Dynamic Semantic Clusters: Where user experience begins

Users don’t read websites the way they read books. They read clusters of pages from many sites that they assemble for themselves or with others they trust; a hunt and gather process that shapes their experience. To help construct a compelling user experience, we content creators need to think outside the box of the website. We need to think about Dynamic Semantic Clusters. One way to do that is with an approach to content design that I call “Every Page is One Page”.

When we think about the user experience of our websites, our biggest mistake may be assuming that it’s our site the user is experiencing. In reality, the user is often experiencing a dynamically assembled cluster of individual pages from many different sites, rendering our carefully crafted site-based user experience moot.

Like most people, my web experience usually begins with Google. My approach is to hold down the Ctrl key and click all the interesting links on the first page of search results, opening each one in a separate tab. I then go through those tabs one by one until I find what I am looking for, or find that I need to do a new search. My user experience, which I have constructed for myself, is not formed by a single site, but by a cluster of single pages from many sites.

This is the kind of behaviour that information foraging theory would lead us to expect. Information foraging, which Jakob Nielsen calls “most important concept to emerge from Human-Computer Interaction research since 1993”, shows that human information seeking behaviour follows the same pattern as the hunting patterns of wild animals. Animals hunt in patches, and must decide when it is better to hunt out the current patch and when it is better to move to a new patch. As Nielsen comments:

If getting to the next patch is easy, predators are better off moving on. No need to deplete all the game in the current patch; once their next morsel becomes a bit difficult to find, they can move to richer hunting grounds. On the other hand, if it’s difficult to move (say, if they have to cross a river), they’re likely to hunt each patch more extensively before going to the next one.

If websites are the patches in which readers forage for information, modern search engines and social media, which now do a good job of surfacing relevant high-quality content, make it easy for readers to move on from one site to another. Thus readers are more likely to pluck pages from multiple sites rather than reading every page of one site.

Dynamic semantic clusters

In fact, modern search and social tools do more than simply make it easy to move on. They dynamically assemble a cluster of the best pages on a subject. It’s as if a fox could type “rabbit” into a search engine and instantly have a cluster of the juiciest and slowest rabbits assembled for him.

Given these tools, information foraging readers do not set out to visit websites; they summon a cluster of semantically relevant pages. If they are searching for products and solutions, those clusters will almost always include pages from competing companies. Even if they don’t make that choice deliberately, Google will often do it for them. Google, after all, is set up to return diverse results for a specified subject, and if competing organisations offer content on the same subjects, they will be part of the result set that Google creates.

Be careful, therefore, if you are thinking of Google simply as the vector that brings the user to your site so that their user experience can begin. Google is the engine by which the reader creates their own experience by clustering many pages of potential interest. That dynamic semantic cluster itself is where their experience begins.

Search engines are not the only mechanism by which such semantic clusters are assembled. Many people will pose a question to their friends on Facebook or Twitter, some of whom will respond with links to web pages on the subject. That collection of responses also forms a cluster of pages which the user then experiences as they seek the answer to their question. If the question is, “what is the best X for me”, the cluster will include pages from all the major manufacturers of X. In fact, product recommendations from one person in a social group may spur others to recommend their own favourites, thus a given user’s dynamically created experience will consist of a cluster of pages from competing vendors.

This is how we really experience the web — not as a library of sites from which we select a single site to be read like a book, but as a series of dynamic semantic clusters that we create for ourselves, and that we create for and with our peers. I call these clusters “semantic” because the cluster is assembled based on the meaning of pages and their relevance to the semantics of the enquiry, not to their ownership, origin, or location. (It is precisely to enhance the quality of semantic clustering that we want the web to be “semantic”.)

A website — a collection of pages with a common design under a common domain name — is an institutional cluster. But readers are not generally interested in institutional clusters. They want a cluster of pages relevant to their current interest, independent of their origin or ownership — a semantic cluster. Even if an institution is the object of the reader’s interest, they don’t just want to hear what the institution says about itself, but what others have to say about it. On the web, no one orders off the menu; we always choose from the buffet.

The decline of the homepage

Are your users experiencing your site as a site, or are they experiencing your pages as part of a dynamic semantic cluster of pages returned by a search engine or recommended through social media channels? It’s fairly easy to tell. Simply compare the number of users who hit your homepage and then use your search and navigation to those who arrive at an internal page via a search engine or social network. The latter are experiencing one page as part of a dynamic semantic cluster of their own creation. They are not “on your site”. In many cases they are not paying much attention to what “site” they are on. They are on a page in the semantic cluster they have created for themselves. That is their experience.

