The WordLift Academy is proud to present the Knowledge Graph Quartet — the world-class experts on Product Knowledge Graphs. Follow them to discover the true value of Knowledge Graphs.
Jason Barnard, Jarno van Driel, Teodora Petkova, and Andrea Volpini will discuss the importance of Product Knowledge Graphs for e-commerce marketing and SEO. Coming from four different areas of expertise, they will help you get the big picture of the role of data in today’s digital marketing and online sales.
Jason Barnard, also known as the brand SERP guy, will show you how a well-designed Knowledge Graph can impact your SERP and result in a richer and more engaging exposure for your products. Jarno van Driel is recognised as one of the world experts on structured data and provides semantic modelling services to leading brands and organizations. Teodora Petkova is specialized in weaving together text, multimedia content and linked data. Last but not least, our CEO Andrea Volpini is a pioneer in the field of semantic SEO and provides tools and consulting services to brands that need to grow online.
If you have an e-commerce website, stop doing what you are doing and create your Product Knowledge Graph.
From the shelf of local shops to Google Shopping’s search results, a Product Knowledge Graph can align all the data and information that your customers can see and find about your products. In this webinar, you will learn why structured data is vital to your business and what is the impact of a Product Knowledge Graph on Google’s SERP for your products and brand.
What Are You Going to Learn?
- What is a Knowledge Graph and why it is important in the context of e-commerce
- What is the concrete impact of a well-designed Knowledge Graph on your SERP
- Where can you start to implement your own Product Knowledge Graph.
- And a lot more!
Meet the Knowledge Graph Quartet
Four experts, coming from four different areas of expertise, discuss the importance of Product Knowledge Graphs for e-commerce marketing and SEO.
Jason is a digital marketing consultant and one of the leading experts in Brand SERPs and knowledge panels in the world today.
Jason is co-founder of Kalicube, a groundbreaking digital marketing agency that pioneered the concept of Brand SERPs (what your audience sees when they google your brand name). He is also a full-time 100% digital nomad, host and keynote speaker at conferences around the world.
Jarno van Driel
Jarno is a structured data consultant who provides semantic modelling services and guides organizations into the era of the semantic web.
Over the years his work has been mentioned several times at W3C discussion groups, due to being an early adopter and large provider of structured data. He contributes to some of those groups helping to improve common standards and continuously learning more.
Teodora is a philologist and a freelance content writer with an educational background in Classical Studies and Creative Writing. She helps companies with content ideation and creation, writing for their blogs.
In the past, she worked as a Latin teacher at the University of Sofia; today she is still passionate about research and she loves to share her experience with web writing.
Andrea, CEO of WordLift, is a visionary entrepreneur, now focusing on semantic web and artificial intelligence.
Co-founder of InSideOut10 and director of InsideOut Today, Andrea has 20 years of world-class experience in online strategies and web publishing. Nowadays, he provides tools and consulting services to brands that need to grow online.
Jason Barnard: [00:00:00.33] Welcome, everybody.
Jarno Van Driel: [00:00:06.30] Thanks for having me.
Jason Barnard: [00:00:07.95] Yeah, I start with you, Jarno. Really quickly, who are you and what do you do? And why are you so good at knowledge graphs?
Jarno Van Driel: [00:00:12.45] Grass used to be a front end developer, became a semantic asiel, which involved in becoming a structured data specialist. And nowadays I look at structured data at a whole larger picture within organizations and provide consultancy based on that
Jason Barnard: [00:00:28.38] Which is wonderful. And Teodora [00:00:30.00] writing a doctorate, I believe.
Teodora Petkova: [00:00:33.33] Yeah, I’m writing my PhD thesis. I’m a content writer, but I decided to dig deeper into what communication on the web is.
Jason Barnard: [00:00:43.26] Absolutely do. I love that, Andrea. We know each other well with WordLift.
Andrea Volpini: [00:00:48.48] Yeah, I’m a co-founder of WordLift. We build the knowledge graph for SEO.
Jason Barnard: [00:00:55.07] Brilliant stuff. Well, I’m Jason Barnard, the brand serp guy, Kaicube. I’m obsessed by brands serps. That’s what [00:01:00.00] happens when your audience Googles you brand name, which is where we’re going to start, because that is my choice, is to say that my experience here is not so much in product knowledge graphs, which is terribly detailed, lots of things all fitting together into this massive marketing mess of moodley selling. I deal with the bigger picture, which is the brand or the person and what appears when you search for the brand’s name. So I’m here to kind of drive the conversation forward [00:01:30.00] and try and drag as much information out of Teodora and the Jarno as I possibly can so that I, like everybody watching, can learn about this specific aspect of knowledge graphs. So we’re going to start off by saying it as we’re going through this. I think we can look at it as saying that, Teodora, you’re really into the theory, Andy is really into the marketing with the knowledge graph aspects and Jarno is really into the geeky tech stuff and actually how to do all of this in a practical manner. Is that a fair assessment, [00:02:00.00] you three?
Jason Barnard: [00:02:06.00] Wonderful. So I’m going to start with Teodora, then, the theory, the overall over arching philosophical doctoral thesis, aspect of product knowledge graph. How are you approaching this whole question?
Teodora Petkova: [00:02:20.86] Wow. Thanks for the wonderful question. That almost answers what you’re asking. The theoretical part [00:02:30.00], quickly for everybody empowered yet distracted user. Overwhelmed marketing department who are thinking that digital marketing is something different from what they did. And and they don’t think it’s functionally or formally different, they just think they need to get all the marketing communications and translate them or publish them on the Web. [00:03:00.00] They don’t understand the hyper textual nature of content, first,. Next, brands and organizations. Our. Very slowly realizing that they are not to sell on the Web, but rather tell on the Web. Another theory thread, the web of data. You [00:03:30.00] can’t ignore the web of data when you’re doing marketing communications, when you’re connecting to the other through the web. You need these vehicles or you need to address that content into interoperable meaning, data that can be read by different devices and exchanged by different devices. So you need your content and your thinking to be informed [00:04:00.00] by this giant web of data. A giant global graph is what Tim Berners-Lee calls it .
