Aengus Lawlor wrote:
> But even Jims suggestion only gives you a site specific metric. Some
> sites are far more likely to be visited by AOL users than others, for
> example, and the ratio for a site with a lot of AOL users will be
> severely skewed by AOLs proxy configuration. And that's the whole point
> - there is no way to express this in a general fashion.
I'm not saying that this ratio has anything to do with known quantities. I
don't think, as you say, it bears any relation to "visit" metrics. Rather,
this is an example of a quantity that can be measuerd and may provide insight
that wasn't there before. Perhaps, after experience and analysis of the
metric, it will be determined to provide a good, fast metric for how balanced
the load is on the site -- something entirely different than "visit" data.
> And I recognize that there can be value in such metrics. But a measure
> such as ROI is only valuable as a "rule of thumb" metric because it's
> based on very solid data. Log files just don't contain accurate
> information on the things that your are proposing to measure. I can be
> reasonably sure that a site that records 100,000 page requests a day
> gets more traffic than one that gets 10,000, but I can't reasonably make
> the same judgement about a site that records 12,000 requests, without
> digging in a lot deeper.
Log files do contain very solid data. In fact, no matter how much you audit
your books, it's likely that you web logs may be more accurate than your
financials. This could be completely wrong too. One of the problems with web
log data is precissely that no one ever does look at it. There could be
glaring errors (how many of us have seen stats that start long before the
server was set up or that contain requests dated in the future).
My suggestions is merely, that if we use quantities that we can trust than we
might be able to come up with "rule of thumb" metrics that provide insight
faster and more accurately than data mining, or heuristic approximations.
Jeremy Wadsack
Wadsack-Allen Digital Group
------------------------------------------------------------------------
This is the analog-help mailing list. To unsubscribe from this
mailing list, send mail to [EMAIL PROTECTED]
with "unsubscribe" in the main BODY OF THE MESSAGE.
List archived at http://www.mail-archive.com/analog-help@lists.isite.net/
------------------------------------------------------------------------
