Which is better, bounce rate or time on site?
 

I'm split testing some changes on a tube site which involve adding some content to some pages, including more banners to click on. So far the results are

Overall it produces more clicks to sponsors, which seems obvious since there are more banners;
On iPhone it increases time on site and lowers bounce rate, by statistically significant amounts;
But on Android, iPad, Windows and Mac it lowers time on site whilst also lowering bounce rate. I can't think why unless it's because users are more likely to click out to a sponsor and therefore spend less time on site.

Which would you prefer to have, lower bounce rate or higher time on site (but not both)?

Both have a different function:

In short:

Bounce: Visitor does NOT like the shit at the first sight. a.k.a. "shit in , shit out"

Time on Site: Visitor likes what he sees if the time is significant. If it is short, then he does not like most of the shit and leaves, If the time is long he likes the content. 15 minutes and more is considered good. Anything less is still " shit in , less shit out".

interesting to know more banners and links have created more interest, we´re just updating more links out on my blog...

thanks for the info :thumbsup

It seems like you should double check all your numbers and how you're running your tests, because your results do not make much sense.

if your numbers/tests are truly accurate -- I would segment down to google traffic and optimize for time on site. Its a better indicator of user satisfaction -- clicks can be baited and lead to a bad user experience -- but if the user stays on your site a long time, that's a pretty hard thing to fake or cheat for.

TOS is very important.

Depends on the type of site and the type of bounce. Specifically, for a tube site, "bouncing" to go to the sponsor site is ideal, but that will lower time on site. Bouncing to go anywhere else would be bad. If you log an event when they go to the sponsor site the event will be seen as another page view, so they'll no longer be considered a bounce. That's the ideal way to track IMHO. In that case bounces are bad (because they don't include going to the sponsor site). And in that scenario low time on site for people clicking on sponsor links is not a problem.

That's what I thought. Anyhow, with more data I seemed to get the results I was expecting, more click thru, higher TOS and lower B/R. Now for my next experiment....

Does a lower bounce rate not automatically lead to a longer average time on the page?

Greetings Beaver

Not necesarily although that's what you'd normally expect. In theory if you can get more people to click on an ad banner more quickly, that might lower TOS and bounce rate at the same time.

If you can split it up, there is always a certain % that will bounce quickly. Remove them and see where the median is for the remaining - that is your real TOS and your real traffic.