Jason Barnard: [00:04:10.49] Ok, so the Internet is a giant web of data. Andy, I know you build internal knowledge graphs. I mean, it’s difficult enough getting your head around the idea the web is all this interconnected data. But even within our organizations, we do a really bad job of actually having any representation in terms of data of our inventory [00:04:30.00] or our content or anything else.
Andrea Volpini: [00:04:32.75] I think the challenge of the knowledge graph is the challenge of representing reality and that’s very difficult when you deal with your own identity and it gets even more difficult when you start looking at the products that you have, that you want to talk about. As Teodora say, we don’t use the word “sale”, but building a knowledge graph it’s challenging in a way, because it’s a form of communication between [00:05:00.00] the humans and the machines that kind of help us get through the digital journey. So we need to build knowledge graph in order to let A.I. system and machine learning and deep learning understand what we’re doing and what we care about. And that’s really, for me, the most important concept of a knowledge graph. It’s like the connecting dot between the human and the machine.
Jason Barnard: [00:05:27.68] Well, yeah. I mean, because we’re basically saying we’re [00:05:30.00] asking the machine to represent us to its audience or its users. And if we want to have any chance of it representing us in a way that we would wish to be represented, we have to communicate properly. And machines understand knowledge graphs incredibly easily. They don’t understand unstructured data. Jarno, I’m sure you can take that to another level
Jarno Van Driel: [00:05:49.55] For me, structured data or even more the process of structuring your data is simply a vehicle for getting things done. In [00:06:00.00] my positions over the years, I’ve barely ever mentioned the word knowledge graph within any organization I work for. I simply started out pointing the single points of the multiple points of failure within marketing systems or within products. So once you start showing organizations where things go wrong and when there’s missing information and how different departments can start to enrich [00:06:30.00] each other, they do not necessarily need to talk about knowledge graphs. You directly start to talk about business impact. And so that’s most of the time my entry points to knowledge graph isn’t so much the concept of a knowledge graph, as simply using the existence of it as a method to connect different areas within an organization or their marketing departments or the product pages or whatever it is. For me, it’s more practice than theory.
Jason Barnard: [00:06:59.27] I like that, and I also like [00:07:00.00]- well, don’t like the idea. We all have this terrible problem of all these different departments with their big stack of data, which is badly organized and they don’t share it. And potentially a knowledge graph internally is a great way to share knowledge. And, what I don’t think we realize, is it goes beyond just products. And if we can bring all that data together and get everybody shown from the [00:07:30.00] same pool of data, we be will winning on every front. Teodora, please expand on that, theoretically speaking.
Teodora Petkova: [00:07:41.48] Yeah, winning on every front… Data… What I thought while I was listening to you is, again, getting back to a theory. That mindset of “Webbiness”. No matter how many pieces of content you create, [00:08:00.00] like “seven ways to use your refrigerator”, it doesn’t matter. You’re wasting money if you’re doing content marketing. You need, as Jarno said, to get your data right, to have that impact, you need to understand that. Marketing, communications, people need to understand that next to content is that idea of publishing data, and this will bring [00:08:30.00] the departments, the creativity and the illogical part of thinking about a CEO when you are discussing the package of your good. How are you going to engage the person in the supermarket? How are you going to get them from the physical reality and warps and direct them to your knowledge graph? Isn’t [00:09:00.00] that exciting? I’m getting goose bumps.
Jason Barnard: [00:09:07.07] Yeah, and that is the same kind of as soon as you start thinking “I’m not publishing content, I’m publishing data and I need to link that data”. All that data needs to have been linked to something else. I mean, Andrea, I think person is a human being. I find it very difficult to start a piece of content, and it’s not the first thing in my mind. This is part of a bigger set of data that all needs to be interconnected. [00:09:30.00]
Andrea Volpini: [00:09:31.88] And I always look at the perspective of the human and then the perspective of the machine. So the perspective of the human, if we look at the consumer, is that we want to see things all together. So when we look at a product, we don’t just look at the product page. We take a long journey and this long journey goes maybe even through the, you know, the the seven details of the refrigerator, you know, but it just one step. And [00:10:00.00] there are multiple facets that that the human will consider even before spending time on or looking for a product. And in marketing, we we then have these very narrow view where we then say, “hey, that’s the product page, this is how it should look”. But in reality, things are way more complex. And the same it applies to the machine, because if you have to write a crawler or if you have to build a knowledge graph from the things that have been [00:10:30.00] published, then you want to collect different data sets and you want to have maybe a feed, which is structured, which can be something similar to the merchant feed. But then you also want to see what people are writing on the Web in a natural language. And then you want to read structured data from these Web pages so that then you can consolidate all these information and store it inside your knowledge graph. So I always see that we have to look at these two perspectives, the humans that it’s consuming [00:11:00.00] the data and then the robots that are using this data to create these this representation and the system that we use every day. I mean, a recommendation system requires these information. You know, an email marketing platform requires this information. So everyone kind of needs this collection of data.
Jason Barnard: [00:11:22.09] I mean, Jarno, right now, and I heard kind of the product feed idea, most companies that I have heard of, and I deal with smaller companies than you do, [00:11:30.00] they’re thinking, “oh, product feed, I’ve got to start organizing my data”. What a bore, how horrible. And that’s unfortunate.
Jarno Van Driel: [00:11:40.05] Yes and no. When you first start doing it, it’s not fun because you don’t know what to expect, what the outcome is going to be of your efforts, and you really need to put in some serious efforts, even if it’s for something that most companies do like setting up Google merchant feeds, merchant [00:12:00.00] feeds. That’s a pretty common thing to do. But getting your data set in order, that’s often torturous experience for a lot of companies. The interesting part is just once you’ve done a good job is look at the possibilities, she created it for yourself. The downside of doing that is that most companies do these efforts in separate silos, so today they decide to start working on Google merchant centers. Tomorrow, they decide [00:12:30.00] to do something on Facebook and the issue often lies already in the early stages of businesses when there is still a small company and they’re just trying to build up their business, and that’s already at the point where companies make that or more informed decisions. And one of the things small companies really have to be aware of is once you start to push into new areas for your marketing, make sure [00:13:00.00] you take the time to already make inventory of the type of information you’re going to be needing to do that and hold that information to any other effort you’re going to be doing, because by comparing different marketing outlet channels, you actually can start to make a good inventory for your future public knowledge graph. As in regards, which values do I actually need for my products to be able to do email marketing, to be able to do AdWords. So it’s not just a thing for big companies. It’s a thing for big companies [00:13:30.00] because they didn’t look at it when they were still a small company. So dearly to start infantilizing your data warehouse, so to say your data needs. And the better you organize that, the easier it becomes to grow, because then you’re starting to plug in additional information as opposed to creating separate data sets for every outlet there is.
