Mailing List Archive

On 7/15/2008 5:03 PM, Natalia Lis wrote:
> Hello,
>
> First, I would like to confirm something. I monitor a site which
> displays a lot of graphs which are shared with other websites. If a
> person visits such a website and looks at our graphs, the website will
> be listed in a referrer report even if the person never visited our
> site?

If a web page at http://abcd.com/graphs.html has one of your images
"embedded" in it, with code like this

< img src=http://xyz.com/graph001.jpg >

then a user that visits that web page will send a request to your web
server (xyz.com, in the example above) and the referrer field of that
request will be http://abcd.com/graphs.html.

If the owner of abcd.com copies you graph and puts it on his own site,
so that the html code looks like

< img src=graph001.jpg >

then the web logs at xyz.com will never see the request, and there won't
be any log entry for a referrer to be part of.

> In other words, report tracks the source of all requests and not
> just requests for pages? I'm asking because in a past discussion I saw
> that somebody expected the number of requests in a Referrer report to
> come close of the number of "requests for pages" (as if it was tracking
> requests for pages only) and this surprised me.

For most websites, most of the referrals are "internal" - if you have 10
images on your home page, then a request for your home page will
generate 10 additional requests, one for each image, and the referrer
for each of those image requests will be your home page. Even if the
referrer for the first request was some search engine, or a book mark or
something else, 10 of the 11 requests in your log file will have your
home page as the referrer. On the flip side, there's only one "Page
Request" in those 11 Requests, and the referrer on the Page Request
isn't internal.

> Second, is it normal that the number of requests in the Referrer report
> is only a fraction (5%) of total requests?

There are no "requests" in the Referrer report, as such. The entries in
the referrer report are URLs, and the "Requests" column indicates how
many times each listed URL occurred in your logfile. It would be
somewhat unusual to only have referrers on 5% of your log file entries,
but you should also check that your "floor" for the Referrer report is
low enough list all the referrers. (By default, Analog only lists
referrers if they occur more than 20 times in your logfile). If you're
getting a million requests a day, then a floor of 20 shouldn't mask
much. If you're getting a couple of hundred hits, and they're from a lot
of different referrers, then you might find that some of your traffic is
falling below this floor.

You can change the floor with REFFLOOR 1r

> Third, let's say that I am trying to find out all the major sites which
> display our graphs. Does the report list all of them? We have an English
> website which targets North-American market; yet, the sites listed in
> the Referrer report are almost all from East-Asia. This is not entirely
> impossible, but I find it curious and this is why I wonder whether the
> report is set correctly or whether we miss something.

The Referrer report lists all of the referrers that occur in your
logfile - if only 5% of the requests in your logfile contain referrers,
then the referrer report probably won't tell you anything that you
should rely on.

You might find the Referring Site Report (REFSITE ON) more useful if
yoru images are being embedded in dynamically generated pages that tend
to have unique URLs.

Aengus

+------------------------------------------------------------------------
| TO UNSUBSCRIBE from this list:
| http://lists.meer.net/mailman/listinfo/analog-help
|
| Analog Documentation: http://analog.cx/docs/Readme.html
| List archives: http://www.analog.cx/docs/mailing.html#listarchives
| Usenet version: news://news.gmane.org/gmane.comp.web.analog.general
+------------------------------------------------------------------------

In addition to everything Aengus said, have a look in your raw
logfiles. You will soon understand which lines have referrers on, and
what those referrers typically look like. Analog just counts those up.

If you want not to count referrers to images, you can use the command
REQINCLUDE pages
This will restrict all the reports to just the lines in your logfile
corresponding to pages. But having said that, I'm not sure that's the
right command for you. If other sites are including your images, you
might actually want to analyse that traffic. Of course, you could do
one report with and one report without that command.

--
Stephen Turner
+------------------------------------------------------------------------
| TO UNSUBSCRIBE from this list:
| http://lists.meer.net/mailman/listinfo/analog-help
|
| Analog Documentation: http://analog.cx/docs/Readme.html
| List archives: http://www.analog.cx/docs/mailing.html#listarchives
| Usenet version: news://news.gmane.org/gmane.comp.web.analog.general
+------------------------------------------------------------------------

Hello,

The site has dozens of millions requests per day so I think the floor of 20
is good. Also, the graphs that we share are embedded images; they are
displayed on other sites, but are hosted on our websites. Based on what you
wrote, I'm not surprised that the Referrer report only lists referrers for
about 5% of all requests. The site is very busy with graphics and I suppose
that most users have it bookmarked.



The difference between a referrer report and a referring site report is that
the former lists URLs while the latter lists websites? So if a website shows
our graphs on two different pages, the referrer report will show 2 urls,
while the referring site report will show one website address?



"If you want not to count referrers to images, you can use the command

REQINCLUDE pages"



This is interesting. So it will only count instances when another website
links to us and a person clicks on that link and actually visits our site?



Let's say we have a graph: www.oursite.com/graph and it is embedded on
another website in the following way: <A HREF="http://www.oursite.com"><IMG
SRC="http://www.oursite.com/graph"></a>


If I follow Stephen's recommendation, the report will only count clicks on
the graph, not just requests for an updated image?

Thanks for your help,

Natalia


On Wed, Jul 16, 2008 at 1:56 AM, Stephen Turner <
analog-author@lists.meer.net> wrote:

> In addition to everything Aengus said, have a look in your raw
> logfiles. You will soon understand which lines have referrers on, and
> what those referrers typically look like. Analog just counts those up.
>
> If you want not to count referrers to images, you can use the command
> REQINCLUDE pages
> This will restrict all the reports to just the lines in your logfile
> corresponding to pages. But having said that, I'm not sure that's the
> right command for you. If other sites are including your images, you
> might actually want to analyse that traffic. Of course, you could do
> one report with and one report without that command.
>
> --
> Stephen Turner
> +------------------------------------------------------------------------
> | TO UNSUBSCRIBE from this list:
> | http://lists.meer.net/mailman/listinfo/analog-help
> |
> | Analog Documentation: http://analog.cx/docs/Readme.html
> | List archives: http://www.analog.cx/docs/mailing.html#listarchives
> | Usenet version: news://news.gmane.org/gmane.comp.web.analog.general
> +------------------------------------------------------------------------
>