Help! webtrends question

geographical location

[ritualcoffee]

You know in webtrends where it gives you the geographical location - can you give me a reason why some say "unknown"?

[agerhart]

mine say that too.......

[ritualcoffee]

how do they figure out location - do they just use the provider? Because I have something like 57% US and 29% unknown (think that is international?) I can't find the info in the oh so helpful manual - getting ready to call and bug them. I don't care about this so much as the product managers that keep calling!

[ritualcoffee]

kind of found an explaination:
[webtrends.com...]

[agerhart]

thanks ritualcoffee....I will definitely be using this

[henki]

Since the country is taken from the domain suffix, there will be a lot of faulty data.

In Sweden 50% of dialed up traffic comes from our largest ISP telia.com, with all IPs resolving to their .com doamin. That goes for many other non-US markets as well.
You need to know your ISPs and filter them with "early search and replace", exchanging the suffix. e.g. my filter says telia.com= telia.se

It is very common that local business operators wonder over all US traffic that their sites generate.

[ggrot]

You can get some very powerful geographical tracking services over at [geobutton.com...] I have no affiliation with them, but you will have to agree to a small button ad on your pages.

[Bolotomus]

If you're a big fan of Webtrends then forgive me, but in my opinion the 'geographical location' is like 90% of Webtrends other features: a bunch of hot air.

It's another example at attempting to know something which is actually unknowable, by making some really wild assumptions. In the end they produce a professional looking chart that looks very scientific, but in fact it's all based on rediculous assumptions.

In this case, the assumption is that a person's IP address tells you something about where they are located.

I heap it in the same pile as "session length." An excuse for another fancy chart.

</soapbox>

Bolot

[ritualcoffee]

yeah - but for product people who put webtrends explaination as a disclaimer it works just fine. IP location means a lot in some cases. It really all depends on what you are using the numbers for. Its marketing for gods sakes - it is all assumption.

[skibum]

"Its marketing for gods sakes - it is all assumption." :)

It probably shows lots of people in Virginia as well if the site gets many AOL users.

"like 90% of Webtrends other features: a bunch of hot air." since having to rely on WT data and make sense of it more often lately, Bolotomus, got any suggestions for other log analysis stuff to run/suggest. It does seems to supply more questions than answers when reporting on web site traffic.

[toadhall]

I agree with Bolotomus. You don't know with any certainty where the IP address holder is physically located. There are too many parts to the equation. For instance: you first have to determine if the IP address you have is the user's IP, and not a proxy server. You may have to extract it from a comma delimited string if there are more than one proxy server. If the net block holder has a dot com address the Webtrend method would report them as American. But the dot com upper level domain name is used world-wide (as well as .org, .net) so the N-B holder could be in Tokyo. And even if you built a script to accurately scan the whois information for a physical address you only get the address of the registered holder, not the physical address of the ISP's server(s). The registered Tokyo company could own an ISP in Korea, serving customers in Seoul.
<added>
See: [samoa.co.uk...]
</added>