Jason Barnard: [00:13:54.01] Yeah, once you’ve got all your little data sets and they’re all separate and fragmented, it’s incredibly difficult [00:14:00.00] to bring it all together, especially when it’s siloed down. Teodora, that’s a big problem is, as you said, companies tend to do things in little chunks with different people who are owning that particular little chunk. with all different ways of thinking. The advantages of centralizing are obvious. But the practicality is incredibly difficult. Teodora that was kind of was an open ended question, that wasn’t very personal.
Teodora Petkova: [00:14:30.26] Yes, [00:14:30.00] I was thinking about maybe touches your open ended question, that idea of bringing everything together and understanding your business. That’s the very heart of relationship marketing and building relationships. Why? Because it’s a knowledge intensive business. It’s a knowledge intensive activity to engage meaningfully [00:15:00.00] with people. Also, a business, need to be able to talk within itself. What Jarno is talking about is also, for me, valid within the business. That is, you have internal audiences. So you have algorithmic audiences, you have external audiences and you have internal audiences. The people who are writing the content [00:15:30.00] or are doing the things, they need that knowledge base, they need knowledge. So again, key word, knowledge intensive. Knowledge is an edge.
Jason Barnard: [00:15:43.04] Yet there’s also the question of being consistent across your brand, but also your products. I mean, I’m obsessed by brands. Building Kalicube, which is a platform for basically making your brand message consistent across the Web, I realized that most companies, including my own and [00:16:00.00] myself, are a long way from being consistent, however, consistently we think we are. And when you take that at the brand level, I’m looking at anything out, we all foolish and aren’t we all rubbish? But it is different departments, even in my company with is only two people, two different departments, two different approaches, three different ways of saying things. And then you multiply that by seven or eight, 10, 15, 20 departments and five thousand products or even hundred products. As you said, Jarno, for a small company, it’s a phenomenal, phenomenal problem that [00:16:30.00] knowledge graphs solve, Andy.
Andrea Volpini: [00:16:31.76] I mean… you create a then you scale. I like a lot Jarno’s example that you have to start small because, I mean, in a smaller organization, you’re still kind of organized stuff. And and everyone struggled with putting effort in managing metadata, because it’s you know, in the beginning, it’s not exciting. I agree. But then, you know, the degrees of application [00:17:00.00] of this work are so huge that even if you just run email marketing and ads or even if you just do SEO and ads, then you see the advantages that kinda scales. So it’s one of these things where you put effort and then you get a lot of value. The big problem is that these values is previously unknown. It’s very hard to kind of estimate, you know, what are we going to get from these, you [00:17:30.00] know, data quality improvement effort that we’re going to make? And that’s hard, because it highly kind of depends on the type of of data that you’re managing and the level of competition. I mean, I was running today is this analysis on a decent kind of a decent size retail shop in New Zealand. And I was looking at an increase in organic by 16 percent by [00:18:00.00] improving the data quality on their product pages. But if you look at it, it’s kind of a minimal impact compared to what you can do now that you had better data for this product, because, I mean, it can help you organize the logistics in a different way, can help you create a better recommendation system for your email marketing campaign. So it’s really a goldmine, but it’s very hard to kind of estimate these. I think we [00:18:30.00] have a question from Jenny Hill. How do you help business see the big picture.
Jason Barnard: [00:18:38.76] How do you sell it to the boss? I was going to be my next question. Thank you, Jenny, you’re a genius, Jenny, the genius. I’m sure you’ve got an answer
Andrea Volpini: [00:18:46.97] I was asking Jarno to take this.
Jason Barnard: [00:18:49.35] I’m sorry.(laughs)
Jarno Van Driel: [00:18:51.23] For starters, don’t expect you hit homerun from every time you’re intending to sell [00:19:00.00] such products projects, whether you’re a consultant or internal employee, it doesn’t matter. You’ll never get a one hundred percent success rate, accept that. Coming from my position, where I normally start to build working on these things within the organization, I quickly make an inventory of what type of data outlets are there, which marketing channels do we need to optimize, which marketing channels don’t we use? So [00:19:30.00] looking at it from a wider perspective, simply as a marketing, you can easily recognize if there are possibilities in better advertisement or better targeting on Facebook or email outreach. So make an inventory of where the opportunities lie and make sure to sell one of them to the to the big boss.
Andrea Volpini: [00:19:54.91] Yeah, one addition.
Jarno Van Driel: [00:19:57.16] Exactly, make the inventory and see where [00:20:00.00] you think you can make and make the biggest impact and start with selling that single site project, as you start building up successes, it becomes more easy to start telling the bigger picture, because if you’re just doing it based on your system, it’s a small case to sell. But once you start involve email marketing and you start looking for content creation, actually start to show within the business case by case that you’re improving numbers, whatever the metrics are, then you slowly start to work [00:20:30.00] towards the bigger picture. Don’t try to sell the bigger picture immediately. You’ll loose the client.
Andrea Volpini: [00:20:35.50] Selling. The bigger picture is always the wrong. I mean, I’ve done enough. So it’s like a mistake that I’ve done many times. Look at the big picture, but but then in reality, yes, we have to start small and we have to kind of target a specific project. Now, in my case, of course, today, I would look at SEO opportunity first because that’s easy to measure [00:21:00.00] and kind of SEO in e-commerce, I mean, Jarno, maybe you know more about it, but it looks to me that, you know, SEO wasn’t kind of designed for e-commerce because e-commerce was so highly dependent on advertising, on Google, and therefore it was less interesting as an opportunity, whether today is kind of an area of expansion, beacuse [00:21:30.00] Google is providing more visibility to, you know, natural listing. So..
