Google Analytics and other statistical software is based on combining the user agent (operating system, browser), IP and browser cookies to calculate visits on a site. While others are captured for different information, cookies are still the base of elementary web metrics, so it's crucial to understand them and have them under control. I did just that each time I did a test – clean Google's cookies, close the browser, open the browser, check it out. I made a hidden post for testing, one which definitely wouldn't get any other referrers other than me, besides using some old and already forgotten posts I made on Digg months ago and some indexed by Google images. Then I started testing different cases, which would help me understand the behavior of all the above mentioned things.
Referring vs. navigating
The first interesting thing I found out was that Google Analytics knows the difference between clicking on a link and manually navigating to a page by entering the URL in the browser. In the first case it recognizes the referrer, and in other one it doesn't (it is displayed as google / organic in the statistics). This is caused by the referrer information captured in the HTML header of every web page. So, if somebody manually enters your web page's address after seeing a link on Facebook, Facebook won't be counted as a referrer, but if they click on the link, it will be.
I tested a few url shorteners, tinyurl, is.gd, skrci.me and the one Twitter automatically uses (bit.ly). I found out that the ones I manually created and clicked on them inside the URL shortening site showed this site as the referrer. But in the case of Twitter, on which I made a tweet, clicked on the link and deleted it within seconds, Twitter was correctly shown as the referrer, even though the click first went to bit.ly and than to my blog. I went further, created a new shortened URL, put it in a hyperlink on a server, clicked on it, and again, this server was shown as the referrer. Because url shorteners only make the redirect, the click is keeping the original referrer, which enables the referrers to be fully captured, even if they go through the shortened URL.
I use a few Web 2.0 services to promote my blog, and it's been a trend for them to provide toolbars, which display the target site inside the parent site. The main reason for them doing this is to keep users inside their site, and in my opinion, it's annoying and it sucks. But at least it doesn't influence the statistics. I tested this behavior on Digg toolbar and on Google images (without removing or closing the toolbar) and in both cases it worked perfectly – the referrer was correctly recognized. After all, upon technically examining both cases, it's only an iframe opening the designated page below the toolbar, so the target page actually does fully open anyways.
The screenshots of Google Analytics below support my theory. In the case of the hidden post, I managed to create the following situations which prove my discussed behaviour of hyperlinks and URL shorteners:
- I clicked on a link from neolab.si (one referrer from neolab.si),
- I manually navigated to the post from neolab.si twice (by entering the url in the browser while being on neolab.si), shown as google / organic,
- I clicked on a shortened link on the is.gd site, later I pasted another is.gd short url directly to my browser (two referrers from is.gd),
- I clicked on a shortened link on the tinyurl site (a referrer from tinyurl),
- I clicked on the automatically bit.ly generated url in a tweet (a referrer from Twitter),
- I generated the skrci.me short url, put it in a hyperlink on localhost and clicked on it (a referrer from localhost).
The second report displays the referrers from Digg and Google images, both services using toolbars. On the first occasion, I opened two different posts inside the Digg toolbar, and on the second, two posts inside Google images toolbar, all of them without closing the toolbar (I went for one post twice, to check out if a session is also created and found out it is). As you can see, the referrers are all there and the toolbars don't corrupt the data in any way.
Look like we don't need to worry about these things anymore. I'm actually quite surprised about finding out the mentioned things work like a charm, not influencing the analysis and statistics in any case. From now on, there can be no more blaming these new features and gadgets on low traffic and weird referrers. The World Wide Web has been well planned and Google Analytics is able to know everything, so if your statistics seem weird, there is probably more chance that you are the one who's wrong.