Web Analytics: Don't Try This at Home
Many companies still fail to have a fundamental understanding that a term like "hits" really doesn't describe anything. Yet some managers are so sure they know analytics, they will make potentially damaging changes in their business or website.
The key principle is to know exactly what you are measuring.
Then, deliver results in a way that is relevant, easy to understand and most
important, actionable. This requires an analyst experienced in behavioral analysis,
web analytics, and business process.
Tougher economic times have some companies using their analyst as a guide to better respond to customers and leverage the competencies within the company. These analysts are experienced in separating "correlations of convenience" from "correlations of confidence." For example, the bars may open at dark 100 percent of the time, but darkness does not cause them to open.
More repeat visitors who spend more time on your web site is
not necessarily a good thing. It could be (still) loyal customers getting really
upset with you, but looking for what should be on your site. You need
significantly more data to prove or disprove, if possible, your theory.
Think of a web analyst of being an "operator" in the movie The Matrix. They view screens of what appears to be streams of seemingly meaningless characters. A good analyst "sees" the world their users live in and can spot changes in the environment. How management does or does not welcome analyst recommendations also could mean success or failure.
These companies may be prioritizing work not by facts, but rather by the whim of the HIPPO (Highest Paid Person in the Office).
There are ample stories of analysts asking for navigation
changes that would have generated hundreds of thousands of dollars or more in
sales. The requests get ignored for lower value work. Products can get
discontinued because of lack of sales and layoffs may occur while a manager
tries to please the HIPPO
"Data scares people," says Chacon, "so the object is to make look easy to understand."
More companies are starting to react to daily or real-time
metrics. This can be dangerous without good guidance or multivariate testing.
Existing web metrics companies like Omniture
and Coremetrics as well as newer
companies like SiteSpect are making it
easier to do complex testing for specific goals.
Testing is different depending on the type of site you are running: sales, content, social media or a combination of the three. Each has a different success metric. Social media sites frequently try to measure the more complex metric of engagement. The problem is there is no one way to measure engagement. The industry group the Web Analytics Association has had a long discussion on the topic. They recently published a list of analytics terms defined for comment. Meanwhile, Eric T. Peterson posted an engagement proposal for public comment at the Web Analytics Demystified blog.
The bottom line: it will still be different for each site,
but this is a great place to start.
So don't try this at home and expect to maintain your sanity. You will have to determine the difference between lies, damn lies and statistics.
Wayne Kurtzman is a senior marketing analyst who loves the shiny toys of technology and online communities. He has led knowledge management and web analytics practices for startups and larger companies including Intel. Wayne also is active at the international level of Destination ImagiNation, a not-for-profit organization that fosters teamwork, innovation and creative problem solving skills in students from kindergarten through college.
