NONE: RE: ONLINE-ADS>> Measuring sessions
RE: ONLINE-ADS>> Measuring sessions
Steve Podradchik (stevep_at_marketwave.com)
Wed, 19 Feb 1997 13:16:27 -0800
bruce_at_sprintlink.co.za wrote:
>I'd be interested to hear from others who have experience with site
>audits about what measures are being used by the auditors and whether
>there's a better way to get a sense of audience numbers than thumb sucking
>a session interval.
Hi Bruce,
The measurement of visits (or session if you like) is more of an art than a
science. Although using a standard measurement system may be preferred for
sites that need to allow advertisers to compare apples to apples, the value
is highly questionable when you have rotten apples ;)
30 minutes is, in our opinion, vastly too long a time period to allow for
visit period. In Hit List, for example, we use a default 15 minute
interval but a) Allow the user to change it and actually suggest that they
do because the # is probably too high for most sites (see below) b) Also
take cookies into consideration, if you use them c) Take the user-agent
string (browser and OS, basically) into consideration. Basically, if the
IP or the user-agent or the cookie changes, it's definitely a new visit
(actually there is some fudge factor in the cookie part because many sites
don't assign a cookie until after the first request during a visit).
In determining the 'correct' value for the timeout period (assuming your
s/w lets you change it), a webmaster should really consider how his or her
site is typically used. Assume two polar examples:
1. You have a search-engine site. People come to you, type a few
keywords, then jump somewhere else. You see an average visit length of
perhaps 2 minutes. For you, a timeout of 5 minutes might be best.
Additionally, as search sites tend to have very high traffic, you should
keep the threshold minutes low to mask-out IP re-use as much as possible.
2. You have an online magazine. Here you might have people spending 5
minutes reading a page, typically. In this case, you might consider using
a threshold of 15 minutes. But if you have low volume, this threshold just
doesn't matter much because the odds of seeing the same IP again within any
reasonable time period is small. In other words, if you changed it to 2
hours, you might not see any significant difference in the # of visits.
The idea is basically to set the threshold higher than your average visit
length but probably not more than 3 times longer (for you stat fans out
there -- think of the old 3 standard deviations, 95% thing)
Also, 25% unresolved IPs is absolutely typical -- that's just about the #
we've observed after processing hundreds of customer logs. However, this
isn't necessarily your biggest problem in terms of not really knowing WHO
is visiting your site. Most sites get a huge # of visitors from online
services and ISPs who, as you've pointed out, re-use IPs. And major online
services likes AOL use proxy servers extensively that further muck up the
works (i.e. you might have 10,000 visits from AOL but only 'see' 2 of them
because their proxy handled the rest w/o even notifying you).
Hope that helps,
Thanks,
Steve Podradchik
Marketwave
http://www.marketwave.com/
Makers of the Hit List family of web site log analysis tools
-------------------------------------------------------------------------
This week's Online Advertising Discussion List sponsor:
Q: Is it true that advertisers pay no commissions in the Adbot Network?
A: That's right. If you win an auction, what you bid is what you pay.
adbot: the auction market for internet advertising (tm) http://www.adbot.com
-------------------------------------------------------------------------
Online Advertising Discussion List To Unsubscribe send UNSUBSCRIBE
http://www.o-a.com/ to online-ads-request_at_o-a.com
