Podcast: Play in new window | Download
Subscribe: Apple Podcasts | Spotify | Amazon Music | Android | Youtube Music | RSS
For the past eight years, Katariina Kari has built knowledge graph teams at giant e-commerce companies like IKEA and Zalando. This practical, real-world experience puts her in an elite group of ontology and knowledge graph experts.
Knowledge graphs offer unique benefits to e-commerce merchants. From better product recommendations to more useful search results, the semantic capabilities that knowledge graphs provide routinely result in seven-figure sales increases.
The knowledge graphs that Katariina builds provide a semantic layer in the enterprise architecture that lets companies capture, use, and re-use the organization’s unique domain knowledge in any number of applications.
Because knowledge graph isn’t one application that does one thing. It’s a paradigm shift in the way we work with data. It’s a paradigm shift in the way we code, because now you don’t need to put business logic into your code.
We talked about:
- her work over the past eight years building knowledge graphs at companies like IKEA and Zalando
- how using knowledge graphs to improve reccomendation and search routinely brings seven-figure business benefits
- the set of skills and talents it takes to implement a knowledge graph project, most of which already exist in most companies
- how LLMs and other AI tools can help transform structured or unstructured data into semantic data, a computable resource that captures business domain knowledge
- some of the specific skills needed for KG work: ontology experts, back-end developers who understand the semantic web stack, data scientists and engineers, knowledge practitioners to capture domain knowledge, and product management
- the need in each organization for a unique knowledge graph team tailored to the needs of the company and the talent available
- the importance of user-centricity and use-case understanding in any knowledge graph project
- the benefits of capturing business logic in a semantic layer which can be used and re-used in multiple applications
- an interesting search-improvement use case that resulted in seven-figure sales increases, as well as experience-improving recommendation and info-box use cases
- how capturing subject matter expertise in a knowlege graph can dramatically improve recommendation systems and deliver unexpected benefits to other
- the importance of showing the benefits of knowledge graphs to organically advance enterprise adoption
- her take on the difference between RDF-based knowledge graphs and labeled property graphs (LPGs) like Neo4j
- the compelling case for knowledge graphs in e-commerce, which she has discovered in her eight years of practice
Katariina’s bio
Katariina Kari is a leading expert in semantic web technologies, specializing in the development of ontologies and knowledge graphs. Over the past eight years, she has worked with prominent brands like IKEA and Zalando, building knowledge graphs that significantly enhance customer experiences by improving search functionalities and recommendations. Her extensive hands-on experience in creating enterprise knowledge graphs has established her as one of the global top talents in the field.
Katariina is frequently invited to speak at international events on the semantic web and knowledge graphs, sharing her insights and practical expertise with industry professionals. Her deep knowledge and passion for the semantic web have made her a sought-after keynote speaker and thought leader in the field.
Balancing a dual enthusiasm for technology and the arts, Katariina holds both a Master of Science degree and a Master of Music degree. From 2012 to 2016, she ran her own consultancy, where she worked closely with classical music organizations and artists, helping them navigate digital outreach. An art-loving and art-serving nerd, she seamlessly blends her love for music and technology in all her work, constantly pushing the boundaries of what’s possible in her field.
Connect with Katariina online
Video
Here’s the video version of our conversation:
Podcast intro transcript
This is the Knowledge Graph Insights podcast, episode number nine. One of the main benefits of semantic technology is the ability to sort out business logic independent of data and data from the applications in which it’s used. Katariina Kari has captured in knowledge graphs the business expertise of e-commerce giants like IKEA and Zalando to power better search and recommendation systems and to generally provide a better experience for both internal users and external customers, resulting in millions of dollars in new sales.
Interview transcript
Larry:
Hi, everyone. Welcome to episode number nine of the Knowledge Graph Insights Podcast. I am really delighted today to welcome to the show Katariina Kari. Katariina is a long-time, deeply embedded in the community knowledge graph professional. She’s done a lot of e-commerce work at places like IKEA and Zalando. She’s currently the head of data at a stealth internet, a stealth startup that we can’t talk too much about. But welcome, Katariina. Tell the folks a little bit more about what you’re up to these days.
