Nutterum: thanks for bringing the topic of page/site personalisation up. It is a topic that has interested me and that I've been working in for years.
Note: I've been writing about this for years over at Cre8asiteforums for those who want to go read.
Note: my sites are evergreen info monetised via direct ad sales, affiliate referral, and AdSense as default filler; search traffic is an average ~40% of total, Google ~22%.
There are two main components:
1. visitor identification. There are broad and narrow workable IDs: broad targets groups based on referer (requires knowing something about the referring platform/site/page/market); narrow targets individuals (or very small groups, i.e. family), which I do via device fingerprinting and cross device usage.
Note: I prefer not to use cookies but others' requirements may vary.
2. personalisation implementation. My first implementation, back in 2007, targeting FaceBook users, were multiple broadly personalised landing pages: each with snippets of information, from around the site, of interest to that subgroup plus some common general info as filler. There were no site links in and SEs were blocked, however there were appropriate links out to the rest of the site. A little band of 'influencers' then posted appropriate mentions with links. It worked quite well, sufficiently so that I replicated for several other SM platforms and subsequently for top referring sites.
However, as I continued with segment targeting I realised that the 'problem' was that I was building rather a large number of pages to say pretty much the same things in different ways so as to better connect with visitors. And that it was because I was thinking in pages rather than in information delivery. Then I read Brian O'Leary's essay Context First in either late 2010 or early 2011:
...my idea in a nutshell is this: book, magazine and newspaper publishing is unduly governed by the physical containers we have used for centuries to transmit information. Those containers define content in two dimensions, necessarily ignoring that which cannot or does not fit.
Worse, the process of filling the container strips out context – the critical admixture of tagged content, research, footnoted links, sources, audio and video background, even good old title-level metadata – that is a luxury in the physical world, but a critical asset in digital ones. In our evolving, networked world – the world of “books in browsers” – we are no longer selling content, or at least not content alone. We compete on context.
I propose today that the current workflow hierarchy – container first, limiting content and context – is already outdated. To compete digitally, we must start with context and preserve its connection to content.
And I went YES!
It just made so much sense. Even before I thought about mobile, which was still only 3-or-4-years (smartphone) old. Initially in my mind context was demographic, was marketing segments. Within a month of research I understood clearly that context (in all it's myriad conglomerations and permutations) was the foundation on which content should be built/delivered.
Instead of thinking:
structure/semantics -> content -> presentation/behaviour for each target also remove structure/semantics from the 'page' such that a given URL is totally amorphous. Think instead:
context -> content -> structure/semantics -> presentation/behaviour.
My second implementation, starting back in 2011, was designed around device fingerprinting to identify returning visitors. The foundation idea was to convert as many first time visitors as possible to returning customers, then work at personalising those customers' experiences where feasible.
Note: I know that a URL is actually a subset (with URN) of URI but I view a dynamic personalised page as a recombination of URIs within a URL. You are welcome to use your own worldview.
In time (still in research stage) a site may become a recombination of URIs packaged according to personalised requirements. rather than move from formal page to page one would move through information chunks (I don't have the computer power to go more granular in real time). But that is probably 3-5 years off. Currently, I'm testing moving through information chunks in various ordering within each page.
A simplified walk through:
Possibility 1:
* visitor is not recognised arriving via SE or non-targeted referrer: default page for that URL is served, while visitor is fingerprinted and added to database.
* visitors actions, i.e. scrolling, cursor, apparent reading speed and interests, clicktracks are logged and associated with fingerprint ID.
Possibility 2:
* visitor is not recognised arriving via targeted/recognised referrer: a broadly semi-personalised page is served based on the referring page's title, heading, description, link's anchor/surrounding text and subheading.
* visitors actions, i.e. scrolling, cursor, apparent reading speed and interests, clicktracks are logged and associated with fingerprint ID.
Possibility 3:
* visitor arrives and is recognised, prior site interests checked against landing page content, and a personalised format based on context of person, device, history, etc. is served.
* visitors actions, i.e. scrolling, cursor, apparent reading speed and interests, clicktracks are logged and associated with fingerprint ID.
Note: I know broadly (P2) and narrowly (P3) which trinkets or widgets or thingamabobs or what-have-you were sold through affiliation or were checked out on my presell pages. So, I can better target each return. I know which ad pages for which products interested you the most over time so I know which coupons and offers to suggest first and second on given days in given seasons. Conversions increase significantly.
Note: only the default (possibility 1) is shown to SEs. All off site links resolve to the default unless (1) the visitor is recognised or (2) a non-SE visitor arrives via a targeted/recognised referrer.
Note: remember that I am NOT changing the URL, just the chunks and their order within, all done server side on the fly.
Important Note: Possibility 1 was implemented in 2011, Possibility 2 was added 2013/14 (held up while sites went responsive with server assist), Possibility 3 is in live testing now. That it works and works well is already apparent.
What is still unknown is how SEs especially Google will respond. So far so good (guess blocking is sufficient) but is early days...
And... let me tell you that context delivery is neither simple nor easy.
* identify kinds and degrees of user context awareness in specific scenarios.
Think personas on steroids.
* develop from the above a context awareness system that is replicable, reusable, and scalable.
* know and build within existing technical constraints while watching to see where they might be eased in n-time.
To add to the difficulty remember that context is not static but a dynamic process with a history. In fact much of what is termed 'context' is in reality a snapshot of a specific moment in context or context state; remember that context is a process, a flow, a series of context states over time. And that history, the previous context states influence future ones. Having fun yet?
To address the OP question:
what about SEO?, I have to say that the only SEO is that I've done my best to block SEs from the personalised pages so as not to confuse the poor critters and possibly damage the query standings of my non-personalised pages. I also have a query in through various channels to several SEs including G but have yet to hear back.
