April 21, 2004

Stats. We Got Stats.

Robert Cox makes some valid points about the unreliable nature of many kinds of blog traffic stats, ranging from how "visit" stats include repeat visits, the impact of RSS feeds on inflation blog traffic stats, and the ways some unethical bloggers cheat to increase their hit counters.

Cox is right, but there are a few things I'd add.

1. SiteMeter, the visit stats tracker popular with bloggers because the TTLB Blog Traffic Ranking uses SiteMeter data, counts the same visitor twice if they visit your blog more than once, but more than half an hour apart. So, indeed, someone who visits your blog five times throughout the day is counted as five visits. However, all popular blogs have their share of regular readers who drop by often, so it still can be useful to compare one blog's SiteMeter stats with another blog's SiteMeter stats.

2. Free hit counters from SiteMeter, Bravenet, etc., are good for tracking trends, but aren't very accurate. They tend to undercount traffic - at least judging from much more robust data trackers that my web hosting company provides as part of my blog's hosting package. I've heard similar views from other bloggers who have better tracking tools than the freebies. Bravenet and SiteMeter are different in ways that make it valuable to have both on your site, even if you also have a better tracking tool from your hosting service.

Having SiteMeter allows an apples-to-apples trend comparison such as the TTLB ranking. And Bravenet tracks unique visitors on a daily basis, making it a useful data source in addition to SiteMeter - that's why I keep both counters on my site in addition to using the more robust data tracking applications provided by my hosting company.

One of the site counters provided by the hosting company, Awstats, counts unique visitors - a much better stat than SiteMeter's "visits." However, unlike Bravenet, Awstats tallies unique visitors on a calendar-month basis, with no day-by-day breakdown of "unique visitors." (Memo to Awstats: If you could provide daily unique visitor numbers and also provide data on how many of each day's unique visitors were repeat visitors during the month, that would be AwesomeStats.)

Visit my blog on April 1 and then visit it once per hour through April 30 and Awstats will count you as ONE unique visitor for the month, while Bravenet will count you as one unique visitor per day and SiteMeter will say my site was visited 720 times.

When I report on my blog that HobbsOnline has had more than 21,000 unique visitors this month, I am using the Awstats data. It means that more 21,000 different people have come at least as far as the home page this month. And when I say that HobbsOnline has about 1,500 regular daily readers, that number is based partly on the Bravenet data and partly on the Awstats data.

Awstats provides other useful data. It can show me which individual post has been viewed the most times. And it can tell me how many times that post was the entry page for the visitor to the blog.

Since I moved my blog to this URL on Jan. 1, the most viewed post on the blog has been this one, posted March 16, titled John Kerry: Let's Play Defense. Since March 16, it has been viewed 13,935 times. And 7,926 times it was the entry page for the reader, meaning they came to HobbsOnline from some other website or blog via a link to that specific article. Also, 7,264 times that post was the exit page. The flip side is that means that more than 6,600 times that post was read, the reader stuck around to read something else on my blog.

By the way: In April, that same post has been viewed just 231 times.

A lot of smaller blogs like mine occasionally see a traffic surge via a link on a major blog or from multiple links on many other blogs. I've often wondered if such links, going to a specific post, result in many readers sticking around to read other things.

Consider this post from a few days ago, titled Citizen-Powered News. It was linked to by several other blogs, among them Glenn Reynolds, Jeff Jarvis and Dean Esmay. (Sorry if you linked to it and I didn't mention you!)

According to Awstats, that post has been viewed 7,772 times so far, of which 4,555 times it was page the reader entered my blog. 3,790 times, it was a reader's exit page. So... what happened to the other 765 readers? Do they still have that page open on their browser? Not likely.

The way I have my blog formatted – the way most Movable Type blogs operate – when a reader is viewing an individual post, the headline to the previous and the immediate next post are displayed at the top of the screen. So I checked the stats and the immediate next post, this one, has been read 1,181 times this month - and that's only the people who opened the item individually, rather than just read it while scrolling the home page. Of those 1,181 readers, only 16 entered the blog on that page. It is reasonable to assume that some of the 765 readers of the Citizen-Powered News post clicked the link to this post as well. (Likewise, the previous post has been read 173 times, yet was the entry page for only six readers).

By the way, 11 different posts of mine have been read at least 500 times this month, and seven of them more than 1,000 times.

Yeah, so what, you're probably thinking. This is all rather inside baseball, isn't it? Well, yeah. But it illustrates a few things. First, while Cox is right to caution that "visits" stats often include repeat visits, surely it's reasonable to assume that an entry that is read nearly 14,000 times is not being read by one guy 14,000 times. It's being read by a lot of people.

Second - if you're a blogger and something you've just posted is being linked to by a lot of other blogs, or a few big blogs, make sure you have interesting posts immediately before and after it, to keep readers around long enough that they might decide to become a regular reader of your blog.

After all, repeat visits may skew the stats a bit - but repeat readers is a very good thing indeed.

[Editor's note: This post, originally posted around 3:30 p.m., was updated at around 11 p.m. with minor changes throughout to clarify a few points. If you are quoting from it, please quote from the updated/improved version.]