But this post is only mildly related to these books, since it started way before: the day I read about Gephi. Gephi is an open source graph visualization tool, to work with huge (or at least very big) datasets and graphs. I've seen it used to graph the friendship network from Facebook, or to graph tweet-retweets from Twitter (these two can be found in the Learn section from the Gephi website). In this post I'll guide you step by step to do a keyword-landing page directed graph, using data from your Google Analytics account.
A directed graph is a graph with arrows. You have a set of sources (in this case keywords) and targets (in this case landing pages), and aggregate sources by name. Plot this, and you have a keyword-landing page graph like this one (data for the last month in mostlymaths.net restricted to the first 500 keywords).
Keywords (arrow) Landing pages
What interesting things you can read from such a graph
- Clustering: In this particular instance, I can find the "clusters" of my blog. You can see 3 big aggregations of keywords and landing pages (left-middle,up-center and right-middle). Each one of these marks pages and groups of pages with several landing keywords. The groupings mark like the "big themes" in my blog, at least from search engine traffic (these are memory techniques, 9 best programming books I've read and seed germination respectively). Oddly enough, I've had one visit landing in seed germination from the keyword "gnus mail", thus the big seed germination cluster includes emacs.
- Big keywords: In gephi you can scale nodes depending on other table variables. For example, I can scale the landing page dots depending on the number of keywords, or the number of visits. In particular, I can show which keywords lead to more of my pages, by using the outDegree (number of arrows escaping) of a node.
- Big pages: Alternatively, I can scale nodes with the inDegree (number of arrows entering) to learn which pages have the most different keywords. I can also label the arrows with the number of visits (and scale them with this metric).
- Huge pages: with a little more work, I could scale with total visits to see which are my biggest pages together with the landing keywords.
Step by step guide
You can get a nice PNG file of you graph by pressing the camera-like icon in the Graph window (clicking the down-arrow will list the properties), or you can export a PDF or SVG with some more work from the Preview window. Both result in very nice images, but a SVG or PDF can be tweaked extensively using Illustrator or Inkscape. But it also needs more work from the Gephi side, so I won't enter into the details in this tutorial. Well, just a little! Be sure to check "show" before exporting, or you'll see a blank page: