Web analytics and a new theme

Five days of work. At the beginning, things went fast. But as often in the world of software development, the little details bog you down: by a rough estimate, I must have spent 50 hours on this new graphic theme to bring it to a usable level, if not a publishable one. All that for what? For the experience, for the résumé, but also to improve this blog's traffic.

This post serves three purposes: presenting the current state of the theme after a short week of work, presenting the first traffic results, and - more interesting for you - presenting the basics of web analytics on a concrete use case.

---

Where does the blog stand?

If creating a theme were a mountain climb, then I would have climbed two thirds of the shortest flank and would be getting ready to spend the night on a promontory not far from there. Concretely, all the basic features of the blog are there, the design has reached a semblance of maturity, and the ergonomics satisfies me fairly well. Time to step back for a few days, even weeks, in order to detect everyday usability problems and find inspiration to give it more visual personality.

Since my previous post, many small details have been changed and big modifications have been undertaken then abandoned. Among the latter, I attempted complete color swaps, the integration of widgets inside the main content, and even the "one-page" approach (a menu taking up the whole page, whose buttons lead to pre-loaded articles). As Drinij mentioned, this type of modern site is a pleasure for the eyes as much as for navigation (mobile or desktop). Unfortunately, such an organization is designed for a static site with few pages, not for a blog with some fifty articles: short of resorting to Ajax, it is unthinkable to achieve this "fluid" effect between sections.

As for the approved modifications:

• vectorization and integration of the logo in vector form (thanks to Isis for teaching me that Illustrator made this kind of task easier)
• integration of the "social" logos (including vectorizing the Sens Critique logo - if that poses a legal problem, do not hesitate to warn me)
• direct integration of the copyrights into the theme, rather than through an extension
• reduction of the line width to 50em maximum (50 characters, in essence), since this would be the ideal figure
• automatic scrolling to the top of the section / article as soon as the page loads and if you have not scrolled before

If we talk about the blog itself and not the theme, one of the next steps will most likely be to run a "card sort" to sketch the ideal organization of the navigation.

I discovered with horror that Jetpack was an undrinkable Rube Goldberg machine, and I hesitate to remove it completely (a post to be expected on this subject). In exchange, other plugins have been installed:

• one for "Related articles" (manual, this time), to improve the site's cross-navigation in accordance with "Kekona"'s suggestions
• a plugin to make pretentious Pull Quotes
• a Super Cache to speed up page generation.

The links have also been rehabilitated. My friends' blogs / personal pages are visible everywhere, and up to four links related to the section you are in appear in the widgets (the address book still needs work).

Thanks to everyone who has given me feedback so far. "Kekona", Isis, Ddril and Drinij offered me an outside view that is sorely lacking for someone who constantly has their nose in the code.

---

And does it work?

The final goal of this theme was to improve the reading rate of the blog's articles and to encourage related readings.

Measuring a blog's reading rate

It is possible to measure the reading rate of a blog's articles thanks to web analytics tools like "Piwik". These record the visitors' activity with varying levels of detail and allow the people involved in a site's design to understand whether it fulfills its role (more details on my use of the collected data). Thanks to Piwik, one can retrieve various measures to assess the visitors' engagement rate on a site, among which three seem important to me:

1. The bounce rate
2. The number of actions performed by the visitors
3. The time spent on each page

Here are a few definitions that will let you shine at office parties where your bosses hang out:

A "bounce" is a visit limited to a single page. The typical example is Gégé-your-neighbor who looks up funny jokes on the Internet and lands, by chance, on one of your old articles in which the term "funny joke" is mentioned. Gégé quickly notes that your post does not match his expectations, so he leaves your site after 3-4 seconds. That is a "bounce": someone who enters your home, skims the content and leaves. To limit the bounce rate, one must obviously work on the site's ergonomics and offer interesting content, but also fill in relevant meta-data for your articles. Search engines rely on this meta-data to serve their results, more than on the content of the inspected pages. In this blog's case, it is the SEO Ultimate plugin that helps me manage the meta-data.

In Piwik's language, an action represents everything a visitor can do on your site. Moving from one page to another is an action. Downloading a document is an action. Watching a movie too. By analyzing the number of visitors' actions, one can estimate their engagement rate, though it would be wrong to believe that viewing many pages is necessarily a good thing: it can also mean the site's ergonomics needs rethinking.

The average time on the site represents, as its name indicates, the average of the time spent by visitors on the blog. A precious but temperamental measure, since it is computed in a rudimentary way:

[time spent on page A] = [time at which you arrive on page B] - [time at which you arrive on page A]

One then adds up the time spent on each page and takes the average. The problem with this technique is that it is impossible to know the time on page A if the visitor does not move on to B. Visits limited to a single page are considered bounces, and the time spent on the last page of a visit (generally the most interesting page) is not counted.

To measure a blog's reading rate, one should therefore limit the bounce rate, obtain an average number of actions between 5 and 10, and have a time spent on each page consistent with the length of the articles (at least one minute for the shortest, 5-10 minutes for the longest). In this blog's case, it would of course be handy to be able to measure other things: the impact of the margin text (read or not? Does it lighten the article or not?), the impact of the blog's color (beyond personal affinity, does Turquoise foster reading?) or even the scrolling on the page, but Piwik does not easily allow retrieving data on these matters. Let us wager it will improve soon!

Isolating my big paws

Obviously, having worked like a madman for hours, I have affected the blog's statistics non-negligibly. It is therefore necessary to isolate segments of visits that could not have been affected by my activity. To compare the blog's statistics, three segments will be used:

• All visits: for the lol, the site's evolution with my activity.
• Visitors from France: living in Switzerland, I could not affect the traffic from France. French traffic is a neutral indicator.
• New visitors: as long as you do not delete the cookies generated by Piwik, a visit counter can tell whether you are a new visitor or a "returning" visitor (even Google Analytics does not translate these terms into French). Isolating the traffic to consider only new visitors also guarantees a certain neutrality of the results. Note however that I used certain browsers for the first time during my WordPress theme tests, so these marginally influence the statistics.

So, the result?

Rather positive, even though it is strictly impossible to draw any conclusion on the basis of five days of use: between my activity that is hard to isolate, the many articles posted recently and the #RPGaDay hashtag that attracted a few visitors, impossible to say whether the noted improvement is lasting or merely due to a confluence of circumstances. Moreover, September 2 was a black day for the blog, since the latter was literally wiped out for 15 hours (4:45 PM to 9:00 AM). As a reminder, the new theme was inaugurated on August 29.

All visits, up to September 2, 2014

French visits, up to September 2, 2014

New visits, up to September 2, 2014

All visits, weekly, up to September 2, 2014

French visits, weekly, up to September 2, 2014

New visits, weekly, up to September 2, 2014

In broad strokes, here is what we notice:

• The bounce rate is at its lowest since the new theme was introduced, in all the graphs. However, the daily statistics show that it is rising again both for the French and for the new visitors. Still, this remains very positive!
• Since the new theme was introduced, the average number of actions has globally increased. This is particularly true for new visitors if one takes the weekly graph into account. Here again, this is very good, even if the volatility seems high in the daily graphs.
• The same goes for the average time on the site. If one considers the weekly graphs, one can see that it has literally exploded.

These first, somewhat hasty measurements seem to indicate that the blog is going in the right direction. All that remains now is to durably increase the size of its audience...