Sent by Chris Kaminski on 27 May 2002 05:05
On Sunday, May 26, 2002, at 05:04 PM, Tuttle Grey wrote:
> Their stats do seem reasonable when compared to my site logs. The
> numbers are not identical, but it seems to me that they do offer a
> useful indication of how many users are using which browsers,
> resolutions, etc. We need some stats, and unless someone can
> provide something better, these seem decent.
For your site, perhaps. Two examples from my own attempts to decide
which browsers to support and how far to push with CSS:
--When developing a CSS-layout site I was told it needed to look
'the same' in NN4.x but to disregard IE4.x/Win due to 'free' stats
about the overall Web audience. When I questioned this decision, we
looked at the logs and found that IE4.x/Win accounted for more
traffic than all NN4.x versions combined.
--When deciding whether to fix a minor but noticeable display bug
in NN4.x on another site, I checked the logs and found that in the
2 months the site was live, *no* NN4.x visits had been recorded.
Not one. I saved my client perhaps a day of debugging and re-coding
by ignoring a bug that didn't affect the functioning of the site,
and only appeared in a browser that wasn't being used on the site.
The sites in question were a newsletter site from a national
tourism bureau, and a community site for disabled teens, both with
in excess of 10,000 visitors a month.
> And, they are free.
That isn't the point. The relevant issue is how much *good* numbers
cost. If any alternative in a decision you're making based on the
stats costs significantly more than getting good numbers, you
should probably pay for the good numbers (or the analysis software
to get them, or the time to scrutinize the logs or whatever). If
getting good stats costs more than all alternatives (i.e.,
re-coding the site to accommodate users of browser x; lost sales
b/c you're not supporting browser x completely, etc.), then by all
means use the free stats.
> CodeBitch wrote an article demonstrating how her own site stats
> were quite different. Of course they were. The sites for *any*
> individual site will be different, as that is a much smaller
> sample. Those sites are by definition much more skewed, as the
> sample is much narrower.
Exactly. And every site you develop will also have skewed traffic,
and so not match the free numbers.
If traffic the sites *you* develop match TheCounter's fairly
closely, then those numbers probably are good enough for you. Many
others will not find that to be the case.
> If someone can point out a free stats service that has better
> stats, based on broader sampling of sites and visitors, then that
> will be good for all of us. Until then we must use what we have,
> and from what I've seen and the discussions I've read, these two
> big services still offer food for thought and helpful information.
They also give people a false sense of security and lead to
misconceptions about the 'average' user. They are poor numbers, and
should be approached with extreme caution regardless of how you
intend to use them.
In another e-mail:
> BTW, contrary to comments that StatMarket overrepresents small,
> hobbyist sites, their site indicates that the published stats come
> only from those sites using HitBox Enterprise, which is a
> commercial, paid tracking service. I doubt that many small,
> hobbyist sites pay for their tracking.
I can't vouch for others, but I didn't say 'small hobbyist sites,'
I said smaller commercial and hobbyist sites, 'smaller' being
relative. In my experience, large sites often handle their own
tracking. That may have changed, but I can't really say as I
haven't seen a client list or statistical analysis of such.
chris.kaminski { functionNewMedia :: <http://www.function.de> }
