Andrea Volpini On Use and Importance of Structured Data For SEO

Hi. Thanks for inviting me. It’s great. Excited to be here today.

I dropped out of university when I was in the second year. I was studying mass communication. But at that time, I was already running a business on the web. I dropped out of the university and continue creating websites. Then I moved from website to content management system, so tools for creating websites. Then I moved on and I started to venture more into the field of Semantic Web Technologies as the content management system was expanding and Tim Bernal S Lee, the founder and creator of the World Wide Web, started to ask for data rather than pages.

Forward or something else? I was starting off a business and I was already at that point employing other people. The business was doing well, but I was the first out of many generation that didn’t get a degree. It was a big thing in the family. They accept it quietly. But yeah, I don’t know how many generation, but it was the first one. But I was an Hacker always. I’ve been always an Hacker in my mindset and always will be. Meaning that in order to understand things and to be able to explain things, I need to do things. Therefore, this path seems to be the most natural one for me. Then I end up, in any case, working closely with the university and doing research for many years with European funding that it’s actually what led us to the first prototype of where they were back in the days, 2015 or 16, we were doing research work. I always valued a lot the traditional education and the importance of spending time on finding innovative solution to problems.

I started on the web in 1995. That’s a long time ago.

Yeah. I’ve been working on the internet since then. In the beginning, I was still actually in high school with my existing co-founder and CTO of WordLift, David Ritchie Telly. We were in school together. The first business was about web hosting. We were buying storage from servers in the US and then we were starting to resell it in Italy for people that wanted to understand what this World Wide Web thing was. That was the beginning.

People usually found the dust. I always been first in many industry and I never had the problem of getting clients because I was one of the few, if not the only, doing with my team certain things, whether that was Web 3D or dynamic web pages or virtual agents, you call it. We went through everything, streaming. We’ve been doing a lot of video streaming. Everything was always first. Therefore, then the market comes to you because it’s just you.

I sold a couple of web companies before getting to the point of focusing more on this intersection between artificial intelligence and search engine optimization. I’ve been doing companies on the web. The first one was really doing web development. Then the second one was a larger agency where we developed these content management systems. Then I started again with David with this idea of creating the platform for the Semantic Web. That’s what.

Doing the same? Yeah, we’ve been pretty much all the way through together in this.

I think it happened in a mix of conditions, whether it was time for me to move on and whether the conditions were good for me to leave and build something new. I think that you evolve as life changes and also as you feel more attracted to new fields. And SEO for me was a new thing. Even though I started building web pages in ’95, until 2015, 2016, it wasn’t a thing, SEO, to be honest.

I’ve been developing very large sites for many years, so it’s not that I wasn’t new to the term. I was new in terms of understanding the importance of making search engines understand your content. I was always more into organizing content, making an effective navigation flow, helping users find stuff on your site, maybe developing the internal search engine for the site. But I never thought about the power of letting others understand your content through a controller and an index that someone else is running, like Google or Bing.

The thing is that we started with this vision of moving from the web made of pages to the web made of data, which was what Tim Bernersley was preaching in the old days when I started with this historical tatt talk to scream about raw data now. And so I moved on into the open data movement as we were doing a lot of work with governments. And so I really got myself involved in open data and the importance of publishing data out in the wild so that others could develop applications on top of it. And then I realized that the web should become all made of data because AI would be trained on data. As we were transitioning into the AI first world, having data at the end and letting people control the data was becoming pivotal also for an open and democratic society. But how could people publish data? How could people create their own knowledge graph? That was the mission that inspired us into building WordLift. All right.

No, because also we started slightly before the creation of the schema or consortium so the thing was really coming out. I was coming out of the open data and link data movement, so I didn’t see the importance of search engines until Boris De Maria, who is the co-founder of VueRank, rang me up and said, Hey, I’ve been following your work and it’s amazing. I come from the SEO industry and we want to invest. Then I realized that SEO was really our playground and maybe this idea of democratizing the web of data was actually a good fit for the evolution of SEO.

Well, this week we’re going to make an announcement. That was in the early days, even before the company was created. We partnered with VueRank. We still have a partnership in place. We started from the ground up a new company while I already was running other two companies within the context of professional services, web development, and also information extraction. This week we’re going to make an announcement about a new round that we successfully closed.

We are expanding, we’re growing, and most importantly, we are enabling a lot of websites to create structured data and also automate the process of structured data.

