Schema App’s CEO Martha van Berkel interviews Aaron Bradley from Electronic Arts on the topic of Schema.org and the Enterprise.
Aaron Bradley shared many insights with his key take-away being:
What the introduction first of Rich Snippets and then later of the Google knowledge graph really introduced into the search engine world is, as encapsulated by Google’s famous marketing phrase for the knowledge graph “from strings to things”, is that prior to the search engines weee essentially indexes of documents. So when you search for something, if you were looking for blue widget, you really weren’t getting results for things that were blue widgets, you were getting webpages that described or use the string the best blue widget.
Whereas now, and especially since the interaction of the Google knowledge graph but also fueled by structured data, what we now have is a web of things kind of one of Tim Berner Lee’s great visions for the web, so that rather than being an index of documents, Google is increasingly becoming an index of things, and facts related to those things. You can’t possibly have that without those sort of linked data technologies working in the background including ontologies, schemas, taxonomy solve that sort of thing.
To get started in this journey in the Enterprise, Aaron suggests three steps.
- Start with fundamentals. If you’re a webshop, look at search engine optimization with schema markup.
- Then look at the possibility of bringing that data into your analytics (Semantic Analytics).
- Connect all of the data points across the enterprise.
Martha: Hello and welcome to Schema Stories! My name is Martha van Berkel and I’m the CEO at Schema App, and I am delighted today to be joined by Aaron Bradley. Welcome Aaron!
Aaron: Hey, how’s it going Martha?
Martha: Excellent. Well, not only am I excited because you’re Canadian and we get to geek out on structured data together, here over the 49th parallel but I’m also excited because you bring a really unique perspective in your role and within the enterprise. Let’s start by having you tell us a little bit about yourself.
Aaron: Sure, I often say I’ve had three careers, that inevitably led me into the place that I am, I was a technical services librarian for about a decade. Following that I was a web designer for about a decade and now for slightly more than a decade I’ve been a search engine optimization specialist, though like most SEOs, my role has enlarged and deviated. So, to encompass other realms of digital marketing and particularly the creation and distribution of digital content across platforms, and right now as I have been since 2014, I work for Electronic Arts where I’m a web channel strategy manager and have a great team that work on these issues alongside me.
Martha: Excellent! Tell me a bit how you got started like within those kind of three careers, when did you first get started using schema markup or exploring structured data?
Aaron: It was in and around 2008 when I was working for a jewelry e-commerce site called ice.com and I first encountered Martin Hepp’s Good Relations ontology for e-commerce. I was very excited about it I think maybe because of my librarian bent and having worked with controlled vocabularies in the past I saw this as an interesting and exciting way of informing search engines and other data consumers about what products were about on a webpage, and began kind of exploring it then.
This is of course well before the release of schema.org. Back in those days, you had microformats were the main structured data type. I can’t recall yet whether datavocabulary.org which is a precursor to schema.org had yet been released. So the search engines and even though good relations was a well-developed ontology, the search engines weren’t doing much with that or anything at all, but following on that Jay Myers of BestBuy began marking up all of their product catalog and RDFa using in good relations. So, just ever since then I’ve been kind of keeping my finger on the pulse of those developments that eventually led to schema.org and the birth of well, Rich Snippets so predated schema.org I believe started to roll out in 2009 and incrementally we saw after that that Rich Snippets were being filled increasingly by search engine sanctioned structured data.
Martha: Very cool! So for those that aren’t semantic technologists that are listening, in kind of layman terms, like how would you describe an ontology?
Aaron: An ontology describes the types of things that you can talk about and the relationships between them in a highly structured way. So, for example an ontology might define a person like Aaron and it might describe a relation , for example Aaron knows Martha van Berkel and those are all highly structured relationships descriptions of classes so that there’s certain attributes associated with those. So kit enables data consumers to have a very precise view of what you’re talking about and applications since we’re just talking ontology is more generally and they are the building blocks then of taxonomies. Most notably the simple knowledge organization systems (SKOS). It rests on various standards including the RDFS so that there’s ontologies behind all of that and schema.org is sort of a de facto ontology of people arguing whether it’s an ontology or content schema, but like other ontologies, provides precise descriptions of things and the relationships between them. From that in and in modern parlance most ontologies are compatible with linked data, that is to say that they reside at your eyes that are de-referenceable, that means both humans and machines can go and then look up information about that class or those relationships at that URI. That’s something that people I think, kind of, don’t necessarily understand kind of more traditional SEOs about schema.org is that when you say a person has name or the name property Aaron, that person is schema.org/Person and name two is a relationship schema.org/name so that it creates a graph. That’s what you know obviously the biggest example of that is the Google knowledge graph.
The really big picture point that I love everyone to take away from this and really what the introduction first of Rich Snippets and then later of the Google knowledge graph really introduced into the search engine world is, as encapsulated by Google’s famous marketing phrase for the knowledge graph “from strings to things”, is that prior to the search engines were essentially indexes of documents. So when you search for something you really you know if you were looking for blue widget, you really weren’t getting results for things that were blue widgets, you were getting webpages that described or use the string the best blue widget.
