This is a post by Hig – about the sprawling parts of this website beyond the blog, and how we figure out what’s going on over there.

Running a web site is kind of odd… it’s a medium for talking to people, but often it’s hard to know exactly who we’re talking to. Sometimes, we do hear from you directly. You post comments on Facebook. Or you send emails with comments or questions. Sometimes you run into us on the street and exchange a few words with us. But these individual interactions are infrequent and not entirely representative of our readership. For example, we never hear from the folks who stop by briefly for a snippet of information and then go on their way.

There are tools for tracking who the “silent majority” of website visitors are. We use Google Analytics, which amounts to having a silent monitor running on our site that records tidbits of information. It answers a variety of questions: What pages are people visiting? How did they get there? What city is their internet connection in? How long did they spend on particular pages? Were they on a conventional computer, or on some sort of mobile device? Where did they go next? If you want to know which pages you need to improve, this is great data.

Figure 1: Organic search traffic for two sites since January 1, 2007. GroundTruthTrekking.org (GTT), in blue, has been actively maintained and updated since its inception in early 2007, while AKTrekking.com has been nearly unchanged in that time (it was launched in 2003 I think). In both 2011 and 2012, traffic doubled during a period of about 2 months in the fall. In previous years, this same period likewise marked a similar increase in traffic. This fall increase, a dramatic holiday lull, a broad but distinct drop in traffic in the summer, and strong weekly cycles – all match the schedule of a typical student, suggesting that a major driver of our traffic may be students researching issues that we cover. Note the vertical axis is a log-axis; see my sidebar on reading log axes if you want to know more.

But there’s more to be known. In the mission of knowing our audience better (to try and provide more useful content) sometimes I try to figure it out. Recently I found myself digging around in our site’s Google Analytics data , and I came across something quite intriguing. I thought I’d share it here, partly out of enthusiastic data-geekery, and partly because this sort of detailed traffic information – proprietary data owned by website operators – is actually fairly scarce on the Web. (An exception: Family on Bikes reports on a traffic spike and then on longer term impact).

Figure 2: Variation in traffic through the week. The variation Monday through Friday is very similar for both sites, each showing a similar drop on Friday, but Saturday and Sunday both show much lower traffic on groundtruthtrekking.org than on aktrekking.com.

On our site, the majority of traffic (right now nearly 70%) comes from what’s called “organic searches.” This consists of people going to Google, doing a search, and clicking on our site. By comparing the traffic on our current site (groundtruthtrekking.org) to that of our old unmaintained site (aktrekking.com) we can get a sense of what the “normal” variation is in traffic for the sort of trekking and resource issue content we create, and what is unique to our current site (Fig. 1).

When we first looked at this data, the strong increase in traffic each fall seemed a striking pattern. In combination with a summer lull, a strong drop-off in traffic at holidays, and dramatic variation between the week and weekends, it seems like a big chunk of those visiting our site are doing so on the schedule of a student. Of these patterns, only the weekly cycle is repeated on aktrekking.com, so I looked in more detail at that pattern (Fig. 2).

Sidebar: Why use a log scale?

In data visualization, anyone who plots values on a log scale, which stretches the small end of a scale while compressing the big end, is viewed with deep suspicion. Why use a scale that distorts normal numbers? Why risk confusing your readers? Are you trying to hide something?

It turns out that there is a specific situation where you “should” use a log scale. It’s when you care not about the absolute difference between numbers (e.g., whether a number is one more or less than some other number) but instead about the relative difference between them(e.g., whether a number is twice as large or half as large as another.) Log scales are commonly used for the strength of earthquakes (magnitude) and for sound (decibels).

In Figure 1, we are looking at how traffic on our site varies through the year. In fall of 2009 our traffic climbed from around 20 hits per day to around 40. In 2012, over a similar time period, it climbed from 200 to around 400. Is it the best description to say that we saw a growth of just 20 in 2009, while in 2012 we saw ten times that growth? Or is it better to say that in each case traffic doubled? By plotting on a log scale, I chose to focus on doubling rather than the amount of absolute growth. The same difference in height on the graph implies the same change in proportional terms (note the bar indicating traffic doubling in 2011 and 2012.) If I’d used a linear scale, then the change in 2009 would hardly be visible, since in absolute terms it was very small.

Edit: This article presents an interesting overview of scientific evidence that log scales are in some ways more intuitive for humans.

The pattern is similar for both sites – steady traffic Monday to Wednesday, a slight drop Thursday, a big drop Friday, bottoming out Saturday, and recovering somewhat Sunday. However, the magnitude of the drop for groundtruthtrekking.com is much greater, consistent with a bias toward students who do research on our site during the week, rather than during their free time on the weekend.

So this is where we are. It looks like a lot of students find our site by doing searches. Further digging in Google Analytics reveals that our readers read pages in many key places on our site – some of the most popular are our articles on Acid Mine Drainage, on Pebble Mine, the True Costs of Coal, and on the Benefits of Coal. Additionally, they visit our map of mine prospects, our blog, and our homepage. A large portion of our readers are from Alaska, BC, and the Pacific Northwest, but we have growing audiences in some surprising places like Texas, Wyoming, Italy, South Africa, and the Phillipines. Plenty of mysteries remain, but I guess we’re doing something you’re interested in.

We’d love to hear from you – who are you? How do you use our site? What don’t we know about how you use our site, and what changes could we make to welcome more visitors?

Now back to work, improving content…


18 Comments - Click to Read and Comment »