Structured data is a way of translating your content into a machine readable format that a search engine can understand. Unstructured is everything that it’s written and can be interpreted by a human mind. Structured, it’s something like a database that a machine can read. Now, structured data is like a database, but it also has the meaning that a human mind will detect out of unstructured data, out of a piece of text. That’s beautiful because it makes the machine able to understand the meanings much like a human would do when reading a piece of content.

Right. Let me share my screen and start from the end. Let’s try to look at why we want to use structured data. Tell me if you see the screen.

Let’s look at these images here. I’m going to pick up one image and then I’m going to search the image with Google Lens. Google Lens, it’s Google way of looking at images and it uses entities in the knowledge graph. As you can see, as I do this, Google is capable from single image to understand that this is now related to Andrea Volpini, who is an entrepreneur. Then it’s pulling a client incent from different websites that are related to, yes, that image, but also related to this entity. Here’s another example. If I go and ask who is the founder of WordLift, then Google is able to provide a rich result that will tell you that Andreo Volpini is one of the founders of WordLift, along with David Richie Telly. We’re providing this ability to the search engine of interpreting the content and providing answers rather than links, thanks to the use of structured data. Practically, what does it mean for someone that writes a blog? The first thing when you write a blog is that you want to make sure that the entity around yourself as a writer and as a publisher of the blog, it’s structured.
You want to start using the class person and the class organization in order to help the search engine understand who you are and how you are present already on the web. You want to start using these same as links and schema URL links that will provide confidence to the algorithm like we saw in this Google Lens example, that that is the person behind the blog and that is the same person that it’s acting on LinkedIn and it’s acting on Twitter and so on and so forth. That would be the starting point. Create your own entity on the web because you want the search engine to be able to answer two questions like that.

It’s likely more complicated than that. But let’s say that if you want to provide an answer to someone making a search, you want that answer, it’s coming from a trustworthy and reliable source. If I have never treated or written about, let’s say, I don’t know, fast cars, because I’m not into automotive at all. Would it be a good source of information for a searcher that it’s making a query on fast cars? Possibly not. Would it be a good source instead when people want to talk about online businesses and structured data? Well, definitely yes. The more we support the search engine in understanding what content we are capable of writing and how trustworthy we are in that specific niche, the more we are providing the ability of the search engine to help us rank for those specific queries.

If we look at the Google patterns that Bislaskovi usually highlights for us, we can see that it actually works in both ways because, of course, it’s rare that you will find content about fast cars on the World of Blog. Most likely you will find something SEO related. But the author behind the piece, it’s equally important. And of course, it plays within the context of the so known EAT factors. Structured data is also a facilitator of these EAT factors because we can entice the search engine to understand more about how specific we can be on that topic. We had done this experiment with the American University of Cairo, where we were able to rank on Google Scholar because we have added structured data behind their blog post and matched them with the deliverable of the research that was behind these articles.

There are two lines of thoughts here that we should follow. The first one is that we are not building structured data for Google only. This year I was honoured to be involved in contributing to the structured data chapter of the Web Almanack. The Web Almanack, it’s a yearly review of the status of the web, where we are. As I was analyzing and dissecting with the others in the team, millions of entries around the web of these structured data formats, I realized and finally saw that we’re not building structured data just for Google. We’re not using structured data just for SEO. We’re doing this because we are helping machines understand who we are. Now, machines like Google, yes, but we are also training our own AI on top of the structured data. Let me show you yet another example where we can see why when we talk about structured data, we’re not merely talking about Google only, but we’re also talking about our own website. Here we can see the graph embedding on the world of. Io website. I can look at the entity for structured data on our side, on our blog, and then I can immediately run these pipeline in order to understand that in WordLift, on our blog, when we talk about structured data, we mostly talk about knowledge graph and then triangular optimization and artificial intelligence and link data and reach sneak back and e-commerce.
Then Bislowski, you can see him here and so on and so forth. I can look at Bislowski or Jason Barton. I can see how these entities are related one to another within the context of the content of the WordLift blog. This helps us understand the cluster of content that we have. We are leveraging the same structured data that Google is using for providing valuable search experience on the Serp. The first line of thought is, look, you’re not building structured data for Google only. You are creating the foundation of the artificial intelligence strategy for your own site and for your own enterprise. The second line of thought is that, well, guess what? Even though the organic opportunities are evidently shrinking because of the AI first approach that Google is taking, our clients, our websites keep on drawing organic traffic. Why is that? It’s because, of course, as we leverage on Semantic SEO, as we nurture the machine with more qualified data, we’re able to surface on different features, whether it’s a top story in a news and media site, or it’s a Google discover traffic, or it’s FAQ content, or it’s the video traffic coming in the carousels.
We’re leveraging so many of these features that the traffic is not shrinking quite the opposite.