Jarno Van Driel: [00:21:39.47] With the current state of things in schema.org, by now, we actually have sufficient vocabulary to build a complete CMS. It’s excellent. All the data is there and all the formats are there to build a new Magento or to build a new Wix or whatever platform you want to build. There’s actually structured data for that. Why [00:22:00.00] is that nice? Because those schema types and properties are based on input from different areas in the world and different standard, so it’s been influenced by good relations vocabulary, which forms the foundation for the product side of things and schema.org work, but with the latest effort is also involvement of Google Merchant Center and translating all their specification to schema.org. And then we also have to use one vocabulary, which takes you a level deeper for people [00:22:30.00] like manufactories or wholesaler’s. So from that point of view, quick win always is as you get those star ratings, get those prices shown in a service, but that’s the low hanging fruit, that’s the easy stuff. The second easy one is always go sit down with the people that do the ads, go look at the numbers with them, go do your keywords research with them and try to discover which areas you’re not doing well, [00:23:00.00] you haven’t exposed yet because most of the time when people are doing ads run into issues, I can’t go into this product group or I can’t make ads for this, is because the system itself on the website isn’t able to filter or slice and dice for the things they want to add, but to advertise for, for example, green sneaker size thirty eight. If you’re web pages can’t slice on those filters, you can also create ads for them. So that’s one of the things I always look at. OK, [00:23:30.00] are there certain types of information we need so we can expand on networks.
Jason Barnard: [00:23:35.61] But then e-commerce companies and sites are faced with a phenomenal problem is that all of these platforms have been built differently and don’t use the standardized schema, which means you’re trying to kind of shoehorn around peg into a square hole or vice versa. So what do we do? What do we do now Jarno?
Jarno Van Driel: [00:23:59.52] Again, baby [00:24:00.00] steps. You first start to look at things, OK, which is the most important channel right now? Is that channel performing well? No, let’s make the advantage there. Because if you take the biggest revenue generator and turn on a higher revenue on that, then you’ll also get more resources allocated to start expanding on the data that you build for that outlet. Don’t create a second outlet. Start expanding on the data set you already built. And that’s actually the key change people have to make. Stop making separate [00:24:30.00] data sets for everything. From the start up from the first feature your building, make sure they all pull from the same data set and that’s where you’ll make the real advantage. That’s the big step. And save millions in the end.
Jason Barnard: [00:24:45.00] Yeah, sorry. Excuse me, I’m interrupting, but that’s a great quote. I really want to kind of focus on the making multiple datasets for the same thing. And the one thing I was thinking about is what you guys do at WordLift, which is start with a small chunk which is manageable, i.e. one [00:25:00.00] section of your catalog, and do that. Prove that it’s going to make you more money, then move on to the rest, which gives you a manageable chunk and B gives you that first first proof of concept. Go ahead.
Andrea Volpini: [00:25:10.43] Also I mean that the linked data infrastructure that we use, it helps us because we create these unique identifiers for each item. So having these nodes as identifiers allows us to kind of keep track of the same [00:25:30.00] product across different outlets or different pages, because we know that in the end, it’s always that same entity that has been described in that video, has been presented in the FAQ, has been blogged about and it’s presented at last thing in that product page. And now this is why we’re looking at omnichannel as a new area of expansion, because of course, in the physical world we also need the same data and [00:26:00.00] working with Teodora, we realized that, you know, these omni-channel challenges that the e-commerce is now ready for handling, you know, the physical consumer. And on the other side, the people at the point of sale kind of locked the same data that day e-commerce website would have. And so there is also kind of the need of bridging these other gaps between digital and [00:26:30.00] physical. And it became so important now because of the pandemic for so many businesses that have to transition quickly to digital, but at the same time keeping kind of the retail experience active. And what it all becomes, I mean, they’ve been kind of analyzed and tracked and optimized as separate revenue streams. But right now it’s pretty much converging.
Jason Barnard: [00:26:55.56] Yeah, another yeah, the offline and online is something I think traditionally over the [00:27:00.00] last 20 years, we’ve done incredibly badly as a a online community. Yeah, yeah. It’s crazy. But yeah, I understand that because actually everybody kind of set up a thing. I feel I’m going to start dealing with the online and offline and not different things. They’re not different environments. It’s the people who are offline who are actually buying online. And as you said, they’re seeing multiple aspects of your product before they buy it, some of which are offline, some of which are online and when it’s not consistent… you’re [00:27:30.00] making a big mistake, Teodora, you were leaning forward to say something?
Teodora Petkova: [00:27:35.24] No, I was leaning forward to write a note, which I thought. The question is… because it looks like it boils down to how motivated a business is to talk to its customers. And then the question is, do we want to sell more by push communications or you want to talk and [00:28:00.00] serve more by creating an environment ready for poor communications? This is where the people pooling information. And the other thing is where you push information. And I know that this is not the mainstream idea of marketing, but on the Web, I believe this is the way forward. That’s why I leaned forward.
Jason Barnard: [00:28:28.00] Which is kind of ironic, because I remember [00:28:30.00] back in the day you would click on links and the idea was you would end up down this rabbit hole, of just clicking on things you thought were interesting, which is terribly poor. And Google and other platforms are going increasingly push, which is contradictory to what you just said and probably contradictory to human nature and what we’re all actually looking for. Can anybody expand on that silly ideological
Jarno Van Driel: [00:28:51.85] How I see this translate within businesses is… businesses tend to [00:29:00.00] ,when creating strategies, tend to have an inward view of things. They are looking at their needs and what they want and what they need of their customers. actually don’t even see them as customers. They see them as resources. And, that’s the nice thing of doing building connected marketing and actually care about the story you’re telling, and if you’re reversing that process, you’re big selling point is [00:29:30.00] not you are there for us. No, we’re here for you. And as vague as it sounds when we’re talking about using structured data and knowledge graph and semantic information for that, there is literally a whole practical side you can implement, which goes beyond the theory, but actually completely describes what the theory is saying about these technologies.
Jason Barnard: [00:29:54.17] The one thing that just struck me, it’s kind of ironic that we’re saying we’re going to use knowledge [00:30:00.00] graph to actually think more about the customer, less about ourselves. But would it be fair to say that the knowledge graph actually then extends out? Because the knowledge graph is just entities with relationships and attributes, that that relationship idea is actually terribly useful for marketing because the customer then becomes a relationship, to our knowledge graph through our products.
