This is my first "Outside the Google Box" topic as per Voyaging Outside the Google Box
[webmasterworld.com]. We'll see how it goes. Backstory:
Way back in the long ago I read The PageRank Citation Ranking: Bringing Order to the Web
[ilpubs.stanford.edu] [PDF 481KB], January 29, 1998. Yes, that PageRank paper.
A few of us looked at it, not as 'how Google does what it does' nor as 'how to beat Google at their own game' but as a simple way to illustrate mathematically or graphically the flow into, within, and out of a site. If, for instance, and to keep it simple, you have n-people a day being referred by some external site to a page and you gave that number a value and then mapped the potential dispersal from that via it's links to other pages, and you did that for every page you could distinguish potential 'driver' pages at one end and 'dead pool' pages at the other of the spectrum.
It was initially fascinating because one could actually map out and test a site architecture before publishing a single page. One could, as it were, get one's internal linking in order. And one's marketing. Out of the Box:
That flow mapping was unusual, I know of perhaps a couple dozen of us that played with it and discussed the ramifications and a handful that were still doing recently. Why would I want to model flow for sites that have been pretty much finished for over 5-years?
Because over time the modelling switched from potential to real; from theoretical gaming inputs to real live traffic inputs. I can map each visitor from each referrer to each landing page. Further, via click-track I can map each visitor link by link page by page until after they leave. And how they left.
Over time there is a traffic flow pattern for each and every referrer. Tied with conversion rate mapping this allows one to put a value on each referrer and so each landing page. Also to know ahead of time what a new visitor from a specific referring link is likely interested in, looking for; the beginnings of a basic personalisation.
By going back to a referring back-link's page one can get the page title, description, headers, link surrounding text, link text, etc. and determine the information scent context that brought them. Using that and the subsequent click-track flow and conversions one can get a sense of whether they found that for which they came or if something(s) could be added or improved. No need for Google query ranking, let the visitor tell the story and improve referrer personalisation. And their conversion rate.
Note: what one does with such personalisation information sets the analytics analysts apart from the analytics gatherers.
I can mathematically explicitly determine the value of each page relative to every other. And that is a critical bit of intelligence when it comes to selling ad space (and for affiliate presell). It allows me to divide a site into three major each with three minor (for a total of 9) pricing/value strata. And I have the data to back up my rates, which is something few other publishers or ad networks can (or are willing to) do and permits confidently offering guaranteed minimums of both traffic levels and conversion rates.
Note: it produces the loveliest colour map graphics for C-level gratification.
Obviously, I sympathise greatly with ergophobe's How "Direct Traffic" May Be Undermining Your Marketing
[webmasterworld.com]; knowing visitors' referrers is a significant input for visitor context, back-link and landing page values.
Another aspect of mapping traffic flow in the context of the referrer is that knowing the meta and not-so-meta data of the back-link page I can search for similar pages and so sites that may be worthwhile enticing to also link. Very highly focussed link building.
Note: I have only bought a few very short term links on first going live with a site (if you need to ask why...) and I have never actually asked for a link. I work quite indirectly; best if they believe it was all their own idea.
One thing I do do is that for every back-link above a certain value threshold I thank the site (as personally as possible) formally by snail mail. It's such an unexpected behaviour that they are shocked and positively affected. A few have actually posted the letter online for additional marketing value. Besides the goodwill, or perhaps because of it, thanked sites rarely remove the links; such sites' (to me) link rot is a fraction of typical. Cheap at twice the price.
Flow mapping turned out to have a great many benefits (not all mentioned here) and is still part of my analytics today. Thanks to the PageRank paper if not to PR itself, which I have pretty much ignored for pretty much ever.