Katariina:
Yeah, thank you Larry, and thank you for having me in your podcast. Yeah, I could say that I’ve been really lucky to have worked in the industries and especially in the lifestyle sector, e-commerce sector for the past eight years. So very early on, before even graph databases were really commercialized, I had the opportunity to start building knowledge graphs. And so now, when there’s someone out there was like, “Oh, I need a knowledge graph,” I can confidently say, “Well, I’ve done it a few times so I can tell you, I can advise you or I can even run a team for you that can build the knowledge graph.” And it’s just shown me a lot of practical things, it’s given me a really good perspective on what works from theory and from research and what actually doesn’t, or doesn’t yet work, or isn’t mature enough yet for an applied setting in commercial use.
Larry:
That point you just brought up, you have a PhD in something, right?
Katariina:
No, no, I don’t have a PhD. I never really went into research. I have a few published articles, scientific articles that I did towards the end of my master’s, but I actually just have two master’s. I have a master’s in technology and then I have a master’s in music arts management, because I’ve always carried this love for both art as well as technology, and I always wanted to combine them.
Larry:
I didn’t plan it this way, but the episode right before this one, number eight, was Vera Brozzoni, who’s a metadata strategist at the BBC, and she comes out of classical music. So if you don’t know Vera, you two have to meet and talk music and data and stuff. But one of the things you mentioned there is one of the things in this community is, and the reason I assumed, it’s usually safe to assume in this world that somebody has a PhD, but you’ve always been more focused on the practice side than the research side, which is awesome, because I’m all about sharing practice, so thank you for being focused that way. And you have all this experience coming up on, what, eight years of experience at IKEA and Zalando, two of the biggest retail brands in the country, or in the world.
Larry:
And one of the things, and I want to gently take you to task for something in a talk I saw you do recently, this hour-long talk, brilliant stuff about a lot of this stuff we’ll talk about today, but right in the middle of that talk, you just kind of matter-of-factly mentioned like, “Yeah, and we’re realizing seven-figure business benefits across this.” I’m like, “Wait, what?” And in my journalism training, we would call that burying the lede. But that’s kind of at a top level… There’s real obvious business benefits to adopting knowledge graph technology. Can you talk about what is it that’s unique about the work you’ve done and this technology that permits these massive revenue gains?
Katariina:
I would say that one part of it is the work, but the other part of it is working with big brands like Zalando, Europe’s biggest e-commerce fashion, and then IKEA, one of the most known trademarks or brands in the world. So their volumes in e-commerce are huge. So if you add a positive change to the customer experience, like giving quality recommendations or just improving a few of the worst-performing search terms, you get a lot of… The volume is so big, the benefit is really big, and it’s already in that category of seven-figure sums. So that’s why I think maybe not every little e-commerce site can invest in this technology first. It’s great that these big brands are actually investing in so we can figure out exactly how to do it, and then that can be brought to maybe a smaller-volume e-commerce setting or smaller-volume industry so that they can then make sense of these best practices. That’s at least the way I see it. But yeah, I mean just being able to improve a big website’s performance, just doing little optimization is already moving the needle quite a lot.
Larry:
Right. And once you’ve articulated those best practices, you can picture it, smaller businesses benefiting from it. But right now, it takes quite a team to put this together. I’ve heard you talk a lot about the skills that it takes, the roles that you need to execute on those skills, and then the human element, the thing that our friend Ashleigh Faith calls the data therapy part of this. Can you talk a little bit, I guess, first, about what does it take, what are the skills that you need, the knowledge, the wherewithal in your organization to actually make a knowledge graph project like the ones you’ve worked on happen?