Dynamic semantic clustering is an increasingly important feature of the web, and one that many sites and apps are actively exploiting. Flipboard, Twitter, and Facebook are all essentially storing search criteria established by the user and executing that search repeatedly to update a semantic cluster of content. In this context, the homepage is not home for most readers anymore. Home — the place they return to when they get stuck or lost or bored — is the Google search they did, the Twitter feed they followed, or the flipbook they were flipping through.

According to the The New York Times Innovation Report, the Times online edition has seen a steady decline in visits to its homepage over recent years. “Only a third of our readers ever visit it. And those who do visit are spending less time: page views and minutes spent per reader dropped by double-digit percentages last year.” And the Times is regularly outperformed by news aggregators like The Huffington Post who are repackaging its own content. Home is no longer where the heart is.

Of course, readers could use Google and social media to find individual sites, just as they once used the book review section of the newspaper to find individual books to buy. Sometimes they do just that. But for most readers, the dynamic semantic cluster is not simply a place to look for a site to read, it is, in itself, the experience they want, the way they choose to access content: a diverse selection of pages on a particular subject, not a site by a particular organisation.

There are a number of reasons for this, several of which I explore in depth in my book, Every Page is Page One: Topic-based Writing for Technical Communications and the Web. To hit just the highlights:

  • The sheer scope of the web ensures that you will almost always find multiple answers to any question. No one site can match that.
  • The web creates a near-ideal information foraging environment, making it very easy for a reader to move from one page to another, regardless of which site those pages are on.
  • The web gives readers access to the long tail of information — content so obscure no one book or site could contain it all, but which taken together meets an ever growing proportion of people’s information needs.
  • The web gives readers access to people with experience as well as to people with credentials. There is a huge comfort in hearing from someone like me who solved the same problem I am having now. Vendor’s websites seldom provide that; dynamic semantic clusters of web content usually do.

The new economy of trust

But there is something deeper at work here, a change in the way that we confer trust. This manifests itself in the rise of services such as Airbnb and Uber, which create trust and enable commerce between ordinary people.

In his New York Times article, The Evolution of Trust, David Brooks explores some of the factors that have contributed to this new social form of trust, and concludes:

The result is a personalistic culture in which people have actively lost trust in big institutions. Strangers don’t seem especially risky by comparison. This is fertile ground for peer-to-peer commerce.

There is a similar change going on when it comes to how people decide to trust content. The old model for finding content was: first find an authority, then ask your question. The new model is: first ask your question, then, if in doubt, establish the authority of the answer. In other words, in the past, we would begin by finding the authoritative book or organisation, then ask our question of that authority alone. Today, we simply Google our question out to the entire web, then sift through the results to find one we trust. David Weinberger makes a relevant statement in his book, Too Big to Know: Rethinking Knowledge Now That the Facts Aren’t the Facts, Experts Are Everywhere, and the Smartest Person in the Room Is the Room (Basic Books, 2011). He says:

We seem to be making a cultural choice—with our new infrastructure’s thumb heavily on the scale—to prefer to start with abundance rather than curation. Include it all. Filter it afterward.

How we decide to trust an answer has changed as well. Whether content comes from the official company site is of far less consequence than the personal reputation of the person who authored it. That reputation is based on the same mechanisms that drive trust in Uber and Airbnb: likes, endorsements, user reviews, and five-star ratings. On the technical Q&A site, StackOverflow uses a sophisticated user rating system for applying both questions and answers to surface the best content. It is the aggregated opinions of ordinary readers — not the imprimatur of editors and publishers — that establishes that reputation. In other words, trust is socially mediated, not invested in institutions.

If this is our trust system, we have little interest in confining ourselves to individual sites, particularly commercial sites. A world in which individuals are more trustworthy than institutions is one in which web pages are preferable to websites. Dynamic semantic clusters of socially validated content make much more sense in the new trust economy.

The storyteller’s dilemma

Novelists often like to set their work in a country house, on ships at sea, in jungles, on mountain tops, or in times past — any place they can isolate a small group of characters and allow them to interact with each other without outside interference. This isolation is often necessary to producing a satisfying story.

In the real world, we are so connected to people, things, and information that most plots could be resolved in seconds. In any emergency, a crowd of trained and anonymous first responders would descend on the scene, save the day, and clean everything up, leaving our heroes nothing to be heroic about. (No wonder most contemporary drama on TV is about first responders. Only they get to be heroes.) Our connectedness makes it easier to live everyday lives, but harder to tell clear and satisfying stories.

But telling a clear and satisfying story is vital to content marketers, content strategists, and information architects. When we try to tell such a story, we often try to construct the same kind of neat isolated world that the novelist uses to tell their story. We create a website. And often we create it as if it were a lifeboat floating on a boundless ocean with never a sail in sight.

