Referrer Data, Referrer Data - Analysis Options

[Nick]

We're just starting to parse out referrer data from our weblogs and store it to allow a variety of analyses of traffic sources etc. Can anyone help me out in what to look for as the different flavors of Google referrer traffic ? eg there could be at least 3 types of activity from Google:
1. Organic searches on Google
2. Adwords on Google
3. Adwords on Google served against distribution partners content pages

What would the referrer data look like in each case and what other data as well as search term and search page can be extracted from the referrer string. Each engine seeems to have their own code structure

Thanks

[Matt B]

For starters, there is also regional referrer traffic from google (for example, Google.it - Italy)
Also, google image search may also show up as a referrer, depending on how aggressively you label your images or how popular the images are for certain searches.

[Haystack]

Content targeted ads (Adsense) will show a referrer of Googlesyndication.com but they don't show what site the person actually clicked through from.

Adwords looks the same in the referrer as a click from a natural search but my tracking software breaks them apart based on the tracking code I add to my ads

[Vertster]

The "q=" designates the query string used for the search. This is not the case for partner sites, such as AOL which uses "query=" for some searches.

What software are you using to analyze your logfiles? Some do this automatically.

[Nick]

We're just doing the analysis ourselves (in an Oracle based datawarehouse and using their Discoverer report generator) and want to be able to look at all the elements of the referrer data and match these to subsequent sessions on our site, paths users take, actions they take etc and look for correlations back to the referrer data.

As such we want to be able to parse out as much data as possible from the referrer data.

Thanks

[Vertster]

We're just doing the analysis ourselves

I am wondering what you are hoping to get out of the parameters that some of the top end log analysis packages can't tell you?

You can do most (if not all) of the analysis you mentioned with Clicktracks. It has a capability where you can tag a user based on things like when they got to your site, whether or not they were a new visitor, which search engine they came from, which PPC they came from, what keywords they searched on, and then analyze just those type of users all the way to purchase or not.