Katariina:
Well, when you work with these industries in e-commerce, they’ll probably have already very brilliant backend developers, and backend developers I think are the most versatile adaptive developers out there. I’ve seen them adapting new technologies really quickly, even into RDF or this whole knowledge graph semantic web technology stack. So you probably will have those already in there. And then the only thing you need is maybe one semantic web professional, like an ontologist, just to bring in the whole RDF and educate the backend developers so that they know how to spin up a graph database and do the data pipelines for semantic ETL or semantic transformation of the data. But then, yes, there’s that human element that process together with the domain experts that Ashleigh Faith calls data therapy, and we agree that we should call it data therapy. And that is something that is implicitly kind of done in companies by a product manager.
Katariina:
So there’s any way business and tech, any way need to talk, some engineers would embrace domain-driven design to kind of get that out, but it’s all stuck in code. So when you do it with semantic web technologies, or you have your semantic web professional, they’ll be formalizing that leading the process of getting the domain expertise out and then, instead of locking it away in some siloed bit of code, they’ll actually make it into a computable resource within the company. So that’s the difference. And what I’m trying to say is most of these things that you need to build a knowledge graph already exists at the company. Just need a professional for the semantic web bit like one person to then put it all together.
Katariina:
Then what takes more time, more resources is when you’re first setting up the entire infrastructure for housing and knowledge graph. I think that’s an investment that needs something like one to two years to build up. But after that, for example in Zalando, I think the infrastructure is there, doesn’t really need team or needs little maintenance currently. I don’t know, I don’t work at Zalando anymore, so I cannot say for sure, but that’s how I understood that it’s now in the background running, and just needs a little bit of maintenance.
Larry:
You’ve just reminding me of the thing that’s everywhere now, the role of LLMs in this new world. Because one of the things about that both… that that’s been an obstacle for a lot of people, for a lot of organizations that one- to two-year up-ramping. But it sounds like from talking to if you, just reading, just having been at the Knowledge Graph Conference the last few years and reading, following people like Tony Seale and Mike Dillinger and you on folks on LinkedIn, it sounds like that the LLMs, or AI in general and LLMs in particular, are helping maybe to accelerate or… Is that cycle of one to two years, is that any tighter these days? Because so much of the tedious work of that seems like can be delegated to these.
Katariina:
That’s true. One part, and I think that’s still being discovered currently, so it’s not something we could go out and report about and be like, “Oh, this is how you’re going to do it.” So everyone’s still figuring out or having their first promising results on how that could accelerate. But there’s definitely the element of transforming structured data or unstructured data into semantic data. So semantic data meaning it’s computable, it’s a computable resource that also has the business domain knowledge in it so that you can do really versatile things and combine your data in different business-useful ways. My bet is it’s going to accelerate the transformation part of the ETL pipeline, the semantic ETL pipeline, which is extract, transfer, and load. Or transform, actually, not transfer, transform. And other places where LLMs could accelerate, perhaps also the human process of… or at least validating.
Katariina:
So if you’re in there with your domain experts, and you’re trying to understand all the central concepts of your business and how that would look in data, then you can see if you’ve missed something or fill in the gaps or verify your outcome, your human process outcome with LLMs. So that can also give a little bit of assurance. But I think I have to say this is all… it’s a little bit hypothetical still. I’ve tried a few things out and see this seems to work, but I think I would want to report back on it in a year or in two years.
Larry:
Yeah, no, and you’re also reminding me, we’ve had a number of conversations where you’ve punctuated the conversation by saying, “Larry, that’s just what an ontologist does.” So there’s still plenty of work for human beings in this process. Well, and speaking of that, you mentioned most of the talent that you need still is often in an enterprise. You just need that semantic web expert kind of guiding and organizing things, but there’s a lot going on in this. You’ve mentioned the ontology development, ETL pipelines, but there in the talk you did, you talked about… I think there were 10 roles that you mentioned in a talk you did reflecting on this. I guess just generally, maybe not all 10 of them, but what’s the general… There’s the ontological understanding, there’s the data sourcing. What else is going on?