Jarno Van Driel: [00:30:20.34] In the multi touch marketing world, where attribution becomes more difficult every single day, it starts to become so important [00:30:30.00] that everything unites together in the way of business visions, express themselves or sells their product or whatever they do online. You have to have an amount – how do I pronounce that word?- a single form of communicating who you are, what you do, what you’re selling, why you do what you’re doing. And for that, you need to be able to connect the dots within your organization. But that means you have to look outward and see what the outside wants to have of you, as [00:31:00.00] opposed to only thinking about what it is you as a business needs.
Jason Barnard: [00:31:04.35] Yeah, I mean, those human beings were all the star of our own film and we carry that over into our businesses. Terribly egocentric. Just a really quick point, and if there are any questions to interrupt me, because I don’t have access to the chat.
Andrea Volpini: [00:31:21.63] Yep, yep, yep. We I think we’re monitoring no questions so far.
Jason Barnard: [00:31:27.15] Oh, but from Teodora, please can [00:31:30.00] you type in an answer?
Teodora Petkova: [00:31:31.68] I have an answer to a question that is not posed yet. I just wanted to say that there is this super thick thread between knowledge graphs and marketing communications, and it’s super hard to argument that in an academic environment, although I bang my head and continue on doing it. But the relationship is look at these relationships, meaning [00:32:00.00] knowledge. So this is the answer to another question
Jason Barnard: [00:32:09.52] That nobody asked, but it fits in with everything else that we’re talking about in that kind of whole philosophical aspect. Now, one thing you mentioned earlier is the GS1, which I know you’re really keen on, because you got this idea of saying we need unique identifiers because these different datasets mean that we’ve got lots of different identifiers that [00:32:30.00] are different than each department. And products end up getting mixed up maybe. And the GS1 is your super favorite solution to that problem.
Jarno Van Driel: [00:32:40.22] Oh, I didn’t imagine you guys wanted to run towards it 70 years already, I believe. All right.
Jason Barnard: [00:32:46.67] I’m going to sleep under a rock then
Jarno Van Driel: [00:32:48.13] The GS1 is the global organization that provides GTIN numbers, practical terms, bar codes and things like that. So they make sure [00:33:00.00] that every product out there on the market that, for example, get sold in the supermarket can be actually be scanned by systems because there’s a unique barcode on that product. So they’ve been providing a unique identifier for identifiers, for products long before the Internet existed and thus making retail possible online. I know from projects in the past that issues companies like eBay and Amazon have is that they had enormous data [00:33:30.00] sets without unique identifiers. So that, for example, things like kitchen utensils could lead to, let’s say, 70000 roughly unique products, which actually aren’t unique products where separate GTIN numbers and variants of one product group. But because those companies didn’t have a unique identifiers to extract that information you ended up looking for table [00:34:00.00] utensils in one of those systems, and you could scroll until you died of old age and you still wouldn’t be seeing the same table utensils. So those unique identifiers allows you to merge all your different products into actually the product you have, which isn’t so much the easy form of an issue for most of web stores, but [00:34:30.00] it’s an issue for organizations like marketplaces, auction websites and organizations like that. They need to make sense of all those different features. They’re getting all those products they are getting into their systems. And that’s where GTIN numbers greatly help you identify which product belongs to which product and which products are variants from each other. So that’s why I’m quite crazy, from that point of view, about the GS1, and next to that GS1 has its own vocabulary. Which [00:35:00.00] don’t let me sell it short, it’s not an extension of schema.org, but it certainly has taken it into account while they designed their vocabulary and they taken it a few steps further. And by now, things of that vocabulary are starting to drip back into schema.org , talk and actually taking each other into account all the time.
Andrea Volpini: [00:35:25.06] I have a question for Jarno. Would you suggest to use the [00:35:30.00]Gs1 product vocabulary today in conjunction with schema or would you rather prefer to wait schema to kind of adopt some of these new terms? I mean, one thing that I like about just one product vocabulary is that he starts taking into account things like sustainability, which currently I mean, we don’t consider much, but it will become more and more important, especially as we again, we get into the [00:36:00.00] physical selling the physical products and kind of creating a unified experience. What would you suggest to these?
Jarno Van Driel: [00:36:09.88] Most of the time when I pull in GS1 vocabulary is when I have data design issues. So, for example, practical issue, a website wants to have more faceted filtering on their category page, but they are not able to offer more simply because they don’t have all the types of properties and values [00:36:30.00] they need to expand on their categorization and create filters for those. The nice thing GS1 vocabulary, it goes into depth around a lot of properties of different product groups and types. Why is that interesting? Because often these also represent the facets you want to disclose in your category pages. Now, of course, if you are deeply committed to a certain type of CMS and the way it operates, then you’ll start [00:37:00.00] extending on that. But still you need to know which properties and values do I need to be able to create that filtering. And that’s where you look at the GS1 vocabulary and say, OK, here I have a food product. What types of properties are there for the properties and values are there for a food product? Which I can, for example, use to create more filtering on my website.
Jarno Van Driel: [00:37:22.11] So by looking at the GS1 vocabulary, is your method to expand your own dataset and create new possibilities within your organization [00:37:30.00] because you use it for that purpose, also using it to generate the markup? Yeah, that’s the added bonus because you already took the base design. So you because you have that information, it starts to become pretty trivial to actually start publishing that information. Now, whether or not search engines are looking for that information is a different discussion. Jury’s still out on that one. But you when you go to that type of advanced [00:38:00.00] level, most of the time it’s you start publishing that information, not so much for SEO purposes as for analytical and business intelligence purposes. So I can extend your pool of information again. You can get information and knowledge from data, like Teodora said. And so that can help your business grow, grow insights. Again, that doesn’t have to be one gigantic project in one go. You can do baby steps. You start with pushing [00:38:30.00] analytical information based on a schema markup. You already have a quite trivial job nowadays. And they you can start spending.
Andrea Volpini: [00:38:43.07] Jason is muted.
