I admin analytics for an online yellow page site, which monetizes many advertisers based on the click action itself.  (Pay-per-click)  Needless to say, segment development is a key part of my traffic analysis.

For those that have not tinkered with Omniture, there are three levels that one can segment traffic to.  (They call them ‘containers’)  The Page View, the Visit, and the Visitor.

Before explaining what each of these containers does, let me first define what each metric is:

1. Page View – A web based file resolved in a browser.  Pretty straight forward.
2. Visit – Starts when the first page loads, of the site  being tracked.. continues until all activity has stopped for 30 min.  (So, I can leave and come back to the site, and still be under the same visit)
3. Visitor – Persistent metric to the prior.  Regardless of the number of visits, if the cookie remains active in the browser, this will be read as the same visitor.  (and a repeat visit)

So to create a segment, I can apply any reported characteristic to whichever of these “containers” I choose, and apply any metric to that subset of traffic.

As such, the page view container is the most restrictive, and the visitor is the most broad.  If I decide to segment traffic based on page name, and only want  to show the amount of page views where page name was equal to ‘Page A’, I would apply it to each of the three layers:

1. Page View – It would display the page views that where page name was equal to ‘Page A’.
2. Visit – It would show the number of total number of page views, from all the pages where ‘Page A’ was referenced in the visit ..at any point in the visit.
3. Visitor – This would be the total number of page views from a browser, if the browser had loaded ‘Page A’ at any point. (Or, in all visits, was there ever been a ‘Page A’ page view..)

I write all this so I can complain about the onclick.

The onclick is the javascript function for an image request when an item is clicked on.  (Not surprisingly)  Onclicks are commonly used amongst publishers to capture usability insight, as clicks do not result in new page loads much of the time  ..which is fundamental for Omniture to function.

However, and this is the rub, the onclick image request fires outside the page being resolved in a browser..  As such, it can only be segemented to the ‘visit container’ and not the ‘page view container’.

This causes an inflation of numbers as there are, to use the example above:

More site-wide total page views, where the visit accessed page A at any point (Visit container) than actual page A page views (Page View container).

This holds true for all onclick segmentation, and this has plagued me since I have worked these analytics suites.

Correlation and subrelation are the only ways to get clean data with an onclick.  (These will be discussed in other blog posts, if you are not familiar) The big handicap with those two is the number of rules/traffic characteristics that can be applied is greatly reduced..

Stay tuned. More on those later.