People and Not People

First off, there’s the difference between Traffic Viewed and Traffic Not Viewed.
In general terms, Traffic Viewed is generated by people. This isn’t a completely
sure thing, but it’s close enough for most purposes. Traffic Not Viewed is generally
generated by things that are not people. This includes robots, worms, or
replies with special HTTP status codes.
Robots are software programs that access web pages for their own purposes.
Search-engine crawlers (also known as spiders) are robots that index web pages
for inclusion in their search results. There are other spiders with less savory
purposes such as harvesting e-mail addresses for use by spammers. Worms
attack your web server, either to shut the server down (a denial-of-service
attack) or to break into the server. Either way, worms can create a large amount
of traffic that is of no interest beyond making sure it doesn’t overwhelm your
server completely. We’ll get into “special status” HTTP requests a bit later. But
in general, these are “noncontent” responses that redirect the visitor to another
page or inform the user that the page cannot be found.

Bandwidth

The bandwidth measurement is a webmaster’s first lesson in the importance
of collecting useful metrics as opposed to useless ones. With the exception of
knowing whether a site is nearing or over its bandwidth limits, there is pretty
much no useful business purpose to a measurement of bandwidth. Most web
sites don’t benefit from knowing the size of the average download.
With one small exception. Here in the United States, we tend to think of
everyone as having high-speed Internet. The fact is that broadband penetration
is less than 50 percent in the United States. According to the Organization
for Economic Co-operation and Development (www.oecd.org) only 137 million
people have high-speed access worldwide. Such figures could mean that half
of the people who visit your web site are using dial-up at 56 Kbps or less.
At 56 Kbps, loading time for pages and other content such as multimedia is
a big issue. It used to be that you had about 10 seconds for your page to load
before a user would abandon the page. Now you have about two seconds. You
can use the average bandwidth per visit along with the average pages per visit
to get a very rough estimate of how much data your average visitor is downloading
and how much time it takes.

Hits

For the first few years that we had web sites, we all quoted the number of
“hits.” It wasn’t until 1997 that we realized hits are another meaningless metric.
Why? To a web server, any access of any document — a page, a script, a
multimedia file, an image, and so on — is a hit. Because one page or site may
have lots of images, and another may be mostly all text, hits become a particularly
poor measure of a site’s performance and an even worse measure of how
a site performs in comparison to other sites.

Pages

Finally, we’ve reached a meaningful metric — pages, also known as page
views or page hits, the subject of Figure 2-4.
Back in the dark ages of 1997, when we were all using page counters, page
views were what we were actually trying to count. In AWStats, the Pages metric
is the aggregate of page requests.
Still looking at the summary on the main page, scroll down to (or click the
navigation link for) Files Type. The Pages total, 37,395, includes 19,037 static
HTML page views, 18,330 dynamic views for pages with a .php extension, 27
CGI script accesses, and 1 “com” page, which has no description. You wouldn’t
be a dummy if you didn’t even know what that file type was. As it happens, it’s
a command file, a program, but exactly what it does is beyond our scope here.
Is that com file a page? Why? A program can output a page. Not always, but
that’s one of the caveats of analytics software — assumptions. AWStats makes
the assumption that a com file is a program that outputs a page, and it counts
an access of that com file as a page.
Is it a page for business purposes? Unless you have a com file that you
specifically know produces a viewable page, it probably isn’t. And that means,
for business purposes, that this portion of the Pages metric is meaningless.
Only pages that are pages should count. If you have a lot of pages that are not
pages counting, it’s a problem. If it’s only a few, a small percentage of your
total, you’re probably safe to ignore the pages that are not pages.

Number of Visits

The Number of Visits a web site receives should be straightforward. That
would be nice and easy, wouldn’t it? Of course, it would.
No such luck, as Figure 2-6 indicates.
Like Pages, Number of Visits has two key assumptions: How long a visit is
and how much time has to pass between page loads to make one person have
two visits? Fortunately, there are industry standards — after all, this isn’t 1997.

A visit is as long as it is. As long as the visitor keeps clicking from page to page,
it’s still one visit. However, when the user stops clicking for 30 minutes, the
visit ends. If the user starts clicking again, it’s a new visit. Thirty minutes is the
industry-standard timeout for visits.
So, say a user toddles into SkateFic.com at 9:00 a.m., and between 9:00 and
9:30 she clicks from page to page, reading her favorite serial fiction. At 9:30 she
gets a phone call. For the next 28 minutes, she talks on the phone. When she
hangs up at 9:58, she finishes reading the page she left to answer the call and
loads the next page at 9:59. That’s one visit, because the break between page
loads was less than 30 minutes.
Now, say the same user is having a Grand Central Terminal sort of day. The
phone rings again at 10:00 a.m. This time the user talks for 31 minutes. When
she goes back to reading and loads a new page, she’s initiating a second visit
as far as AWStats is concerned. Same person, same day — and, if you asked
the user, same visit — but for pretty much every stats and analytics package,
it’s two different visits.
The average of 1.23 visits per visitor varies in meaningfulness. For a site that
gets a lot of returning visitors, it might have some meaning. For a site where
90 percent of visitors never return, the average doesn’t mean much, because it
is dragged down by the vast bulk of people who never return. You could have
10 people who average three visits per month and 90 people who come once and
never come back. Average visits will be 1.2, but it won’t be a very useful metric,
except to tell you that most of your visitors don’t return after the first visit.