Switching from web server log to page tagging
Job of the week
Featured threads
- How relevant do links need to be? 14 replies
- Tracking Online Response to Marketing/Communications Activities 8 replies
- Behavioural targeting software 4 replies
- Penalty avoidance on English-speaking foreign sites 5 replies
- 3 way linking - good or bad? 21 replies
Most viewed threads in last month
Most active threads in last month
- Effective use of anti virus software 0 replies
- lulgefeafCarf 0 replies
- Best Practice SEO Guide Jan 2012 0 replies
- Entry level search function 0 replies
- Introduction 0 replies

Head of Web Product Management at Penguin Group (UK)
22 November 2005 12:17pm
Has anyone made a switch from a server log approach to a page tagging approach for web measurement?
For many reasons our server log data had never been processed to meet ABCe requirements (we don't sell advertising space for a start!) and undertandably the page tagging approach is showing a significant difference in unique visitors and visits.
I'm having trouble getting buy in on the new figures so any 'me too' stories (anonymous or otherwise) would be gratefully received,
Jeanette
Technical Director at Box UK
22 November 2005 13:40pm
Hi,
Are you seeing a drop in the figures for the page tagging approach? Whilst developing a log analysis module, we investigated both approaches, and found that stats are often lower when using page tagging because search engine spiders usually do not parse the necessary javascript to register a page impression (i.e. your page tagging stats may be missing most search engine data). Unless you've already taken account of this in your stats? In theory, the page tagging approach should also register some 'missed' data that web logging won't have (the pages viewed from an ISPs or proxies cache).
Is there a particular segment that is missing in one data set (browser, etc), or is it a general difference in stats across the board?
Sorry I can' t be more help at this stage,
Dan
Head of Web Product Management at Penguin Group (UK)
22 November 2005 13:51pm
Hi Dan,
Thanks for your quick reply. We're seeing a significant drop in unique visitors, visits and page views across the board. It seems to big a drop to be apportioned to search engine's only but wondered if anyone had a similar experience.
Jeanette
Multi Media Developer at SportNetwork.net
24 November 2005 09:39am
Our raw logs say we do about 11 million page impressions, our Urchin log system which is javascript and log processing say we do 8 million. Urchin strips out all spiders etc.
Also as we have advertising on every page our advert server also says we do 8 million.
How did you count your unique users before? As using just I.P. addresses will give you very strange results, as most AOL users will have the same i.p. address but a BT Broadband user might have 10 different i.p. addresses in a week.
We have a difference in unique user numbers between our advert server and urchin, we put this down to our users not accepting 3rd party cookies as I can't fathom it out anyother way.
John
CEO at Econsultancy
25 November 2005 13:17pm
Hi Jeanette
I've never met anyone whose traffic/stats have gone up as a result of moving from log file analysis to page tags. And that's always been because the log file analysis stats have been overstating traffic - unfortunately it is almost always the lower number which is more accurate!
As John says the 2 most common reason for overstatement of traffic for log file analysis are:
1. Non-human traffic is being counted. And there's increasing amounts of non-human traffic out there! As much as 20% of site visits for some sites. With page tags, which tend to use JavaScript, the non-human traffic is not counted.
2. The same user is being counted as multiple users. This is where a 'user' is uniquely identified by IP address. If that IP address keeps changing (as it does for most consumers using ISPs) then that same person will be counted as a different user each time.
The above is somewhat of a simplification and can, to a large degree, be avoided even with log file analysis. For example, you can try and filter out non-human activity from your stats; you can also use a cookie instead of an IP address to identify unique users.
For an excellent overview of the pros and cons of the various approaches you should read Dr Brian Clifton's "Web Traffic Data Sources" paper - something we contributed to.
Regards
Ashley Friedlein
CEO, E-consultancy.com
VP Operations & Marketing at Celebrus Limited
25 November 2005 17:34pm
Ashley, your right increases are unusual, BUT there are reasons why it can happen.
If (for example) your content is cached then the cache can deliver a page without your web server ever noticing.... its another reason why you can't trust your logs...
Malcolm
CEO at Econsultancy
25 November 2005 17:38pm
Decreases can also occur with page tags if they are the kind which need to be put into every page / template and you omit to put them into all the relevant places, of course...!
Ashley
Head of Web Product Management at Penguin Group (UK)
25 November 2005 17:43pm
Hi Ashley,
Thanks for the responses (and to everyone else to). We were expecting the numbers to drop we just weren't expecting them to drop that much!
I am seeking an acknowledgement from others that the numbers are in fact very different (which I think you've all given me). Hopefully I can now persuade the site owners that although this is a bitter pill, it's a necessary pill. They might need a good stiff scotch to wash it down with though ;-)
Regards,
Jeanette