It does because search is becoming more conversational, meaning that we can see already with Google MUM features like topic Zooming. For instance, if I make a query like Kali Kube Tuesday, then Google MUM will fetch the upcoming event from Kali Kube Tuesday. Then the searcher, even though he doesn’t know that the upcoming episode is about X, Y, Z, will learn from these is a topic Zoom. How Google is capable of training something like MAM to enable such a feature using structured data, because that’s how we are transferring human knowledge to AI. Again, it’s not just Google. The behaviour of the searcher is changing. It’s not that it’s only Google doing that. We, as people, are becoming more conversational. We are getting more used to speaking with the automated assistant.

Everyone can use structured data. Structured data, it’s simply a form of open data. Now, open data, it’s the way in which we can communicate with today and tomorrow’s AI. If we want to be heard, we need to speak to the AI. The knowledge graph and structured data, it’s the language that we can use to talk to the AI and to let the AI understand us, whether the AI is from Google, Bing, or our own AI. Structured data is for everyone. But of course, we have to focus on things that matter because, of course, is it important that I am recognized as an entity person? Is it important that my products are recognized? Is it important that I specify, for instance, the information about the shipping details of my products? Do I want to facilitate a visual search like we saw in the beginning? Or would it be better, for instance, to focus on the identity of the brand? These are very strategic decisions that need to be taken before thinking about what type of knowledge graph I want to build.

Right. WordLiftt automatically will read the content of a blog post and will automatically extract the most relevant entities using natural language understanding and name entity recognition. This will automatically provide Google with a lot more meanings about a specific blog post. Even by doing something which is similar to tagging, WordLift is providing search engines with a wealth of information. As we saw, it’s not just for Google because then I can use this data in the graph, for instance, for creating this topical cluster around a specific area of interest like structured data on the WordLift blog. I can use this graph also for creating maybe a headless set up. Rather than going through WordPress, maybe I can build these headless in JavaScript to make the page super fast. But all the metadata about the content is within the knowledge graph.

Right. Let me quickly share the screen and look at a blog post so that we can see these practically. This is a blog post about employee burnout. Here the wheel of the Earth is detecting a set of entities that might or might not be related to the content of the blog post. As I am notating, I can also decide if, for instance, I want to add the links to this content or not. Just by doing this, I am already doing a lot for SEO because I’m providing Google with entry points for multiple queries that are still related to this blog post but somehow attached to these concepts. I can do this with simple tags. Now, of course, when I do this, I can see in the front end that WordLift is already creating these context card widgets. These content cards widget are providing the reader with some understanding of some of the concepts that I’m using. Of course, I can decide if I want to have pages behind these entities or not. This is, of course, an SEO decision that needs to be taken. Of course, it has to do with the theme content question that you ask.
Typically, WordLift will start to create these vocabulary of terms by looking at the content that you already have so that we don’t have the problem of theme content. But we’re going to also get the content from Wikipedia in order to provide some of these entity description. But the entity description can be simply used for building up the structured data, or it can be used for creating pages. But these are decision that the user can take and that will impact the SEO strategy. If I limit WordLift to use the data only for annotating the blog post, I don’t need to worry about the thin content. You can see here that WordLift is saying that this is an article in structured data form. But then it’s saying this article talks about burnout and talks about the other concept that has been detected. Automatically, the world is also creating links to the web of data. It’s not only referencing a generic concept, but it’s also referencing the same concept across multiple data sets on the Semantic Web. This is creating a knowledge graph around the article that any machine can leverage on for understanding what the article is about.

Absolutely. First of all, there are different cases, of course, in every site, it’s different, but there are some unique aspects. In general, as I said, there are different levels. But the starting point is that I add semantic annotations to an article. Only by doing that without creating links, without creating entities, without adding additional content, only by doing that, on average, we see between 6 % and 8 % increase of the organic traffic over a period that can vary between the two weeks and the two months.

It’s 6 %, 8 %, which on a larger platform, means a lot. It’s pretty stable growth. Now, as you start to use the tool and you start to add links to pages that you have, maybe for a specific concept, or as you start adding widgets, which is also another area where WorldLift allows the reader to discover more content using the data in the graph, then these can go up to maybe 30 % or maybe 40 % increase on a period of time that varies, again, from the two weeks for sites that have a lot of traffic to maybe the three months for sites that have less traffic.

No, not on my side, not on a third party website. On my side, I create a page for, let’s say, grouping everything about structured data. Then this page becomes a category page, but it can be editorial as well.