The problem is, websites are not isolated, and cannot be isolated. They exist in the riotous confusion and ceaseless commerce of the web, with readers dropping in and out constantly, often with little idea of what site they are on at a particular moment. The story we are telling ourselves when we create a “website” as an artefact designed to stand alone is pure fiction. This is not how readers consume content today.

There are two kinds of stories that we are telling when we create a website this way. There is the story we are attempting to tell the reader — a story that all too often depends on the fiction that the reader is visiting our website the way they would once have read a book or perused a catalogue. The second, and more insidious, is the story we are telling ourselves: the fiction that we can create an isolated world in which the reader will have the experience that we design for them, unadulterated by anyone else’s content.

The only organisation that might be able to come close to pulling that off is Facebook. But most of the content on Facebook is not Facebook’s content. It is user content. If Facebook has the power to hold people’s attention, it is because it dynamically clusters content from people that its users know and care about. Can Facebook manipulate that for its own ends? Of course it can. But the experience that it manipulates is not one it creates, but one that is created by its users through their own posts and photographs. Facebook does not craft a little world, it opens as a bazaar, lets everyone in, and makes its money on concessions.

The role of the page

In my book, I present two diagrams illustrating the fiction we tell ourselves about how our content appears on the web. The first represents a website as a tight integrated cluster of content — a little world all to itself in the vastness of the web. This is how we think our content appears on the web.

Diagram depecting a big web circle and how we think of content in it as a small grid of pages.
Figure 1. Diagram depicting a website as how we conceptually think about them, as tightly packed clusters of pages (by URL) locked into position in the expanse of the open web. (From "Every Page is Page One: Topic-based Writing for Technical Communication and the Web". Used with author permission.)

The second shows our content as individual pages scattered all over the web. This is how our content actually appears on the web.

Diagram depecting a big web circle and how pages are available to us.
Figure 2. Diagram depicting a website's paged content as it actually is, dispersed through the vastness of the open web. (From "Every Page is Page One: Topic-based Writing for Technical Communication and the Web". Used with author permission.)

How so? Because this is how the web works. It’s not a static publishing media, but a vast dynamic content filter that dynamically clusters content according to how the reader searches and according to how the web’s billions of readers search, surf, tag, comment, and like content every day.

How does the reader’s experience of your content begin? Does it begin with them typing your homepage URL into their browser? Not usually. Most often, it begins with a Google search or by following a link on Twitter or Facebook. (This is why you invest in SEO and social media, right?)

And where do they land when they follow these links? Seldom your homepage. Even if your homepage is your most viewed page individually, its numbers are almost certainly dwarfed into obscurity by the aggregate of all hits on other pages. Readers land on one individual page somewhere on your site. But even then they are not really on your site. It’s a good bet they don’t know what site they are on. Where they really are is on your page.

Even when readers do start on your homepage, they often go straight to the search box. They are not looking for the site, but for one page on the site, for one piece of information. The experience they seek is the experience of a page, not the experience of a site.

Living in the semantic neighbourhood

Pages do not stand in complete isolation, of course. They have surrounding and associated pages, which readers may well want to view for background, follow-up, elaboration, or alternatives. But don’t assume that the reader’s idea of surrounding pages is the same as yours. To the reader, the surrounding pages are the other pages in the dynamic semantic cluster they are using. They are, in other words, the pages that surround the current page semantically, because they have the same or closely related semantics. The surrounding pages are the other results in the Google search, other recommendations from Facebook friends, and those pointed to by other tweets in the Twitter hashtag they are following.

Other pages on your site are not part of the reader’s current semantic cluster unless you do something compelling to make them part of it. Creating a relevant semantic cluster is work, and if you can do that work for the reader, in a way that they find satisfying, then you can keep them on your site — not because they have chosen it as a site, but because it happens to provide a semantic cluster that is relevant to their needs and interests.

This is how the web works, and why the web works. It creates clusters of content dynamically in response to the semantics of reader enquiries. This is what the most successful web companies do, from Google to Twitter to Facebook to LinkedIn to Flipboard and Pandora and NetFlix. They are, literally, the dynamic semantic clustering of content from many sources.

The user’s experience of your content, therefore, is not the one that you have so painstakingly constructed from the top down within the artificial boundaries of your site. It is whatever part of your content appears in the dynamic semantic cluster that the reader has created for themselves in the moment. It is not the experience of the site. It is the experience of the page. Any hope you have of keeping the reader in your cluster begins not on your homepage but on the page they are looking at right now.

Towards a dynamic semantic content strategy

A content strategy for addressing readers who create their own dynamic semantic clusters must begin from a different starting point than one that simply seeks to build a web site.