Jason Barnard: [00:38:45.26] I’ve got people banging away next door, banging nails into the wall. I was trying to save everybody. I was just thinking of a lot of those marketers and businesses tend to think, “oh, Google wants it and I’ll do it because I might make some extra money, get some extra traffic”. [00:39:00.00] And in fact, that that kind of internal business logic is like I need to build an internal knowledge graph. Whatever Google chooses to do is phenomenally powerful and, bringing in Jeannie Hill, who has insisted on the fact that we repeat “you will save millions in the end”. What an amazing tip and what a great way to convince people to actually start. But I agree, whatever the disadvantages today, starting to build today means that you’re going to be more [00:39:30.00] efficient and more effective in the future. And that will, I mean, on my level, will save thousands, but on a bigger level would save millions, which is very, incredibly exciting. And I didn’t realize, Jarno, that was quite that granular.
Jarno Van Driel: [00:39:44.67] You know, one of the reasons it is that granular is a lot of the GS1 vocabulary is used to exchange data between manufacturers and manufacturers need to have that tiny gritty detail level [00:40:00.00] to be able to to do what they do. So it’s from that point of view, it’s for a different need and for a completely different purpose. It finds its origin.
Jason Barnard: [00:40:15.87] Right.
Andrea Volpini: [00:40:17.95] We got a question for Jason. How important our knowledge graph panels for direct to consumer brands? And is it correct that [00:40:30.00 getting a knowledge panel, it kind of depends on the popularity of the brand or how much it has to depend on the structured data? And this is, Jason, your bread and butter
Jason Barnard: [00:40:46.23] It is my bread and butter. And there are multiple questions there. And I could easily talk for an hour about this all on my own and not shut up. So cut me off if I do go on. But, number one, wiki isn’t necessary. I think that’s kind of something a [00:41:00.00] dangerous mistake people are making. A, polluting the wiki world and, B, thinking that’s the only way. And what I’m saying is that Google has multiple verticals with multiple knowledge graphs, all of which have knowledge graph IDs attributed to them. So Google Maps as well on Google Books has one, Google podcast appears to have one. Google scholars, obviously, and also the Web. The Web index has a knowledge graph with knowledge graph IDs attributed to certain entities within it, images, too. And what [00:41:30.00] what I’m looking at here is Google is posting all of these little by little into the main knowledge graph. So that’s that’s a great trick. Those are great. If you’re in the vertical and you’re sufficiently well understood within that vertical, it’s actually relatively easy. I mean, I don’t say it’s fast. It’s easy. Simple process takes a lot of work and it’s not fast, but you can actually pull things across. I ported my sister, which is a bit weird, from Google Books into the main knowledge graph, and it was pretty easy and it didn’t require a Wiki page [00:42:00.00] and it just required providing the information, getting an entity home, which is vastly underrated as a lever to do this. Make sure Google understands where it should be, looking for the information from the horse’s mouth, then go going and corroborating it so that Google can say, I am confident enough to put this into my knowledge graph and I know I’m not going to be polluting the main knowledge graph with rubbish information.
Jason Barnard: [00:42:26.16] So that’s number two. And number three is the popularity of the brand. [00:42:30.00] That’s a question I get asked all the time. Wikipedia, wiki data rely on notability guidelines because they don’t want everybody and anybody in there who aren’t actually interesting to people, human beings, or worthy in inverted commas of a place. Google doesn’t care. It just wants to understand. So the idea of notability within the context of Google’s knowledge graph is not a concept you need to worry about. Google simply wants to understand everything and everybody in the world so everybody [00:43:00.00] can be in the knowledge graph. After that, “do you get a knowledge panel” is a whole different question, which is based on confidence, geolocation and the probabilistic the probability that the person who made the search is actually looking for that brand. Well, then you move into the idea if you have an ambiguous brand name, that ambiguity immediately becomes much higher. If there are multiple brands within the same region or multiple people the same region, Google will tend to go with the more popular one because it’s more probable. So I’ve answered several [00:43:30.00] questions really quickly. If anyone wants to learn more about that, just ping me and contact me. Thank you, Andy, for the question, I loves it, but actually Jeannie was asking a question about “pull”, which is for Teodora..
Andrea Volpini: [00:43:44.10] Yes, that’s for Teodora.
Jason Barnard: [00:43:44.31] Off you go, Andy.
Andrea Volpini: [00:43:47.46] Yeah, yeah, yeah. Jeannie is asking to Teodora, if, you know, working on data and making data more more clean, it’s creating, you know, the push. And so [00:44:00.00] it kind of letting marketer focus on the human relation pull, because the push is done by the data or the clean data. Is that correct?
Teodora Petkova: [00:44:14.94] I understand what Jeannie says, but this is not what I meant. I meant something [00:44:30.00] radically different. And it’s it will be translating analog days marketing into digital, literally, if we use a knowledge graph for pushing ads. I know this is not what I’m supposed to say., maybe, but this brings me to something I wanted to mention. And it’s regarding the value of knowledge graphs and [00:45:00.00] our role as marketing people. I want us to be braver in explaining clients that there is no one single square answer to things and working with data is also a creative endeavor. And again, everything should come from their urge to build [00:45:30.00] more relationships.
Andrea Volpini: [00:45:33.90] Right, in a way, a relationship, it’s also providing the means for a search engine to answer to questions, which is I think it’s pull. So if we focus on creating better data, we are able to provide more answers to the consumer without, you know, pushing the sale to the [00:46:00.00] other human at the other end of the line. I think that’s the idea that Teodora is displaying.
Jason Barnard: [00:46:10.95] I’ve just found the YouTube feed and I’ve just realized that I can do that. I hadn’t noticed before. So I’m now going to show that we’ve got some people watching on the YouTube channel with cups of tea and saying hello. And here’s a question that Andy is going to have to answer. His face is covered up… That Nik Ranger got from Australia. [00:46:30.00] Can you dynamically create new pages by means of invisible rule-based category pages? These are created based on the keywords pulled from either your search bar or what trends based in GSC data. Does that make sense to you?
