A few months ago I came up with a working definition of "semantic SEO" that goes something like this.
For most of their existence what search engines have done have is match keywords in queries to keywords that appear in a web resource. They've done this with extraordinary sophistication, but search has been by and large played out on a linguistic stage.
Technological developments have allowed search engines to make sense of the things to which those keywords make reference as things – specifically by identifying each of those things with a unique URI. The stage has shifted to one of meaning, on which language plays a leading role but is no longer the main attraction.
But it is not so much the existence of these uniquely identified things that marks what a big leap this is for search engines – or what big changes it demands for SEO – as what the process of identification facilitates. Namely, that connections from one thing to another may be reliably made and mapped. It enables the search engines to understand the web as a series of interactions, and to use that context as it negotiates what it returns in response to a query, the negotiation itself changed from a simple catechism of questions and responses to a dynamic exchange where the search engines make sense of queries both temporarily and by their context.
Optimizing for this new environment, as one might expect, entails targeting entities rather than the keywords that describe them. To a big degree, though, this simply entails targeting something that's identified by a bunch of keywords, so one ends up targeting … keywords. This hardly represents a bold new approach to optimization.
What has changed is the ability – to be successful, the requirement – to target the interactions themselves, to be present at the right time in the right context. Participating in those interactions are now key: being seen in them, helping to shape them, initiating them, refining them. And in the course of charting these interactions, continually learning new lessons that can be used to shape and refine future strategies and tactics.
And that, in a nutshell, is "semantic SEO": optimizing for the relationships in which people play a role – real life relationships between people, real life interactions between people and things in the world, rather than only the relationships that can be described by keywords.
I've been using the quotes around "semantic SEO," however, and the reason I've been doing so isn't because I'm uncertain about whether the activities it describes represent a significant departure from "traditional" SEO. I think a semantic approach to SEO is in many ways a big departure from previous optimization efforts – though many of these "traditional" SEO activities must still be vigorously pursued.
Why I've started to stumble with "semantic SEO" is that I'm no longer certain "SEO" adequately describes the optimization target. If you're optimizing for relationships between things, what's to say the benefit of such optimization will be limited to how you appear in search engine results? Or that in a given instance the optimization effort is even really directed at search engines when the route to superior search engine viability is by means of another digital channel?
I'm relatively good at practicing what I preach, so two things have struck me as I've been doing things in order to "optimize for semantic search." First, that these efforts almost always impacted other things in the digital arena aside from search – and sometimes were obviously directed at areas largely divorced from search. Second, that many of the activities necessary for this type of optimization, being relatively new, fall into a kind of no-man's land within organizations.
For example, I've implemented or tweaked the implementation of Open Graph more times than I've had hot breakfasts. By staying on top of this code you control the way the snippet for an article or a video or whatever it is you're sharing appears in Facebook, increasing the chance it will be shared, discussed, emailed, and ultimately linked to from meaningfully spidered environments.
At a deeper level, finessing this code means you're better defining and more clearly disambiguating all the nodes that appear in the vast graph that includes both Facebook and the things outside of Facebook to which it connects, things that can use Open Graph to stand up and be counted. All leading to miraculous things like Facebook users more frequently being shown a more relevant ad.
So the benefit to search is real and tangible, but the shorter distance in terms of awarding a big winner here is obviously a company's social media efforts, rather than their appearance or ranking in organic search results.
Who ends up implementing Open Graph? From my experience it's people like me, "SEOs," that have taken up the mantle of things like Open Graph that are clearly social-media facing.
That's not a complaint, and – to finally get to the point – it makes total sense as structured data markup for search is based on exactly the same model as structured data markup for social media networks, which in turn is conceptually identical to product data provided in XML feeds for Google Product Listings.
That is, identifying entities and providing information about them in the form of structured property/value pairs.
I just happen to work a lot with structured data markup, so I recognize it readily as an obvious way of identifying and describing entities, but it is by no means the only way – or even the most important way – of doing something that ripples across seemingly separate realms.