One of the first questions to ask is how large a role a single site should play in that strategy at all. Part of the strategy of a site may be to attract readers from all over the web by inserting itself, by fair means or foul, into the dynamic semantic clusters that they create, and then try to keep them locked within the site — a kind of Venus flytrap in the content jungle.

Unlike the Venus Flytrap, however, you have no way to actually trap the reader in your site. The web is really a collection of pages, and it’s only through a common design and internal links that a collection of pages becomes a “site”. They have no cohesion beyond this, and no power to hold or compel the reader to stay. Some sites do try, of course, especially in the mobile world where they try to force an app down your throat, but the reader can always escape. You can build fences on the web, but only to keep people out, not to keep them in. The rose, rather than the Venus Flytrap, is a better example to follow — allow the insects to come and go, carrying your pollen out to the world.

It is your message that matters, after all, not your site. Getting your message out to many sites across the web may be a better way to ensure it gets included in the dynamic semantic clusters being built by your target audience.

But if you do want to capture the reader — and there are many excellent reasons to do so — you have to make sure that every page in your site actually works individually to accomplish this goal, since every page on the site is a potential landing page. That is, make sure that every page is independently capable of catching and holding the reader’s attention, and that it guides them naturally to your other pages along the lines of their current interest.

This isn’t done by putting a home button on every page, or by pushing the reader back to your navigation bar or table of contents. The reader is not interested in your site, or its organization, but in the subject they are reading about. To keep them, you must help them follow their interests by providing topical connections to pages that have an immediate semantic relationship to the current page.

I call this “Every Page is Page One”.

Every Page is Page One

Every Page is Page One is an approach to content design that begins with the assumption that readers can enter your content set from anywhere and can and will move around in it — and in an out of it — in any way that strikes their fancy. It assumes that readers will often encounter individual pages in the context of a dynamic semantic cluster, and that they will continue to navigate along semantic, rather than organisational lines, after they arrive on it.

Every Page is Page One is, in other words, an observation about how people seek and use content today, and an attempt to deal with the consequences. Among those consequences:

  • A content audit is no guide to the nature of the user experience you are providing. It may be a useful way to detect and remove the dross on your site, and perhaps to find obvious errors and omissions, but it tells you very little about how visitors are experiencing your content, since they are largely experiencing it one page at a time. As well as, or instead of, an audit of your site’s content, consider an audit of the user’s content experience when they look for information on subjects that matter to you and to your customers.
  • Your competitor’s content is part of your customer’s content experience — it appears in the dynamic semantic clusters they create for themselves. So does the content created by your users and by your critics. You cannot change the fact that your readers experience this dynamic semantic cluster — but you can influence where and how your content appears in that cluster — and whether it performs well as a member of that cluster, in comparison and contrast to the parts of that cluster that you don’t control.
  • You cannot control the vocabulary in which the discussion about your products and services takes place. It is nice to imagine that developing a taxonomy across your company will banish all confusion and ambiguity and put an end to all content navigation problems. And while that is dubious at best — language is not nearly so neat as this approach supposes — it is meaningless in the context of a dynamic semantic cluster which was assembled using vocabulary of the reader’s choosing. If you want your content to feature prominently in that cluster, and if you want the reader to select it from all the other items in the cluster, it had better speak with the reader’s vocabulary, not the corporation’s. As Jakob Nielsen advised in his article on information foraging, “Don’t use made-up words or your own slogans as navigation options, since they don’t have the scent of the sought-after item.
  • Whether a particular piece of content gets found and used depends less on where and how it is featured in your site and more on how well it responds to the needs of actual users, and the semantics of the queries they use to assemble the semantic clusters they use to meet their needs. The information scent of your individual pages matters more than the information scent of your site. Indeed, the information scent of your site is really nothing more than the information scent of its pages.
  • You are creating a collection of pages. But the reader is interested in the pages, not the collection. Every Page is Page One. The emphasis in planning and in writing must be on the page, on how well the page works for the reader and how successful the page is in inserting itself into the dynamic semantic clusters that the reader creates. And if you want the reader to act, you should provide the means for them to act on each individual page.
  • Your success in retaining the reader beyond a single page will depend on how effectively that page works as a node or hub of a semantic cluster related to its subject matter. This is not about navigating your site. It is about navigating your subject, and the subject that the reader cares about.

The web is not a publishing medium. It is a colloquium. A content strategy for the web is not a publishing strategy; it is a strategy for a colloquium. It is not about creating and living in a walled garden, but about participating in an open bazaar. Your reader’s experience is the experience of the bazaar, not of your walled garden. Your content strategy must be a strategy for the bazaar, for the colloquium, for the dynamic semantic cluster.