Andrea Volpini: [00:46:45.50] Yeah, I mean, I don’t know… I would ask this question to Jarno because I… yes, I mean, technically, that’s what I’m planning to do. So we’re planning to create dynamic category of pages, like we call it in [00:47:00.00] WordLift. we have this research project with the University of Innsbruck and other organization, and we’re working on creating these dynamic pages that can go and provide information for a long tail queries. So we want to be able to create these dynamic pages and we want to be able to cluster content that fit with with such demand. So once we understand the search [00:47:30.00] demand and the idea of this prototype that we’re creating is that we create these dynamic pages that look at what Google is presenting on its results page, look at the entity behind these results page, and then try to see in the local knowledge graph what are the most similar entities to create a collection. OK, so that’s so we are actually doing what Nik is talking about. Will this work in SEO terms? [00:48:00.00] That’s for Jarno because we haven’t tested yet. I think it might, so we want to invest on it, but I don’t know if it makes sense.
Jason Barnard: [00:48:08.62] So yeah. Jarno, can you expand on that because Nick was asking the question to you and I thought I was to Andrea.
Jarno Van Driel: [00:48:14.45] I think what Andreas says absolutely makes sense, more generically for an e-commerce site might be, um, start looking at indeed your internal search data. What data do people pull in from the search bar? A little [00:48:30.00] bit more effort might be also get insights into which category filters people use. So you slice and dice for colors, to slice and dice for certain sizes of clothes. If you get all that information together, you actually start to see usage behaviour and what people are actually looking for. I often take that information to the people who do the ads and simply ask them, are there any terms here [00:49:00.00] you think you can generate good ads for that actually generate revenue? So I often take the shortcut before letting the ad people test all those variables and those variables that actually create success on ad basis. And we’ll actually unlock on site and start pushing through ads immediately and then you can build on that within your strategy. But if AdWords already shows you can create ads based on your place, you can directly make money without doing a lot of [00:49:30.00] effort. So getting that info in there, creating them dynamically, I’ve done so in the past as well. But that greatly demands from the type of pages that will generate automatically. I’ve seen that get out of control as well, where systems started to generate twenty thousand plus category pages per day. Well, that’s out of control then. Nobody can keep up with that, let alone create a good, clean graph of your site. So it depends on how out of control it might get if you do that automated.
Andrea Volpini: [00:50:01.00] Yeah, [00:50:00.00] yeah.
Jason Barnard: [00:50:02.80] Yeah, I mean, things gotten out of control with faceted navigation is already a big enough problem for a lot of websites. Adding this to the mix would be slightly concerning if it went, what would you call it, went wild? No, that’s not the right word.
Jarno Van Driel: [00:50:18.01] Went out of control.
Jason Barnard: [00:50:19.51] Thank you very much, Jarno. Your English is better than mine. Wonderful. I don’t know why you want to take this conversation for the last [00:50:30.00] ten minutes. I mean, I don’t want to bully anybody into talking about things they don’t necessarily want to talk about, from an e-commerce standpoint. I mean, Nik once again is asking about ROI. I think a lot of people like Nik are faced with clients and or bosses who are saying “I just want to make money”. When you were saying earlier on, 16 percent uplift in traffic and presumably sales for e-commerce website in two months and well, that’s not much. But most offline business, it [00:51:00.00] would take 16 percent in a year.
Andrea Volpini: [00:51:03.41] Yeah, I mean, it really depends on… I have to dig into the orders in this case to see, you know, exactly what it means plus 16 percent. But I think now it’s not so difficult to prove the ROI, if you are able to to measure properly things. The most complicated part is really to set the environment in such a way that you can control the metrics and [00:51:30.00] not just, you know, going after the numbers without meaning, but if you can do some basic experimenting and isolating pages where you do some experiments from pages where you do other experiments, then you come up with numbers that are quite clear in terms of, you know, what is the actual impact, where, you know what how is all of this impacting on the bottom line? As we said before, I mean, here we are going to one project to another because I mean, the [00:52:00.00] data is the same and then it can bring, you know, as we say, to an impact on a performance of the advertisement. But it can also be good for CEO and maybe can also help, you know, the logistics people, you know, get their stuff around at a faster pace. So there is a lot of key advantage and different type of ROIs that you want to measure. But I think now proving ROI of schema, it’s pretty easy. I mean, if we compare it to, you know, when we started, [00:52:30.00] I mean, right now it’s a dream. You know, there is so many carrots that Google is throwing at us, you know, for doing this or that or, you know, that rich result. I mean, that the end game, we know it’s not just that. Right. So that’s kind of the beginning of it. But it’s so easy now to get these positive ROI, if that’s what you’re looking for. You know, if that’s what the boss is looking for in the short-
Jason Barnard: [00:52:56.36] And moving up kind of philosophical point. I mean, a lot of people are saying or [00:53:00.00] some people are saying, Google throwing these carrots so that we will implement schema markup so it can train its machines and then it can drop the schema markup and get on with it itself. And it won’t need it anymore. Jarno, I’m pretty sure you’re going to disagree with that.
Jarno Van Driel: [00:53:13.60] I’m sure Google would love to be the Star Trek computer. Surely that’s going to be their big aim at the end. But we’re a long way from there yet still obviously structured data in part is used to set [00:53:30.00] up machine training sets and there are areas where it’s being used. But pragmatically looking, OK, Google wants to compete with Amazon, so Google needs to create a shopping environment in its organic search results. You know what Google needs for that? Properties tipes values to be able to create categorization, to be able to create faceted navigation. And Amazon already has that information and Google wants to have it. So [00:54:00.00] just even looking at keeping it to something practical like that, there’s a lot at stake for Google to get this right or they’ll definitely forever lose from organizations like Amazon and Alibaba,
Jason Barnard: [00:54:14.40] Which brings a couple of points, one of which is Google’s giving us the carrots. And if we take the carrots, Teodora, we’re actually going to end up building our own internal knowledge graphs and making sense for ourselves within our organizations and being able to use across the organization. And [00:54:30.00] as Jarno said, save millions.
Teodora Petkova: [00:54:33.97] Yeah, we’re going to enter the endless space of the giant global graph. No matter how the dark, how small the door is, this flow will burst the door. So don’t worry.
Jason Barnard: [00:54:50.85] Yeah, I mean, go with the flow, as it were. Build your internal knowledge graph. And Jarno, are you suggesting that GS1 is a good way to have a focal point within [00:55:00.00] the specific aspect of products to say this is my product, it’s unique and I can build my knowledge graph with reference to GS1.
