web stats and proxy servers

[cecilia]

I'm trying to get my head around the web stats for my very new site. The IP address for 2 visitors is for Tiscali and 1 for BT. These, I think, are myself and my friend. But these are not our IP addresses. I read somewhere that the IP address sent out can be that of the proxy server, not the originator. If this is true, it makes it impossible to identify the location of visitors, which is a useful feature of webstats. Has anyone got any thoughts or information on this?

[degsy]

It is true and it depends on what the script uses to record the IP.

e.g.
goto http://www.degs.co.uk

Alot of scripts will record the Remote Address.
In the case of many ISP's this will be a Proxy Server.

[cecilia]

Thanks, Degsy. If it depends on the script, is it possible to tell which script won't record the proxy IP and then find a webstats co which uses this script?

[degsy]

Sorry, it would be on a script by script basis.
If it is not in the documentation then you would have to test it yourself.

[cecilia]

Doh! Sorry to be dumb - what documentation??? Webstats don't seem to come with any but the most basic instructions, no tekkie info.

And - is this a can of worms? If the web stats are recording the proxy IP address, are they also recording other data from the proxy rather than the individual user?

[degsy]

Sorry, I was referring to different Web Stat scripts.

Some are better than others.

Data depends on the configuration of the Proxy Server and the Individual User machines.
Many these days have internet security programs that restrict or falsify data for privacy purposes.

[cecilia]

This gets worse and worse! Has anyone ever done any research on the accuracy of web stats? If so, do you have a reference? Make my day, give me some good news.... this has all been depressing so far :eek:

[degsy]

I would suggest that you try different script until you find one you like.

http://awstats.sourceforge.net/
http://www.needscripts.com/Resource/19644.html

[cecilia]

Thanks, Degsy, you've been a superstar. We're setting up a comparative trial using our existing webstat provider and the two you suggested. Will report back on results in about 2 weeks in case anyone else needs this information, if you think this would be helpful. It sounds from what you say as if each case will be different but it might add to our communal knowledge base.

I've taken webstats as gospel for years, it wasn't until I sat down to some serious analysis and found the discrepancies that I began to question. There's a lesson here for all webmasters.

I won't even think about the numbers which don't add up at this stage - am I the only one who takes a calculator and adds the individual hit figures and compares them with the total????? On the other hand, nor did I before this week.

it's been a revelation.

[degsy]

Hope it works out for you :)

[Steven.Bentley]

The problem is this - if you record that there has been a visit from a proxy server, you record it as 1 visit. It may have been a thousand users on the other side of that proxy, particularly if it's a caching proxy, but it still registers as 1.

I would take all web logs with a pinch of salt for this reason, among others.