FAQ

I understand Reporter 9.1 uses a different algorithm to calculate browse time than the 8.X version, can you tell me what it is?

FAQ ID:    FAQ267
Version:    8.0
Status:    Published
Published date:    07/23/2009
Updated:    08/11/2011
 

Answer

Some reports provide a datapoint called Browse Time. The intention of this statistic is to estimate how long a user spends browsing a particular website or category of a website.

 The formula:  Reporter calculates this by matching each source IP address and each user in the logs with a website. After a match occurs, Reporter tracks the activity of each user as seen in the access logs. As Reporter procesess each log line in each log file, it finds and adds up browse time for pairs (client IP address and username). A pair determined to have browsed a website for 30 seconds or more constitutes a page view and counts as browse time. However, if the view is fewer than 30 seconds and another page view is found within the Page View Combiner (PVC) cache time window (10 seconds by default), Reporter subtracts the time of the previous page view from the next and counts the result. This calculation is more accurate than the previous alrgoithm used in Reporter 8.x versions, especialy for cases where a
user is using two browsers at the same time or has background applications making requests. Reporter calculates browse time in real time during log processing. Furthermore, Reporer can only subtract the time difference from the last page view if it still exists in the PVC cache. For example, if a Reporter administrator sets the default browse time to 60 seconds per page and leaves the PVC cache time windows to 10, the 60 second value applies by default unless another page view is found for the same client IP address and user agent within the 30
second PVC window. Therefore, you could have pages with anywhere between zero and 30 seconds or 60 seconds of browse time. Typically, the default browse time is set to 30 seconds by default, which means all pages have a browse time from zero to 30 seconds, but never more.
NOTE: This does not mean browse time for a website will only ever be 30 seconds. This means each page view will be 30 seconds. If a customer actively browses a website, by clicking through pages after pausing for over 30 seconds, browse time for each page will be added up for that website in 30 second intervals.

Older versions: With Reporter 8.3.X we use to  attempt to break user traffic into sessions. However, we found this to be problematic due to simultaneous sessions, and background requests from desktop apps.  In Reporter 9.1, we no longer attempt to track user sessions., instead, we now track browse time, per user and client IP pair, based on page views.  This new algorithm,  we believe, will improve the accuracy of the PVC ( Page View Combiner)  to better differentiate between page views ( a person viewing content)  and non-page views ( an application making background requests) in an attempt to prevent the counting of background javascript requests as user browse time.

Definition of the Page View Combiner:   While processing each log line, the PVC combines multiple HTTP requests that are associated with the pair and a single Web page into a single log line. When a user browses to a Web page, often that origonating page triggers requests for more content, either from the same Web server or another server (for example, a media server that stores video or image content). Rather than regard each of these as separate requests, the PVC combines all of the views thereby only counting the origonal page view. For more information, on the mechanics of how this works, complete with diagrams of the process flow, please see chapter two of the Reporter 9.1 Adminsitrators Guide.  

NOTE: more information on the Page View combiner can be found here: KB1774

NOTE: To change the default values for Browse time, in version 9.1.x, see this article.:FAQ554 

 Examples: 

Example Notes: These examples assumes the default values of 30 seconds for browse time with a default PVC cache of 30 seconds or fewer. For example, if a user visits cnn.com and never loads another page (does not click through the various articles links) for three hours, the resulting browse time is 30 seconds. 

  •  Scenario 1: Employee A visits cnn.com for 40 seconds, visits yahoo.com for 20 seconds, and then leaves his browser on youtube.com for 2 minutes but does not watch a video or click links in the site.
    • Reporter calculates 30 seconds for cnn.com, 30 seconds for yahoo.com, and 30 seconds for youtube.com for a total browse time of 90 seconds.
    • If, however, the same user browses videos on youtube.com every 29 seconds, the resulting browse time is 30 seconds for each video, resulting in a total browse time of 120 seconds.
  • Scenario 2: Employee A opens two concurrent browsers—Internet explorer and Firefox—at the same time and performs the above scenario. The result is double.
    • The result would be double. 
  • Scenario 3: Employee A uses the same browser. By default all page views are given the default browse time, which is 30 seconds (this value is configurable). If Reporter processes another page view from the same client IP address on the same user agent while the first page view is still in the PVC cache (which is also a 30 second window by default), Reporter lowers the browse time for the first page view to the time difference between the page views.
  •  Scenario 4: Employee B vists images.google.com for 5 seconds and then clicks a picture, views it for 15 seconds, clicks the back button, clicks a different picture, and views it for 45 seconds. Reporter records 5 seconds for images.google.com, then 15 seconds for the the first picture (plus whatever time it takes to click back and click on the second pic), and then 30 seconds for the last picture.
  • If I watch a video on youtube.com for 1/2 without clicking away, it will be counted as 30 seconds of browse time.

 NOTE: Reporter calculates 30 second for every page view, regardless of how long you spend on it, but it's only counted if you spend more than 30 seconds per page.

For reasons why some browse times may be recorded as zero, see FAQ421
 


Rate this Page

Please take a moment to complete this form to help us better serve you.

Did this document help answer your question?
 
 
If you are finished providing feedback, please click the RATE CONTENT button. Otherwise, please add more detail in the following text box and then click RATE CONTENT.
 
 

Your response will be used to improve our document content.

Ask a Question