Well, at that point, it’s not just structured data, but it’s also the time that the user will spend on the side because, of course, if I’m reading an article about modern SEO and there is a reference to, let’s say, JSON LD, and I’m not familiar with the concept of JSON LD, and you allow me to understand on your side with an internal link what JSON LD is because you have written already about it. By creating these topical clusters around the article, we are increasing the opportunity for the user to discover content. Semantic SEO, it’s all about thinking outside of the logic of keywords and pages.

Think in terms of topical clusters. When I write a blog post, I want to cluster everything that I have in my galaxy of terms and content pieces that can help the reader discover more. That’s the beauty of it. That would create at least a 30 % increase.

Writing it? As I said, there are two ways. You can, whether you can look at the content on your site and then, let’s say, for instance, transform every tag into an entity. Therefore, your tag pages will become these new pages. Or World of can get the data from the link of data, so Wikipedia or any other external source. Or in some specific implementation where we start maybe adding more integration services, we can also generate it using language models. There are various ways in which you can start looking at scaling up the content production.

Let me get back to the previous example for the article that we were working on. Then we can see the widgets in action. Here we are. We are building these knowledge graphs inside the article that represents the information about this piece of content. As you can see here, I’ve also added a video and Wharrley has detected the video that has been embedded in the editor and it’s already pulling the metadata from YouTube automatically. Whether this is my video or it’s someone else’s video, now it will appear as a video object markup bound to the article and then I can personalise the description, the name of the video and all the metadata that the WordLift test has created for me. But let’s look at the widgets. When I do this, I can add a block and for instance, add the Faceted Search Widget, which is one of the widget that World of provides for connecting things that are semantically relevant. Let’s have a look at this. Here, it needs to be styled, but World of this, of course, and this is a demo site, so it’s not necessarily the best association of terms. But here, World of this is providing an interface which is similar to the one that we have on, let’s say, Google image search, where entities are used to filter the content that is related to.

Article. Then this content that it’s coming from my own site. Usually, just by adding these, we get 30 % more page views per session because, of course, the content is semantically related. The user can filter the list by tweaking on the network of entities. It proves yet one other way in which we can increase the traffic and create value for the reader.

Related articles. Usually when we place the widget, we look at two metrics. We look at the increase of the page per session on the pages where we have added the widget, and then, of course, we look at the average time spent on the page, which will dramatically increase, maybe by 40 % or 60 % even in some cases. These plays on the metrics are scalable because then you can add this widget at the template level and then every page will benefit from this, even from the archive that you have. When we look at Semantic SEO and how much traffic it can generate, it usually talks about a minimum of, say 6.8 just for the semantic annotation, then 30 % if you start creating your own vocabulary. Then, of course, you can go up to 60 %, 500 %, whatever, depending on, of course, the amount of content and the ability that you have to publish content.

You don’t want to overdo it with structured data. We want to keep things as simple as a child would understand. This is a good practice for Google, but it’s also a good practice for your own AI. Because if I start to overstructure content, because if I start to overstructure content… Because of course now I can structure content by type, so I can use podcast episodes for all the podcasts and then I can use, let’s say, service mock-ups for the pages where I describe my consultancy service. Then let’s say I can use, I don’t know, local business for the pages where I talk about my presence on the ground. As I do these, I have to keep things tied up together in a meaningful way and I don’t want to overdo it. I want to do things that a child would understand. Because if I overdo it, then it’s going to create noise greater than value. You want to keep things simple and consistent.

On using? WordLift, it’s a SaaS. It’s priced with basically three starting plans. Plans defer one to another based on, for instance, the type of entity they support. If you want to, for instance, use the video object extraction that I show you, which will also create a video sitemap automatically for all the videos that you have created, then you will need to use the business. Whether you need to use, for instance, the local business and you have more than two sites, then you want to use the professional. And if you have just one single site, you want to get into the process, you can start with the starter. The starter is €49 per month and then the professional is €79 per month and the business is €199 per month. In the business, you also get 10 keys so you’re going to use Whirl lift across 10 different websites. Then, of course, a big chunk of our clients require us also to get into the consultancy for supporting the creation of the knowledge graph in the beginning. We have packages that start from €1000 per month and then they go up. You have to imagine the world is something that can run on a single blogger site or on a website like Airbnb or Orkney.
It’s an infrastructure that scales from a single blog to a large e-commerce. We’ve been lucky enough to work with a large corporation that makes hundreds of millions of euros on a website and with the average blogger. The infrastructure is.

