in the dark — that is critical genius.”

I’m not advocating you consider your site’s users to be imbecilic. But Wilder’s observation has some application to Web site analysis. When the audience laughed, did everyone really think it was funny, or was it just two really loud guys at opposite ends of the auditorium?

Sometimes those things sound the same, but they’re not.

Want more search information? ClickZ SEM Archives contain all our search columns, organized by topic.