What's notable is that those is that those ripples run through many disciplines, and that more and more there's a class of digital work that doesn't truly belong to any one of them. It is a world of things that is, or is becoming, the tuck and trade of social media, advertising, analytics and – finally – search.
The entire existence of successful social media networks is based on disambiguated verified identities interacting with other identified things. Innovations like the Open Graph protocol or Graph Search are only Facebook's latest efforts at harnessing the potential of the world of things, which has always been Facebook's world.
A social media network provides probably the clearest example of how the interests of search and social merge at a structural level: the multiple Rich Pins supported by either schema.org or Open Graph markup. Pinterest's logic was spot on here: why create another protocol to solicit the same property/value pairs that pages are already dripping with?
Semantic analysis – understanding the referred to things that are lurking beneath the surface of content – has a little bit of relevance in the world of advertising because, at the end of the day, figuring out underlying meaning is pretty much all a company named Google cares about. If you can understand what things are being referred to in a query, or understand the things that being referenced by a web page, you can deliver better ads. This doesn't just mean making more money off desktop search, but not getting slaughtered quite so badly in mobile with its smaller real estate and lower click through rates and mobile devices' general disdain of interruption.
Analytics, always a key player in the all digital marketing endeavors, has extended its ambitions from counting individuals as they visit a single web site to understanding their behavior make their way through digital and physical worlds. Sentiment analysis, for example, produces insights that are light years past "keyword phrases harvested from social media that include my brand name and the word "sucks"."
Google's next-generation measurement tool, Universal Analytics, explicitly escapes the confines of web resources and web-only identities. Universal Analytics' focus is on the behavior of visitors – actual visitors. UA isn't just visitors on a website identified by a cookie, but actual individuals who browse to multiple websites, are exposed to ads, use apps on their phones, play games on consoles and make purchases at brick-and-mortar stores.
So search engines have now vastly improved their ability to identify and interconnect things, but search isn't the only part of the digital universe undergoing this sort of transformation.
Where these paths lead, the point that's at the center of the every graph found in search, social, advertising or analytics, is plain. It's the individual. The searcher, the member, the viewer, the visitor.
So when you're optimizing for relationships, you're optimizing for the appearance of an individual as they move in and out of digital and physical environments, as they create even more connections between more things over time. More descriptions. More data.
You're not optimizing for search engine rankings or social media engagement or the best ad at this level, but for the overall digital presence of a brand in all these situations and beyond, for the connections that are possible now and are yet to be made.
I think that the increasing trend we've seen towards SEO subsuming other types of digital marketing is a logical outcome of the structured world around which these efforts are coalescing. And the uncertainty of search marketing's new identities, the tenuous removal of "SEO" from taglines and business names makes sense too: much of the work that needs to be done is for the search engines alone, but more and more required work – work having to do with the description and identity of things – isn't restricted in scope or benefit to a single arena.
It's quite a scream that just a few weeks ago I grumpily observed that there was such a thing as SEO, and that pursuing things like social media engagement or better customer experiences was a hell of a roundabout way to influence one's visibility in search. I still think that's true: my point here is that there's not things we should doing instead of SEO, but things we be should be doing that transcend SEO.
What sort of things?
This article is ostensibly a write-up of my 2014 SearchFest presentation, "Approaches to Structured Data for SEO and the Emergence of Digital Presence Optimization" – or its far less pretentious and much more accurate alternate title, "How to Say Stuff About Things that Everybody Gets."
In that deck you'll find some of these approaches enumerated, and you'll find that deck embedded at the bottom of this article. I won't go over all of the points that I make there, but some of those points might be improved by some further explanation.
Think like a contemporary data consumer
That is, a beast that's on the hunt for information about identities and descriptions. Feed it with property/value pairs. Lavish it with triples.
You can lean on the Thing class page of schema.org as a pretty good crutch to help frame how to think about things in the context of data consumers. Don't think "what's a good <title> tag for this page," but "what's a good name for this thing?" What's image represents this thing? What's a good description of this thing?
The term "data consumer" itself is necessary: framing a search engine or social network or advertising network as the data consumer too narrowly binds the things on which data is being provided to a specific realm.
