It just hit me... after posting a couple of threads about John Mueller's recent Google Hangout, and also reading through various Hummingbird discussions, why, going forward, referrer data would probably not be helpful for serious SEO.
John Mueller's Hangout video is on YouTube here... Webmaster Central 2013-09-27 trt 1:08:38 https://www.youtube.com/watch?v=R5Jc2twXZlw
I posted about it with regard to... Google Authorship
and... Google's View of Product Descriptions
Particuarly in the latter, I got the sense, from John's comments following the mention of Hummingbird at about 23:50 into the video, that Hummingbird looks at the different aspects of content included not just on a product page, but on a site, and it tries to return sites that satisfy the intent of a wide range of relevant and helpful queries.
If so, an SEO's concern is not so much about what keywords are included on a page, so much as what pages are included on a site... and, beyond that, how helpful the pages and the site are to a broad range of users.
For a while now I've been pushing clients to build sites that anticipate a range of user intent and provide content useful to various kinds of users in different places in the buying cycle and in their engagement with the product... and perhaps even with the product niche.
What Hummingbird is about, I'm coming to understand... or at least I'm guessing from John's comments... is simply that it's better at evaluating the range of useful content that a site might return, and it increasingly will be sending users to such sites.
So it's no longer just a single page and its title satisfying a query... It becomes a whole site satisfying a range of users. With that kind of scope, the individual referrers are both less easy to specify and less determined by the landing page itself. Actually, not so different from what some of us have been preaching.
So, it's up to the SEO to guide in the development of what is a genuinely useful site... not just one that matches keywords. Over the long term, this is going to depart from 2 and 3 word phrases... and the long tail phrases that comprise the "query" will also become increasingly harder to define, as they will perhaps to cease to be just words. Intent will very much be part of the picture (as, eg, mobile search/local purchase vs online purchase). I can't imagine chasing keywords under such a situation.
I also can't imagine that it's going to be very hard to look at a page and describe what keywords the core page should rank for. The issues are going to be with the broader rankings and the site overall.
This is my guess, at any rate. I feel that overall, this will be an improvement in the web, which will become less a collection of content farms and more a collection of pages created with the user genuinely in mind.