Whereas now an especially since the interaction of the Google knowledge graph but also fueled by structured data, what we now have is a web of things kind of one of Tim Berner Lee’s great visions for the web, so that rather than being in index of documents Google is increasingly becoming an index of things, and facts related to those things. You can’t possibly have that without those sort of linked data technologies working in the background including ontologies, schemas, taxonomy solve that sort of thing.
Martha: Love it. So with this movement from strings to things in your role at EA, how do you see this evolution playing out in the enterprise. This may be a different than how we often talk about this, as like doing SEO for solution for different companies or within agencies. How do you see that application within the enterprise and how can they benefit?
Aaron: Certainly, it’s a crucial and key component in personalization and especially personalized content delivery. In order to deliver at scale-personalized experiences you need to know two things you need to know something about your user and you need to know something about your content, and the more model that is, the more rich you can make those experiences and the better the relevance of the content that you can present to a user. Now, to give you an example, so we might know that say a user plays Battlefield One, one of our games and in our own systems that’s linked to a taxonomic term which is the genre first-person shooter described on ontology and other games, then, are described by that same structure. So, therefore, we might say that Star Wars Battlefront 2 might be of interest to that player because they’ve expressed an interest or we know that they play Battlefield 1 and therefore we might want to surface content in, you know that very kind of play an example about Star Wars Battlefront 2 because it’s also a first-person shooter and you can infer that through having that structured universe. So, if I have a Star Wars Battlefront 2 article that I’d like to surface for a player it’s not necessary for me to label that, tag that and sort of typical content parlance first-person shooter that can be inferred by making a query to the system, “Okay, I know that this player likes Battlefield 1, show me all other games, content about all other games that are also of that genre so as long as we know that that that that piece of content is correctly associated with the game Star Wars Battlefront 2, we also know it’s a first-person shooter, we also know it’s playable on Xbox 1 and PlayStation 4 and PC. So, there’s all these attributes about games that then we can associate with content and then we can provide that data to players. That’s a high-level and very simple example, but I think you can see how that’s sort of inferencing can start to be really powerful in recommending content across the enterprise.
Martha: Absolutely! And I love how you sort of used to “ask the question to the data” right you know I think that’s something that I’ve seen in other interviews with you is where you say to think about what would you ask could Google assistant. In the enterprise you’re saying what kind of questions would you want to ask your data. Often we sort of ask that question around Semantic Analytics . we ask what do you want want to get from your analytics and can you sort of use some of that structured information to gain insights.
Aaron: Querying is really core to any sort of many of these processes, right, whether you’re querying your analytics or whether you’re querying your content or where whether you’re querying user events. If they’re described in the same ways, that starts to really present powerful possibilities of joining one with the other and when you’re dealing with data, you always have to think from it think of it from a machine data consumer perspective, right? So you’re not going to be able to get answers to questions that your…your data isn’t aware of, right? So it has to be structured in such a way that it supports those queries, and in building a lot of the systems that are worked on, that’s really where a lot of the MVP (Minimum Viable Product) requirements start with, its can it satisfy this query and this query and this query and if not, then you modify your architecture in order to satisfy that.
Martha: Which is a good lead-in to the next question around, “I work in on enterprise, for me I put on my hat from my Cisco days, and I want to know, how do I get started doing this right? If I want to move in this direction, while also sort of thinking about the search applications, what would you recommend to people sitting in an enterprise as a first step to help them get started or to get thinking about this?
Aaron: Excellent questions to which I don’t necessarily have a ready answer. I think, you know, familiarising yourself and starting to use schema.org is a good lightweight way of starting to be introduced to these technologies. I would actually go the extra mile, I’m just such a big-linked data guy…
Martha: Yeah, yeah…
Aaron: I think and I think to truly understand something like schema.org you do need to know a little bit about link data and how it works. What we would call the Semantic Web but link data is now more of the term. There’s a couple of canonical documents about that if you. Anyone watching this, go ahead and google Tim Berner Lee’s work. I forget the year but it’s called Link Data Design Issues in which he provides the five rules of link data. If you understand those 43 words, you’ll really, really understand linked data and the Semantic Web.
Aaron: Manu Sporny who’s actually one of the pioneers along with Gregg Kellogg and Markus Lanthaler of JSON-LD has an excellent series of YouTube videos on various introductions to the Semantic Web, JSON-LD, things like that. So that’s where I’d start in terms of knowledge and then, I go ahead and put this into practice by starting to apply schema.org schemas to existing content. At least if you’re in a web environment and seeing how that’s structured and what the outputs are and what the results are.
Martha: Excellent! Yeah, I’m a big believer in start with some of the standards and then extend because it’s you can move a lot faster, right? I understand sort of getting started from the technicals perspective, right, so will sort of include a lot of those resources. You obviously have great executive sponsorship and the work that you’re doing since you’d be doing this, yes since 2014, and sort of getting really deep into it. How do you translate this into the business value, like how do you tell that story, you talked a bit about personalization. Anything else with regards to sort of the business outcomes that get the executives excited about investing in this?