Determining what things you're dealing with is a critical first step
I've used "thing" instead of "web page" above because a web page is just one example from one of the two broad categories of things which you should be identifying and describing – categories which are not mutually exclusive. It's important to know which sort of things you're dealing with, as that helps identify the sort of properties for which you should be supplying data.
All things found on the web are some sort of resource type: a web page, an image, an audio recording, a video, a data file. Each of these resource types have properties that are understood by diverse data consumers and should be provided to them.
What's described or depicted or quantified can also represent something out there in the real world, be it a physical entity or a concept, that itself has properties entirely separate from the web resource that describes it.
In each case what type of things these are helps determine what sort of information is most useful to data consumers, the most useful way of representing data for machine consumption, and what sort of data provision to avoid. Both an image and an audio file can have a size in bytes, but only the latter can have a bitrate value. An island has a location – best described by geographic coordinates – but a toy being sold online does not.
I think approaching things in this manner helps make the concept of "content" more useful. From a structural perspective "content" is a combination of a resource type and the things that it references.
An HTML document is "content," a tweet is "content," a Facebook post is "content," a video is "content," a .CSV is "content," and they cannot all be used in precisely the same way. But in any case the container is useless unless it contains something useful, and "content" that is typically thought of as "thin" is typified by representations of things that aren't all that meaningful, or aren't very much connected to other things – like good old fashioned advertising.
The holy trinity of data sources
While the obvious connection between, say, editing a Wikipedia entry about a brand and how that brand is represented in Knowledge Graph or Snapshot results makes mucking in with Wikipedia or Freebase or Wikidata seem mostly an activity in aid of search engine visibility, the wash thrown up by these data supertankers extends further than search.
Enterprise-level recommender systems and document classifiers lean on exactly these data sources, so what there's a direct relationship between, say, what appears in Wikipedia and what product might appear in an ad, or what articles might be featured in a news vertical.
Merging entity and keyword research
What "entity research" brings to the party that keywords can't readily provide is information about edges. Traditional keyword research mostly, kind of, sort of covers what's uncovered about entities if they're viewed through a property/value lens, but not always the relationship of the objects being researched to other objects.
Keywords are good at saying what attributes things have, or identifying other things that they're the same as, but not so good at saying what things or related to, or even who the parent or a child of a thing might be.
Put an entirely different way, perhaps it's useful to stop thinking of "keyword research" – an activity specifically undertaken to in aid of organic and paid search optimization – altogether. It might be more useful to start approaching keywords as being (as indeed they are) an attribute of an entity.
Whether it's at the resource level (e.g. "web page," "video") or at the representational level (e.g. "acme blue widget," "king henry viii") core research into the thing being described – its properties, its relationships to other things, its appearance in context – creates a sort of data profile that can help shape how a thing, a thing like a brand or a product, appears in multiple digital settings. In organic search, paid search, social media networks, digital advertising, data repositories, news streams, email messages….
As described above, verified human identities are the bedrock upon which social media networks are built.
Verifying identities, along with engaging in interactions in verified identity networks, are themselves types of data structuring. You're unambiguously identifying yourself and exposing your relationship with other entities out there in the world. That sort of structure is baked into the very fabric of social media networks. An Oxford definition of "network" as a "group or system of interconnected people or things" makes the point perfectly.
The graph perspective? Once you've identified yourself as a node, it's all edges all the time out there in social media land.
The brave new world of DPO
There's more examples of emergent, cross-disciplinary optimization activities in the deck embedded below, and some very colorful pictures: you should check it out.
While (hopefully) a brand's visibility in search or effectiveness in advertising or traction in social media conversations might be improved by taking some of the approaches I suggest, at the end of the day does it matter if we lump these approaches together as "digital presence optimization," or just tackle the tasks as they come?
While the label itself doesn't matter, I think that understanding how changes in technology have brought a number of digital environments – and digital marketing disciplines – together is fairly important. Optimizing for the world of things by pursuing a set of related activities is more efficient than taking on each digital environment in a piecemeal fashion, and produces more consistent representations of a brand across these environments.