Web content allows for a level of tracking and analysis unseen in other forms of media, but I get the sense some publishers are a little hazy when it comes to the established analytic measurements. This primer touches on the main measures I’ve used in my own efforts, but it is not exhaustive. I encourage other analytics folks to chime in with their thoughts and techniques in the comments area.
A few notes before we get into it:
- Note #1: If you check stats religiously and you’re a whiz with your analytics tools, this post will be elementary and quite dull. You’re better off perusing the excellent conversations at Webmaster World.
- Note #2: The term “hits” was outdated in 1999. I won’t be using it here and I implore you to avoid this word — and anyone using it within a Web traffic context — at all costs.
With that out of the way, let’s dive in …
Visits — When you access a specific Web site, that counts as one visit. If you leave and return, that usually counts as a second visit. I say “usually” because most analytics tools use a timer. For example: If you leave and return to a site running Google Analytics within 30 minutes, one visit is logged. But if you return after 30 minutes, a second visit is added to the tally.
- Caveat — “Visits” should not be equated with “people.” Even with a timer in place, it’s possible for a single person to rack up multiple visits to your site.
- Recommendation — Track visits over a period of months, not weeks. Long stretches will reveal the overall growth of your site and your audience.
Unique Visitors — Unique visitors represent individual visitors to your site (in theory). This is an important metric because it gives you a sense of your audience size.
- Caveat — Analytics tools rely on cookies to track unique visits, but cookies can be deleted or rejected by the user. There’s also no way to differentiate between people using the same Web browser. Public terminals, lab computers and family PCs will all register as single users.
- Recommendation — Limits on privacy (a good thing) and technology (not so good) prevent analytics tools from achieving the 1:1 visitor tracking utopia. For the foreseeable future, the unique visitors metric offers the best approximation of audience size. Just make sure bosses and advertisers understand the limits.
Page Views — A page view represents a single view of a single page under a certain Web domain. If you click to another site and then click back to the original site, you’ll log another page view. If you refresh the page you’re viewing, another page view will be counted.
- Caveat — A single visitor can log dozens of page views, especially if they’ve got an itchy refresh finger.
- Recommendation — Page view figures should be used for general analysis. Their real value comes from the manual parsing of page view data. Close examination will reveal popular pages and topics, which can help guide future editorial efforts.
Pages Per Visit — The Web’s built-in context makes it possible to attract visitors with one piece of content, then present them with additional material on the same site through related links, embedded links, recommendations, etc. A high pages per visit average (3+ pages is quite good) means visitors are interacting with your content. A low average means visitors are viewing one page and quickly moving on to other sites.
- Caveat — Want to see how the pages per visit average can be manipulated? Visit any major media site and look for the photo galleries. Placing a single photo on a single page and then encouraging users to click the “Next” button is an easy way to boost the pages per visit number. Pages per visit is also influenced by traffic spikes. If you receive an inbound link from a popular recommendation site (Slashdot, Digg), you’ll likely see a huge increase in page views but a dramatic drop in pages per visit. Most visitors from these sites look at one piece of content and then move on to the next popular destination.
- Recommendation — Like most analytics measurements, the pages per visit average should be examined over multi-month stretches. Traffic spikes should be disregarded — not ignored outright, just disregarded in this case. If you see the average go up by a full page over the course of 3-6 months, you’re doing something right.
Average Time on Site — The more time users spend on your site, the more you can assume they’re engaged with your content and your brand … and your sponsors’ brands. Given the hyperactive nature of Web browsing, holding visitor attention for a full minute or more is considered a success.
- Caveat — As the Google Analytics FAQ notes, some visitors leave unattended browser windows open. Analytics tools make no distinction between an engaged viewer and a distracted viewer with messy browsing habits.
- Recommendation — Analysis over a multi-month period is the best use for this measurement (sound familiar?). Consistent growth = good. Consistent decrease = bad.
Again, this primer is the tip of the analytics iceberg. There are many related topics worth further discussion and inquiry, including search engine optimization and Web advertising models.
There’s an interesting shift that’s also worth monitoring. Some publishers are looking beyond site-based statistics to gauge their overall reach across social networks, recommendation engines, RSS, mobile applications and other distributed platforms. Douglas McLennan, the founder and editor of ArtsJournal, touched on this topic in a recent interview:
I’ve come to the realization that ArtsJournal is not just a Web site anymore. Only 25 percent of our users ever come to the Web site, the rest get it through newsletters. We have 35,000 newsletter subscribers. Others get ArtsJournal through “newsbeats” that we provide on other Web sites. Some people get ArtsJournal through RSS feeds. In the course of an average day, there are 45,000 to 50,000 visitors — people who use Artsjournal every day. The unique visitors per month is probably 250,000. We probably get 500,000 to 600,000 visits a month and a few million page views. So ArtsJournal is not huge by the scale of large Web sites, but it’s substantial.
We may eventually see Q scores — or a variation on that concept — integrated into future analytics toolsets.