How site speed affects bounce rate and pages per visit. Or doesn’t.
A problem with a load balancer means that our site speed has been much lower for the last week or so. This is obviously pretty worrying, but it did give me the opportunity to look at the knock-on effects.
The first graph below shows our average page load time over the last month. The effect of the issue with the load balancer issue is stark, with page load time increasingly significantly (I presume the recovery at the weekend is due to less demand).
I expected that this would have an equally dramatic effect on the behaviour of our site visitors. The two obvious measurables to me were bounce rate and pages per visit. I presumed that as the average page load time increased, we’d also see bounce rate increase and pages per visit decrease. But, as the graphs below show (covering the same period of time), there’s been no discernible effect at all.
I don’t think this means that site speed isn’t a problem – I’m sure that if visitor satisfaction were measurable we’d see a corresponding dip – but I do find these (non-)results surprising.