Relationship between distinct hosts and actual visitors

anyone know this exactly?

[webdevsf]

Given that IP addresses are not an accurate measure of how many users have been to your site, has anyone ever tried to figure out the "variance" between distinct IP addresses and actual unique visitors? I would guess that it would be relatively constant - (ie, similar numbers of proxy servers, etc)

It's not particularly scientific, but anyone have any theories here?

[hakre]

i think you can only measure it for one site or specific sites en group and not for the complete net. on the other hand it will be problematic to differ-out what unique visitors are if you can't count on the ip adresses. even if somebody visits via a proxy you (can) get the remote address of that guy in your logs i think. so proxy shouldn't be the problem. i think about lans connecting via a gateway to your site. you can differ anymore unique visitors out of that gateway.

[indomitable]

I can give you a couple of examples of which I have first hand knowledge:

January 2003:

Site A has c. 150k PIs per day (Commercial site)
Site A Unique Users = 961000
Site A Unique IPs = 283108

Site B has c. 180k PIs per day (Charity)
Site B Unique Users = 747544
Site B Unique IPs = 325100

Hope that helps.

[webdevsf]

How are you calculating unique users? Cookies?

[indomitable]

Yes. Cookies have a unique ID number and these counts are the number of distinct cookie IDs recorded during the month.

The cookies are first party and we find that the number of people who reject them is very small. If they do we only lose some of the cached traffic anyway.