Aaron: Sure. I mean obviously you need to point to in any sort of digital marketing efforts do you do need to point to business value and I think that there’s a number of ways of doing that and I’ll give three examples.
One obviously from a search perspective if you start using structured data mark-up you will start to see rich results surfacing in the search engines and you will absolutely and unequivocally see a rise in engagement and CTR from those results because their information is provided to the users directly in the SERP, about that they tend to have higher click-through rates. You’re providing the search engines with really explicit information about the things on the page that are represented there, so you’re more likely to show up in query results for relevant queries and this should all at the end of the day translate into a better bottom line, particularly if you’re looking at you know the typical and correctly, so web metrics of a conversion rate optimisation and revenue.
You touched a little bit ago on and extending that another example is analytics and I think still one of the underutilized promises of schema.org is using it to tie in weather as Google Analytics or another platform so that you can start to slice and dice that data as well by the things that are represented in the content rather than by the strings that identify it. For those that work in Google Analytics at the end of the day Google content groupings are rather blunt instruments but if you’re already marking up your content particular using JSON-LD, then you have at your fingertips the capability of then exposing that information within your analytics and being able to make some really interesting and useful queries against that data to then extract business value.
Finally, if you’re looking this, you know, to truly enterprise skill if you’re a large company with, whether it’s a single domain that’s highly focused or an e-commerce site, if you start to build linked data applications you can start to use these same technologies to bootstrap your efforts particularly. That’s what we’ve done at EA as we built out our initial ontology for games is that we use schema.org to bootstrap that. It’s a very useful all-purpose sort of vocabulary that you can then locally extend but it tends to cover kind of most use cases that are most common, that are most common in across businesses.
So to recap them
- Start with fundamentals. If you’re a webshop do the search engine optimization.
- Start looking into the possibility of bringing that data into your analytics and start to think about. (Semantic Analytics)
- Pie-in-the-sky, connect all of the data points across the enterprise in some sort of semblance of order.
I think properly framed, you get executive support and buy-in from that and particularly if you follow those incremental steps because you’re going to show literal business value by those first couple of steps. It will lead to an increase in conversion in revenue.
Martha: Love it. Two more questions and then we’ll wrap up today. How do you know, who do you follow or what are you watching to stay on top of the trends, like in this space, especially around the enterprise?
Aaron: Yeah. In the schema realm, I follow the schema.org mailing list but more importantly these days is the schema.org github repository. That’s where most of the development work and discussions around it takes place and it’s great because you get to work with the people that actually built that the vocabularies and working on. Aside from that, it’s a broad spectrum of people. I actually have Twitter list that’s open, called Semantic Web where kind of everyone that I’ve ever been able to find that works on link data is there. I also you know go to conferences. I’ve been to the last couple of taxonomy boot camps in London. I spoke at Semantics 2017. For those that maybe aren’t that heavily into those sort of more extreme applications of link data but you know the search people, obviously I’ll promote my own community group that I moderate along with my friend and colleague Jarno van Driel Google+ Semantic Search Marketing which is a great starting point, particularly if I have any questions about structured data markup. We’re a very active and responsive community.
Martha: I love the debates in that, thank you, yeah. I think there’s a new joke (2.22) that I come in and weigh so that the business questions or applications, You can ask early technical questions, you can ask business-relevant questions, you can be a beginner or advanced and you guys are so welcoming to everyone. So my last question is – if you could wave a magic wand and change one thing about schema.org, what would that thing be?
Aaron: Yeah. Hard to say because it’s not especially lacking. Should point out too that schema.org really benefits from the experience of a lot of people that worked on the project from the early days that have been involved with it most importantly, Dan Brickley, who along with Libby Miller developed the first really broad scale web ontology which was friend-of-a-friend vocabulary for just grabbing people and the relationships between them. So, it’s really solid stuff and you have lots of luminaries of that space like Martin Hep and Kingsley Idehen who are involved in that space. So it’s it’s quite solid. I think if I could do one thing, I would vastly extend the number of examples and the clarity on schema.org. There’s probably for many practitioners too many pages where there are no examples or too many pages where the examples are a little esoteric when it should start with a straightforward display of information about that particular type and get more esoteric after that. So I think it has the potential to confuse some kind of novice webmasters there. In terms of the vocabulary itself, I don’t see any major holes and as those are identified, they are being filled.
Martha: Awesome! Yeah, we try to help our customers with those examples or like changing the definition. So you can actually understand if you’re not a semantic technologist what it means. Well, thank you so much for your time today. If people want to find you, where can they find you online?
Aaron: Oh, I’m everywhere. So Twitter is the easiest, twitter.com @aaranged. I don’t blog enough but I do blog sometimes, that my blog SEO Skeptic and as I said Semantic Search Marketing, the community I run Google+ Semantic Search Marketing.
Martha: We use it for that search community if nothing else. Well, I hear you’re scheming NINJA tshirt today. Maybe you can give us a flash of your T-shirt just for fun.
Martha: …with awesome JSON-LD on the back describing how you are indeed a ninja. Thank you again for joining us and have a great day!
Aaron: Thanks a lot, Martha, you too!