Web analytics, mining visitor data for trends, has been around for about fifteen years. We web workers have attempted to find good ways to convert the numbers to something that resembles user engagement, with varying success. Now that videos are a major part of the web, new ways of measuring streams are emerging. The nascency of video analytics serves as a reminder that we still don’t have standard web stats completely figured out.
In the late 90s, “hits” became a popular term, despite not being very useful from an analytics standpoint. Hits refers to the number of requests a server receives, including images, CSS files, and more. The term does little to help us compare sites to each other, because the number of requests can vary wildly.
Technical publisher O’Reilly asks, are streams the new hits? Different sites have different definitions of a stream. For some, it means an entire video. For others, it’s a segment of video. That makes comparing a site’s streams about as useful as comparing its hits.
We’ve known that hits are a bad metric for some time, but what about other ways to analyze web traffic?
Pageviews are probably the most popular means to evaluate traffic, as well as dole out ad dollars. Much as video can be split into multiple streams, text articles are often separated into several pages. So, our beloved pageviews aren’t necessarily the best gauge of user engagement.
Visits is another common web metric, but it doesn’t take into account what users do while on the site. Viewing one page, immediately closing the browser and spending an hour clicking through archives are counted equally. Time per visit and Pageviews per visit attempt to rectify that problem, though each could also signs of user confusion.
Bounce rate has recently become a popular sign that user’s lack engagement. If a user views only a single page during a visit, they’ve “bounced.” For some types of sites, such as those selling products, this might be a good metric. Information sites might be unfairly docked if a user quickly checks the page for the latest news and then closes the window. Following a link to an external site from the first page counts as a bounce. That means that if you search Google from a toolbar and find a result on the first page you just bounced.
Any metric probably has its downsides. The same will be true of video analytics, as we search for the right data. Recently Google examined brain waves to find new ways to evaluate user response to overlay ads in YouTube videos. Even though click rates are likely abysmal, the company is searching for proof that the ads are effective.
I don’t know the details behind the methods that Google calls “more technologically sensitive.” My guess is these video metrics, like the examples above, also have their problems. For example, if the brain reacts to an ad, it could just as likely be caused by the user wondering what showed up as it could by the user being engaged by the ad.
Just because metrics are imperfect doesn’t mean we shouldn’t use them at all. Both striving for better measures and being aware of the downsides of current ones are important. We’ll continue using less-than-perfect numbers because they’re better than nothing and, in some cases, all we have.
What metrics do you use?