rotating SERPS to refine relevancy?

[rcjordan]

I can't get my thoughts together on this, but here's the basic recipe (some parts courtesy NBCi LiveDirectory):

Take the top 50 or 100 returns and present them in different rankings for the same search term. Score both the clickthroughs (small positive) AND the bail time (large negative). In the short term, basic SEO would rule but over time those sites that kept the traffic would float towards the top.

Bail time: the visitor returns to the SERP and clicks another choice or searches on the same term again.

[rcjordan]

<added>
toolman, I posted this just before I read your post about direct hit. In a way, they touch on the same thing from different perspectives.
[webmasterworld.com...]

[MaliciousDan]

I'd think any time someone came back to the engine it'd be a bad thing, it would mean they didn't find what they were looking for once they reached the site.

Personally when I go to a search engine I open any links that look relevant in a new window, there really isn't much they could do to track this information (correctly).

[rcjordan]

>any time someone came back to the engine it'd be a bad thing
>
Exactly. That site would get a large demerit, erasing the credit for the clickthrough and then some. It's using your user base as human editors. Individually, they're 'opinion' is not worth much, but collectively...

But what a load this would put on the SE to collect this "bail time" data!

[ggrot]

The bail time data could be stored in a cookie per user and then as just one number per site in the db...not a big deal.

The question is that almost all users will return to the SE eventually. How long until the bail time stops being negative?

[rcjordan]

>How long until the bail time stops being negative?
>
I've seen that number mentioned somewhere previously, I think it was 45 seconds.