Forum Moderators: DixonJones
I know on a Webtrends report, for example, it only gives an average per VISITOR, not visit.
Does anyone think the "per visit" metric has any merit?
My point is that page views per VISITOR is much more useful as a metric, and that's the metric I would always use when talking about page view behaviour.
I always say take the stats with a pinch of salt and think more about trends, as this is the only fair way to compare.
Hope that helps a little
I think duration (time per session)
is a much better metric.
aspdaddy: I'm interested in how you are measuring duration. If you can't measure the unload time of a page then presumably your session duration is measured as the time from the loading of the first page of a session to the loading of the last page, thus losing the length of time spent looking at the last page. As the last page in a session is often the one with the longest duration (i.e. they have found the content they are looking for...) then you should only really use duration as a metric if your stats package is able to measure page unloads.
IMO definition of a visit is whatever you want it to mean for the analysis you are doing, AFAIK only Nettracker & Site Analyser support this.
>measuring duration
Typically, I look for a end of session indicator, such as a "thankyou" page. Theres lots of other ways though, depending on the site.
In terms of a low page views per visit number. There are a few things to look at.
1. How many people never get past your first page? If your site is not a simple one or two page affair, then you could have one of two problems: poor site navigation, or poorly qualified traffic. One means the user cannot get past the first page, the other means that they have no interest in getting past the first page.
2. Can any request to your server be considered a visit? This is something you need to check with your reporting package. Unfortunately ours measures any request as a visit, so if someone is deep linking into an image that request counts as a visit, but it should not.
If you can run a report of traffic by external referrer, then do that and look for a high hit/page view ratio or a page view/visit ratio that is less than 1 in your external referrers. The second one only works if your package counts a hit with no page views as a visit. This is a great way to spot image thiefs. A high ratio means that a lot of people are seeing images/other files but not a lot of page views. A page view/visit ratio below 1 also tells you the same thing.
3. Check session time out, or if you are not setting your cookie until a couple of pages into the site. Both of these can screw with your ability to count uniques and visits accurately especially when it comes to page views.
If you are not setting your cookie until they enter the commerce side of your site then your reporting package will probably go ip/user agent for the first couple of pages, then see a cookie and use that for the remaining pages. Then you end up with 2 different uniques, both with low page views. When a visitor returns, the cookie is already set and the reporting tool won't make the same mistake. All the page views of visit 2 then go to that cookie.
So under this situation, if visit 1 and visit 2 both have 4 page views, then the tool might see 2 uniques. One would have 1 visit and 2 page views (this would be ip/user pre cookie), and the other would have 2 visits and 6 page views.
