Lapizuli - 11:11 pm on Aug 1, 2010 (gmt 0)
Great scenarios. That's exactly how it happens.
1. Rankings go up but traffic goes down
Isn't that an artifact of personalized SERPs? I.e., you're not really ranking as highly as you thought you were? Otherwise, the only thing I can think of is if an algo change put a different spin on a keyword and showed it to a different demographic, and they don't like what they're seeing. In which case, the high ranking should go down after a bit.
2. Traffic cycles between poorly-targeted and well-targeted, even though the total stays level
My total traffic has been pretty level, and yes, it has minor fluctuations. I don't know how to tell when my traffic is targeted or not. I'm at the mercy of conversions, and those are at the mercy of reporting indiosyncrasies and other "fluctuatables." Is there a way?
4. Traffic fluctuates wildly during the day, or from day to day
In recent months, my traffic doesn't seem to fluctuate wildly during the day, or from day to day. Today, August 1, it's normal for a Sunday. Note that I don't make major changes to my pages, as I mostly publish on sites I don't own.
3. Rankings are lost after a template change, even though textual content and URLs do not change
I think it would be easy to lose rankings after a template change. With Wordpress, some of the templates are so fiddly, and so are the plugins. For non-experts (e.g., me), one template's SEO features might result in double meta tags, for example. (And one might have to spend long hours figuring out what's going on and fixing it and thanking one's stars one noticed it before it became a problem for one.)
Back to bounces...
So Google knows it can't discern the meaning of a bounce and so doesn't try to use it to rank a site?
Interesting definition of bounce. It's slightly different from what I thought it was.
Bounce is a single pageview in a visit session.
Where a visit is defined as either closing the browser window or a session timeout of 30mins (inactivity on the site).
So if I arrive on your homepage and then leave and don't come back for 30mins or more that is a bounce.
So it's not counted as a bounce if you visit site - leave site - visit site again within 30 minutes. Surely they picked that 30-minute window of time because that's a meaningful way to distinguish types of bounces.
Though they seem to take a pretty unilateral view of bounce rate here:
Use this metric to measure visit quality - a high bounce rate generally indicates that site entrance pages aren't relevant to your visitors.
I'm just not sure I trust that. Surely they must already use relational data, or whatever, like returning visitor + bounce = good thing.
And they seem to say they do lots of usability testing with experimental search rats, whom they can ask, "So what were you doing when you bounced just now?"
And they love algorithms, and trying to read minds.
Maybe I'm being too Miss Marple, but I think maybe bounce rate isn't as noisy as perhaps it once was.
I have some pages with a highish bounce rate that rank well. I have others with a highish bounce rate that rank poorly. Either they don't take into account bounce rate very much, if at all, or they interpret it deeply.