2008-11-25

Self Referrer

Just got asked an interesting web analytics question. The client was reviewing the "Referring Domain" report in WebTrends and noticed a lot of direct traffic and that the site itself is listed as a referrer. They wanted to know why.

This is a question that I'm frequently asked by users of various web analytic products, and so I thought I'd post my response for others.


Direct traffic (as reported by WebTrends) comes from the following sources

1. People entering the web site URL directly
2. People having added the web site directly to their bookmarks and then clicking on the bookmarked site
3. People clicking on a link in an email
4. Other possible sources for links that are not part of a web site
5. People who choose to hide the referring web site information (a setting now available in some browsers)

In the report where the website is listed as its own referrer is a bit harder to explain, but here is my explanation:

Every page has a referrer, so when someone browses from one page to another page within the site, the site is in itself a referrer.

Normally, these should not appear in a referrer report (many analytics products strip this data out), but WebTrends does show this data under certain circumstances. What a site self referring traffic generally means is that someone was on the site who's visit session had time out (stayed on a page more than 30 minutes) and then clicked on a new page. So that a new visits was initiated and the visits referrer is now the site itself. There are a couple of other possible causes of this, but this is the easiest one to explain and understand.

1 comment:

Jacques said...

Add also the influence of the selected session tracking mode. If you use IP - User Agent, and the IP changes during the visit (which happens with some ISPs such as AOL). When the IP switches, WT identifies it as a new visitor, new visit, hence referred by the site the visitor was on. In WT, a high self referring rate is usually indicating that WT has some difficulty sessionizing.