We don’t look at the traffic because we want to create more traffic. That would be heating, working backwards in the models. We have to look at the number of pages, especially when we, for instance, look at e-commerce. One of the e-commerce companies that we’ve been working with has more than 700,000 products in catalogue and is building a product knowledge graph for 700,000 products. It means a lot of data. WordLift hosts the data in the knowledge graph for you. We have to price it by the number of entities that are published. But this happens outside of the three plans.

Future plans.

Yeah. That’s a very smart question because balance is really the key to success here. We are a SaaS and our goal is to host as many knowledge graph as possible to create these reputable, scalable SEO tactics that allow an organization to get structured data, to use it in the most interesting way, but also to create these AI infrastructure on top of it, which one day can be used for generating the product description, and the other day can be used for recommending the best content. We need to keep on developing the platform in order to serve these advanced use cases. Everything that has to do with an advanced use case will go through the professional services. But the platform, which is what we are capitalizing on, it’s the same. The same platform that served a starter for 49 is the same platform that is responsible for the structured data on a very.

No, we are keen on training other people. We need this thing to move on its own. We’re not only training the editorial team or the SEO team on the other side, but we’re also training specialized SEO agencies around the world to take it forward for us because, of course, there is so much need and we want to keep the team small because we need to stay focused on the technology. I want to enable more use cases. Right now, we’re working on enabling, for instance, the Neural Search on e-commerce because we realized that we have a wealth of metadata inside a product knowledge graph that it’s used for Google. But this metadata can be effectively used for on site search. We want to bring a better on site search experience to all of our clients and we need to get this technology in place and then we need to bring it to the end of the people, which means that we need a network of professionals that take it forward.

It’s still an unsolved problem on site search. But the reality is that it’s not a technological issue much as it is a content structuring issue. The problem is not that the Algolia or the elastic search that you might already have are not capable enough. Not at all. The problem is that the curation of this data is not as precise as we think. In SEO, though, we’re already doing this work because we’re doing it for Google. Why don’t we leverage on the onsite search?

That’s exactly what we’re doing. We are actively working with the team of GenII in order to create the best possible neural search for e-commerce. I can also, in this case, give you a quick preview of what we’re doing in this area because, once again, I mean, World of it’s a technology platform for automating SEO at scale. Working on the internal search, it’s part of this process. If I say I would get something like this, but then I can also do multi model search, and then I can leverage on images to run a query. Do you see that? I’m adding a pair of sunglasses here. I’m running the crop, and then I’m just going to hit here and I’m going to use this image as part of the query that hopefully will work and will help me return relevant results for this type.

That’ll be okay. This is a true multi model search, meaning that, of course, I can then combine visual query like the one that we saw with the text and say I want this in pink or I want this in black. We’re building this infrastructure that once again, uses the structured data in the knowledge graph as a.

Yeah. We’re going multi-platform and at the moment we definitely are pretty strong on e-commerce sites outside of WordPress because we start from the product data and reach the product data and bring in the metadata, the structured data right on the page. We are less fast on editorial content that is not running on WordPress. We do have, though, a couple of large implementations where the CMS is proprietary and we have a set of APIs that can be used for doing the integration. We work outside of WordPress for e-commerce. It’s pretty easy. Outside of e-commerce.

Absolutely, Drupple. It’s definitely one of the CMS that we’ve worked on the most. Definitely, it would be interesting to start integrating WordLift.

Your life? One thing that I learned is that I get very happy by building an amazing team. When I can aggregate like we’re doing now, motivated professionals that discover something completely new that it’s mind blowing as building the web of data, I am satisfied. That’s one thing. The other thing, of course, is that my mind thinks in terms of what’s going to happen in the future. For me, it’s always about what technology we need to build to enable the future. In the future, your website is going to be a chat bot, and I want to build the technology that you need for making it easy for companies to create, let’s say, a chat bot or a neural search. Staying there and building the tools that people need, it’s also what makes me happy.

Maybe a product launch site. It is. We see that the knowledge graph, it’s the building block of your AI strategy, and every organization needs to have an AI strategy. It’s crucial for SEO. Our angle will remain SEO, but we see SEO becoming more and more holistic and covering different aspects of the user experience.

For it.

Three words? Italian, technology addicted, future.

Building.

Building an amazing team. Okay, awesome. And working with clients. Working with clients, it’s vital because I learned so much from our clients and we have such an amazing set of clients from all over the world. I’m very thankful because of what we learn every day.

In a parallel universe, I love to meditate a lot. I would spend more time in a mindful state, which I do at the moment. Sometimes writing code, sometimes doing crazy stuff with clients. Mindfulness will.

Thanks.