Web Site Statistics
The real object of Search engine optimization is to popularize your Web site and increase the number of visitors. To
see if it is working, you need a statistical program or at least a counter.
Before starting your Web site you should have installed a reasonably good statistical program that is based on a
Javascript installed on each page of your web site. Most statistics programs only analyze the log of the server.
"Before Starting" usually means just that. Most good tools require that you install a Java script or meter code in
each page. It may be hard to do that if you have already hand coded 2000 pages, but legacy pages can usually be updated
by mass find and replace editing even in manually coded Web sites.
Urchin seems to be the best tool around. It uses javascript as well as the log. Google Analysis is based on Urchin
but is free only for the first 50,000 page views a year apparently. If you do not have accurate statistics you cannot
judge the impact of your optimization efforts, but if your site is large enough you can judge from Alexa.
Various free counters like sitemeter.com provide a reasonable overview of traffic that is adequate for a small Web
site, and can be supplemented by host log statistics that are generally provided by hosting companies.
Statistical analyses are relative, because each system of measuring visitors or sessions or hits produces somewhat
different results. No two counters will show the same number of visitors, but they should give within 20% of the same
results for the same statistic. Statistics based on the Log of the Web host do not require a javascript. What they show
as "visitors" are generally pageviews or sessions (a half hour of connection as seen by the Web server. One visitor may
generate two or three sessions in my experience, and about 5-8 "pageviews."
"Visitors" are defined differently by different tools. For Urchin, a "Visitor" is a person who was at the site at
least one time in 24 hours. Others may define visitor as a person who was at the site for at least an hour, giving a
larger number of "visitors."
"Hits" or file downloads are not a good measure of Web site activity, because a single page may have a dozen graphics
files associated with it.
Pageviews, hits and file download statistics may include meaningless data from downloads of files such as favicon.ico,
which is the little icon that shows a symbol for your Web site in the browser tabs or favorites list, or robots.txt, a
file used by robots, or data from sessions created when a spider or aggregator or other device accesses xml or other
files used for propagating RSS.
Good tools will also give you information on what pages are entrance and exit pages to and from the site, which tells
you what pages are attracting visitors, and click-through paths - what paths visitors follow at your site. A good
program will also be able to separate out actual visits by people from visits by search engine or spamming robots.
Pageviews per visitor is supposedly an important statistic. If people aren't clicking links in your site, it may be a
sign of poor navigation design. On the other hand, if your site is a dictionary or other reference, it might just
indicate that they found what they wanted in the first page they visited.
Urchin and other tools may also tell you what key words are used to find your Web site, though these statistics do
not seem to be reliable and may not be broken down for each page. The tools will also tell you how much time an average
visitor is spending on a page. Obviously, a visit that is less than one second in duration is not a real visit by a real
visitor - probably robot software of some sort or an artifact. If many visits are for times less than 10 seconds, that
may indicate that a lot of real visitors are getting to your site and not finding what they want there.
Comparison of the figures given for "visitors" in your site statistics to those given by a different counter can be
instructive, because they can often tell you what the program is really counting. If the host statistics program says
you have 30,000 "visitors," but a counter like Site meter or adfree reports only 2,000 visitors, then the host program
is counting something other than visitors.
