If you're looking for industry averages, especially for ecommerce, I just ran across the FireClick Index (http://index.fireclick.com/).  I have always been asked, "How is that compared to the industry average?"  Funny how everyone always assumes that such a thing exists.

Unfortunately, there is no detail on how these metrics are calculated.  Of course, being the thorough web analyst that I am, I emailed FireClick.  Here are the definitions:

Cart Abandonment Rate - the percentage of sessions where visitors add an item to the shopping cart, but do not make a purchase.

Conversion Rate - a percentage calculated by Fireclick as the number of sessions where an order occurred divided by the number of total web site sessions for the specified time period.

Average Session Length - average number of pages viewed in a visitor's session.

Average Session Duration - average time in minutes a visitor spends on the web site.

Average Page Display Time - average time in seconds it takes for a web page to load in a web browser window.

Average Page Read Time - average time in seconds the visitor spends viewing a page, from the complete download of the current page to their click to the next page.

Average Connection Speed - actual throughput in Kbps that visitors view the web site.

Special thanks to David Maloney at FireClick for replying with these definitions.

Today I had another enlightening discussion with the same coworker I discussed the bounce-back rate with.  Suppose you place a new promotion on page A.  As in the previous case, there is concern that visitors who follow the promotion will not return to perform an action called for on page A.

For simplification, let's say the action you want to measure change in is a visit to page B from page A.  The most obvious measurement is the fall-out rate from page A to page B.  If the fall-out rate increases, then the promotion impacted the pages original call to action.  Right?  Wrong!  If you expect visitors to return to complete the original action after following the promotion, then the fall-out rate will increase.  What, the fall-out rate will increase even if they still do what I want them to do?  Yep...

So how do I calculate the maximum expected increase in the fall-out rate?  Good question!  One assumption you will have to make is that the traffic is consistent before and during the promotion.  Take the average number of daily visits that did not complete the action (page A to page B) before the promotion divided by the sum of the average daily visits before the promotion and the average number of daily visits clickings the promotion (obviously during the promotion).  If the fall-out rate during the promotion exceeds this value, then your promotion may be driving traffic away from the original call to action (page A to page B).

You must determine what amount of traffic is acceptable to lose.  Sometimes, losing conversions to another promotion is not necessarily a bad thing.  For example, the promotion may generate valuable leads at the expense of decreasing product downloads for a short time.  Costs, such as these, can help determine how long to run such a promotion.

I recently had a discussion with a coworker about a current promotion.  There was much concern that the promotion would drive traffic away from the page that otherwise might have stayed and followed its call to action.  The most reasonable approach was to watch the fall-out rate of the pages original call to action.

But then he posed an entirely different question.  How many visitors that followed the promotion came back around to the original page?  Herein lies the difficulty - there is only one segment of these returning visitors that we can measure - those who immediately return via their brower's back button.  A simple path report for paths containing the original>promotion>original page pattern.  This is what is called the bounce-back rate.  It helps answer the question, "Does the visitor see what they expected from the promotion?"

Now back to the difficulty.   Here are some other segments of the returning visitors that I came up with:

• Visitors who open the promotion in a new tab or window and later return to the originating page.
• Visitors whose visit ends before returning later.
• Visitors who navigate beyond the promotion page and return sometime later in the same visit.

I can imagine there being a way to track the last two by assigning a unique ID to every visit, then counting the number of unique IDs that visit the promotion page at least once and the originating page one more time.  A method like this would also include the back button segment.  However, it seems that the first segment is impossible to track.

This proves, once again, that web analytics is not an exact science.