In-Reply-To: 
Message-ID: 
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Status: O
X-Status: 



On Sat, 13 May 1995, J.J. Linden wrote:

> On Sat, 13 May 1995, Ric Manning wrote:
> 
> > Now that my editor is getting interested in publishing something on the web, he
> > wants to know how many people read online editions? I can tell him how many 
> > hits my Gizmo Page gets and that I understand San Jose gets clocks 50,000 a 
> > day (or did before it imposed its subscription fee).

> > Would any newspaper Web publishers be willing to share a few details about how
> > often their creations are read and perhaps how much progress they are 
> making > in getting Web sites to pay for themselves.

> As a non-newspaper publisher, I can't address that question; but a quick
> caution about numbers of hits.
> 
> There are probably zillions of different ways of reading the access log 
> files of web servers.  Raw hits is probably the worst of all of them.
The accounting varies from server to server, but most servers keep a
"url accessed by hostid.port" log.  This is sufficient to provide some 
good statistics.  In addition, many web servers ask for client info
(to help determine best presentation) and log that as well.  A lycos
search for web statistics gives you several different resources.

> The problem is that if your page includes a large number of small 
> graphics, every visitor's client needs to access your server once for 
> each file.  A single visit might well represent ten or even twenty 
> "hits."  Needless to say, since many people use different methods of 
> reporting their sites' traffic, and very few of them describe those 
Give the log information above, you can pick specific items such as text 
pages (number of unique visits), cachable items (number of visitors)
and unique IP addresses (how many firewalls you found).

> methods in their reports, the impression of whose sites are how busy are 
> badly skewed.
This is true, for the most part, the estimates are extremely 
conservative.  Remember the "census taker" sketch by Lenny Bruce?

> The bottom line is that because of caching proxies and other doo-dads, 
> it's impossible to really know how many unique accesses you're getting; 
> furthermore, without careful -- very careful -- study of the access logs, 
> it's difficult even to state how many of your visits are from unique 
> people and how many are from repeat visitors -- "regulars."
If you think about it, you can't even do that with television.  When 
comparing the metrics for Web pages and servers, you can get incredibly 
precise demographics from a web log.

> The bottom line is to take responses to your question with a grain of 
> salt in many cases, and where you know from an answer how the figures are 
> derived, recognize that they still need further study to give you 
> information which is really useful for your own case.
What is most important is:
Does the potential advertiser believe that your page will bring him 
qualified customers?  The second question is:  How many referrals
has he actually received from your page?

If the Mercedes Benz dealer gets 5 qualified customers and 3 buy cars 
each week, he doesn't have to worry about whether you are getting 35,000 
hits on your home page.

On the flip side, the guy who wants a $50,000 Benz doesn't have lot's of 
time to go searching through the universe of the Web.  As a result, he
may find your page as a result of a search that caused only 1 hit
(Mercedes Manhattan), and you passed it through when he hit the dealer's 
button.

Either way, the dealer nets about $15,000/sale.  Paying you $1000/week
for one page that isn't even the home page is a trivial investment.
If you happen to have 200 pages that consistently get hit by a few 
qualified customers each day....

Kismet!

	Rex Ballard
	Standard & Poor's/McGraw-Hill
	Opinions expressed do not necessarily reflect
	the Management of the McGraw-Hill Companies.



From rballard@cnj.digex.net Tue May 16 02:34:32 1995
Newsgroups: comp.text.pdf,alt.hypertext,comp.text.sgml,alt.fan.mozilla