NONE: Re: ONLINE-ADS>> web tracking
Re: ONLINE-ADS>> web tracking
Kim Brooks (kbrooks_at_cortland.com)
Sun, 20 Oct 1996 12:58:03 -0700
In response to several replies about web tracking solutions:
>While getting a summary of the traffic on your site is something your
>developers can do in-house just as well as I/Pro, ABC, BPA etc., their
>audits are valuable as "proof" of your audience size for your
>advertisers and potential advertisers. Audits let potential advertisers
I have heard this reasoning from several 3rd-party developers, but I have
never had the issue brought up by the advertisers themselves. None of our
advertisers question our numbers because they come from in-house solutions
(ie, Interse). Maybe we are just lucky because as a magazine, we had an
established ad base and reputation. Maybe advertisers know that you can't
lie too dramatically because their logs will show click-throughs and such.
Have other sites run into the doubting-Thomas phenom yet?
>How can a standard log detail a path through site, and if that claim is
>made, what is the accuracy of it? All you know about a hit in a standard
The path is assumed by the sequence of requests logged by an IP address
within one visit. For example:
OCT 18 1996 10:00 196.54.234 home.htm
OCT 18 1996 10:03 196.54.234 news.htm
OCT 18 1996 10:06 196.54.234 story1.htm
Would tell you that the visitor from 196.54.234 first went to home, then to
the new menu, then to story #1. More or less. Yes, it could in theory be
multiple visitors from one domain. But really, we are looking for an
ESTIMATE of patterns over time, and with 100,000 visitors aggregated per
week, it gives you a good general idea.
>Geographical resolution is the biggest crock of bull I've seen,
>especially in I-PRO ( I believe ). They say when you get 500 hits from
>Microsoft (microsoft.com), that means that 500 hits are logged to
>Redmond, Washington. Well, the local Phoenix Microsoft office's domain also
ends
>with microsoft.com. So, your tracking software will say these guys are in
>Washington?
Again, I realize that there are inaccuracies in logging. But again, we are
looking for general patterns and trends. I assume that some percentage of
our hits logged from Pheonix are actually from NM, CA, WA or who knows.
Some percentage of our CA hits are actually from WA, OR, NV, etc. So
hopefully, alot of the bleed balances out and the numbers are useful within
a range.
>...And 404 error recording doesn't belong in a "marketing analysis"
>package. A 404 error is when you dial up a page and it isn't there. This
>type of stuff gets recorded in the server error log by default, so this
>is not a feature of a stats package as far as I can tell (or would pay
>for).
Of course it's not a marketing feature, but it certainly belongs in web
tracking discussion. We use our web tracking not just for blowing numbers
at our advertisers -- we want to improve our site. This includes watching
how well new features do, tracking when old ones are burning out, AND
eliminating errors.
Lastly, there were some interesting comments about how cookies and/or sample
surveys are the only REAL marketing information out there. Personally, I
say that's true statistically, but both solutions are invasive and may not
give you any more of any accurate sample than logging. We definitely use
surveys to complement our logs -- we have an on-going registration system
with demo questions - but the response to surveys is low enough that I
wouldn't want to rely just on those. Cookies won't tell you any more than
logs unless you get the user to surrender that information, and we all know
what a debate personal information stirs up.
Again, I realize that web tracking is not the end-all, be-all solution to
understanding the web audience. But in terms of recording the trends and
patterns of your site, it's a darn good start that doesn't bother anyone.
Kim Brooks
U.S. News Online - http://www.usnews.com
