On Thu, 05 Jan 2006 07:38:32 +0000, Roy Schestowitz put finger to
keyboard and typed:
>__/ [John Bokma] on Thursday 05 January 2006 07:17 \__
>
>> Roy Schestowitz <newsgroups@xxxxxxxxxxxxxxx> wrote:
>>
>>> I sometimes use a fairly recent version of Webalizer (now that I
>>> check, it *is* the most recent version) to analyse logs. I suspect
>>> that visit counts in Webalizer differ widely from the figures reported
>>> by AWStats due to handling of requests for feeds. Has anybody noticed
>>> the same behaviour? It's as though the definition of a "visit", let
>>> alone detection of crawlers, makes the term very subjective. What is
>>> it that people consider a visit if one tool reports numbers twice as
>>> big as the other's.
>>
>> AFAIK you can configure what webalizer considers a visit. IIRC it is "the
>> IP address has to be unique for x minutes". (And IIRC x = 15, default)
>
>I don't think I can tune it. Only administrators can, unless I am totally
>wrong to assume complexity. I never saw a Webalizer back-end even though I
>have used it since 2001. It used to be the only thing I thought existed.
>
>If so, that still doesn't make "visits" any measure which is objective. I
>guess that statistics should either come with some fine prints or taken with
>a grain of salt.
Statistics are fine, when used for their primary purposes. There are
two situations where any stats package will give you extremely useful
results:
1. Comparing relative popularity of different pages on your own
site, including (less accurately but still usefully) the most
common entry and exit pages. This is obviously useful for
determining which of your content is more popular with users,
but it can also play a key role in highlighting any navigational
issues - if pages are under-used, maybe they're not as easy to
find as you expected.
2. Comparing relative popularity of your site over time - whether
you are gaining hits or losing them. This helps you to decide
whether or not you need to put more effort into marketing your
site, and it's also an early warning of potential bandwidth or
server load issues in future (if you're gaining traffic).
Secondarily, stats can also give a good approximation of where your
traffic is coming from, both in terms of referring sites and
geographic origin. They can also give a fair estimate of the browsers
and operating systems used by your visitors. (None of these can ever
be 100% accurate, but it doesn't need to be that accurate to be
useful). This is less important from a sysadmin point of view, but can
have important implications for marketing a site.
The one thing that stats can't do very well at all, but everyone wants
them to do, is to give a reasonable comparison between your site and
someone else's. Claims of "x thousand" hits, views or visitors per
day/week/month/year should always be taken with a whole barrel load of
salt; not just because these figures depend on the software used to
generate them but also because they are affected by the design and
layout of the site.
Oddly enough, a well-designed site can often have fewer hits and page
views for the same level of popularity than a badly-designed one.
Consider, for example, the difference between a site where the main
content is accessible with one click on a front page menu, while
another requires navigation of a tree hierarchy several layers deep.
The former is more useful, but the latter will generate more hits and
more page views. Equally, having lots of external file dependencies
(such as images, javascript files and stylesheets) in a page will
increase the number of hits. A site can get a lot of visitors,
especially if it comes up highly ranked in Google, but if the site is
rubbish then those visits will be very short and unlikely to be
profitable. A less popular but more content-rich site may have fewer
visitors, but they may take in more of the content while they're
there. But this may not necessarily be better; sometimes what you want
is a lot of short-term visitors. Only the site owner knows what he/she
is trying to achieve with the site, and therefore only the owner can
know what constitutes "good" statistics.
Mark
--
http://www.GoogleFun.info - fun and games with Google!
"Everybody's changing and I don't feel the same"
|
|
