I was doing a periodic look through some sites for stats on screen resolutions (interesting in itself: 1280px width accounts for 35% of traffic and only about 15% is smaller than that).
Anyway, I decided to look at something I'd never checked in Google Analytics: bounce rate versus connnection speed over the last 5000 visits.
I wasn't that surprised to learn that bounce rates tracked exactly with connection speed - 49% for dialup users and dropping into the 30s for T1 and OC3 users. I've known for a long time that the site in question would be more user friendly if a lot of cruft were cut.
Then I decided to get a bigger sample and go back another month and look at it over 15,000 visitors. Strangely, the pattern then appeared random. ISDN users, a statistically insignificant group (10 visits) bounced at 80%. But after that, it looked like this:
Unknown - 49.54%
OC3 - 48.94%
Cable - 48.24%
DSL - 47.33%
T1 - 47.25%
Dialup - 44.38%
Here the bounce rates are clustered around 48% for all high-speed access and down around 44% for Dialup. Of course, the massive numbers of unknowns could easily skew things one way or another if there were a systematic bias toward some connection methods getting seen as unknowns.
Just makes me curious what other people see? Do your stats show a bad user experience for dialup users based on bounce rate, or are they more or less random?