Jarno Van Driel: [00:55:08.34] And especially if you if you’re starting to have the need to expand the product line products you’re selling, start looking at GS1 and the properties and values they have, because and often they form a wonderful basis to start expanding your business. Then at least you have an idea of what type of information you need. The biggest issue for small companies optimization is making sense of the gazillion [00:55:30.00] different product features they get from manufacturers. Well, if you actually can create a organized template to merge all those feeds through that template and create your own general feed, then you always have to start building a general knowledge graph. You’re still just using XML. You’re passing a bunch of different XML files. You’re passing through some logic. And, voilà, you have an XML knowledge graph. That already can help a lot because a lot of different systems eat XML. And [00:56:00.00] they are still far away from the abstract worlds of rdf and owl and triple store databass. Now, simply create an XML feed and store your data in there. That already does miracles.
Jason Barnard: [00:56:13.07] Right. Yeah, and one question there, because with schema markup, I started to look at schema.org and think if somebody or a group of people have thought that attribute is sufficiently important to have in the schema vocabulary, it must therefore be more useful [00:56:30.00] to the real world. So the other people and therefore I can actually start reflecting on my own business according to what I’m saying, in schema.
Jarno Van Driel: [00:56:37.46] I mean, you know, you’re absolutely right, but you have to pay attention because you can give a volcano a telephone number. And they’re due to inheritance of classes. Some things will have certain properties that won’t make sense at all for the specific type you’re looking [00:57:00.00] at. But because a volcano is a subtype of a place and places can have a telephone number, a volcano can have a telephone number as well, when you’re looking at schema.org, always make sure to be awake and realize you ask yourself “does this property make sense here for my case?” But also from schema.org perspective, don’t apply everything without thought.
Jason Barnard: [00:57:22.97] And what’s at the top of the page is what’s been created specifically for that type of entity or that should be. And so that would be [00:57:30.00] the one you would tend to look at first. And the further down the page you go into schema.org, the more general it is, possibly the less applicable it will be. Don’t be naive like me and just think, oh, this is a good idea, I’ll get my volcano telephone number. I like the example.
Jarno Van Driel: [00:57:46.34] So it’s a good example from the early days of schema.org.
Jason Barnard: [00:57:51.20] OK, I thought you were being incredibly original and funny.
Jarno Van Driel: [00:57:54.65] No, no. I said documented example on numerous occasions [00:58:00.00] on W3C mailing lists.
Jason Barnard: [00:58:03.89] Brilliant stuff. So we’re coming to the end. So can we have some closing comments? We’ll start with Andrea. Closing comments on what we’ve gone through and what your feelings are.
Andrea Volpini: [00:58:13.73] Yeah, I mean, I think it’s good. We discuss why we need to focus on data quality and the “webniness!, if I got it right from Teodora. So this idea that we start connecting things and that’s the [00:58:30.00] way in which we create an experience, I think that’s kind of for me, you know, the main idea is that we focus on connecting dots. That’s the point.
Jason Barnard: [00:58:44.81] Absolutely brilliant. I love that. Teodora, do you want to make your closing comments as though this were a court case, but it’s actually a PHD meeting.
Teodora Petkova: [00:58:53.84] Well, I always do. I just try to say something profound. Now, really practically [00:59:00.00] when we think about content, let’s think about data also and let’s work with people like Jarno and like Andrea and make that marketing thing a creativity burst, which doesn’t have to be the cool copywriting thing you can write down. But it can also be the creative way to engage people through data.
Andrea Volpini: [00:59:28.34] So we can add the phone number to the [00:59:30.00] volcano, and, in the end, thanks to this creativity…
Jason Barnard: [00:59:34.28] Teodors’s advices: work with Jarno, work with Andrea, don’t work with Jason (laughs). Jarno, can you pull me out of the fire here, please?
Jarno Van Driel: [00:59:42.68] Jason is much more eloquent in the way he expresses things. I’ll probably rub your toes in the process.
Jarno Van Driel: [00:59:49.85] I’m a blunt guy from the Netherldans. And so, yeah, if you want somebody who is actually nice to you, go to one of those three without me.
Jason Barnard: [01:00:01.39] So, [01:00:00.00] concluding comment from Jarno, schema, structured data and GS1 for products..
Jarno Van Driel: [01:00:08.39] Don’t be intimidated by the idea of building knowledge graph. It starts with very basic steps. You can build a knowledge graph in an Excel sheet. You can build a knowledge graph in an XML feed. Just pick a point, pick a point somewhere in your business where the biggest need for the organized data is and start organizing that data. And once you’ve done that, [01:00:30.00] start building on that and make sure you hook other systems into what you’ve built. It could be an XML file right now. Maybe it’s an mysqll database in the future and maybe it’s an entire Amazon cloud for 10 years from now. You don’t know, but you need to start as early as possible creating organized data and don’t build multiple datasets, merge it all together.
Jason Barnard: [01:00:53.79] Absolutely brilliant. That was genius conclusion. Thanks, everybody. I’m just going to show this, which is a special [01:01:00.00] offer for WordLift E-commerce SEO, 40 percent off… Get more of your products on Google search. Andrea is saying it’s a lot. I think it’s a lot
Andrea Volpini: [01:01:09.96] That hasn’t been approved. I mean, who did this, Jason?
Jason Barnard: [01:01:18.57] Ops… well, I’m not part of WordLift, so I’m allowed to offer 40 percent off. It is actually a serious offer, if anyone wants. Actually, Jarno, thinking about what you said. I mean, I’ve been building [01:01:30.00] Kalicube Pro, I build out a relational database, which is incredibly complicated because I don’t know how to do big data and I’m beginning to think it’s not a bad thing because it really means I’ve had to organize everything in my own brain and not rely on a machine to do it all for me. And I can actually explain my database to people and I can find things in it and I can dig data out that I’m quite surprised that at times and that just comes down to sitting down… actually, it comes down to waking up at three o’clock in the morning, pretty much every night and thinking, oh, I could reorganize it like that, and move that there and add [01:02:00.00] this there. And it’s been an incredible experience, so, Kalicube Pro, incredible experience for me. Jarno, Andrea, Teodora, thank you very much. I thought that was absolutely brilliant. I learnt loads. And you’re all delightful,
Ciao-ciao.Thank you. Awesome. Bye bye, everybody.