Katariina:
Yeah, there’s definitely, understanding semantic web, so knowing how to design ontologies, then how does the semantic web stack work, so that’s more like specialized. So a backend developer could develop itself or themselves to knowing about the semantic web technologies, but then you still need a bit of an ontology brain to really understand how to create ontologies, so you need to understand description logic. But then the other one is data. Because you need to build data pipelines. That’s not going away. You need to figure out how to load a document, parse it, work with it, do things with it, and then serve it performantly to customers, all the way to customers, whoever those are, if they’re e-commerce customers or if they’re just internal customers. And then you do need the domain knowledge. Yes, and I would say the fourth axis here is product, so knowing how to actually prioritize the things you’re doing.
Katariina:
And then I kind of got that into ten roles, because I was using different combinations of things. I was saying that, “Well, if you combine ontology knowledge and domain expertise, then you have this person. If you combine product knowledge and domain expertise and you have this. Or if you combine product and tech, then you have this.” And so I was just kind of combining them and being like, “You can have these kind of people,” because my experience is also that you still need to build your team case by case. You need to see what talent you have and where these people are willing to grow to.
Katariina:
So for example, I had in one of my teams, I had a person who’s really into engineering and has a PhD in semantic web, but they didn’t really like the social aspect of the semantic web thing. So they’re like, “I don’t want to sit with domain experts in a meeting room arguing about where this class should go.” So then that person clearly is more like a knowledge engineer type of person. Whereas there was one who’s like, “Yeah, I like to work with data, but I don’t want to stare at my screen all the time. I also want to work with people.” So that person clearly is more an ontologist. Same set of skills, same background, but then personally fitting better into different kind of roles.
Larry:
That’s interesting. So there is just that as a leader, if you come with that semantic web expertise and you look around and you go like, “Oh, all this great talent,” and then that’s a huge part of that fine tuning is like, “Oh, he’s not going to talking to people. She’s going to love talking,” whatever it turns out. Tell me, did that team setup that you’ve talked about, does that reflect the actual life you had at IKEA and Zalando? Did you have teams about those sizes or…
Katariina:
Yeah. Yes. It depends on the use case. Because these teams, I was always building knowledge graphs together with building use cases. So we would put up APIs that would then serve this information further to the e-commerce. So it was always also a part of building up a use case within the company. And we needed to sell the thing as well. We couldn’t just say, we’re bringing you the most perfect optimized knowledge graph infrastructure, and then you’ll figure out how to use. We can’t do that. So in order to make a knowledge graph project successful in the company, you need to have your first low-hanging fruit use case, probably has a little bit like, not the most fantastic description, logic semantic web behind it, but has more simple semantics behind it. And then you’ll gain traction within your company and you’ll start doing more.
Katariina:
So ultimately I would say a knowledge graph team within a companies like a startup needs to figure out what the company needs today and what other teams are not able to solve. So we’re like, “Okay, we can definitely solve that with knowledge graphs.” Because knowledge graph isn’t like a one application that does one thing. It’s a paradigm shift in the way we work with data. It’s a paradigm shift in the way we code, because now you don’t need to put business logic into your code. I was talking about those code silos. You don’t need to log it away in there. You have your – maintain a semantic layer in your company that has all the business logic, or some business rules, whatever, and then the applications make use of that and reuse that information.
Katariina:
So I think Dave McComb from Semantic Arts has talked a lot about data-centric revolution, and it’s not about centralizing data, it’s about having data at the center and having the metadata at the center so your semantic layer, and then applications become more lightweight, because they don’t have to rewrite over and over again the business logic into their code. They don’t need to create their own data silos for the application to work. They’re more like reusing the data from the data layer.
Larry:
Yeah. That generic benefit, it seems like… I’ve read Dave’s book and follow him closely, and it seems like the benefit is so clear, but so many organizations, they still just source their own data from wherever, whenever they build a new application. I see the same thing in the content world that I run in. But that’s interesting. But you were just saying too, that doesn’t have to be one centralized data source. You just need that semantic layer that helps you understand what you have and can work with.
Larry:
And you mentioned a minute ago, I want go back, because so much what we’re talking about I realize might be abstract to some people, and I want to ground this in some of those use cases that like, where does that million dollars of benefit come from? And so some of the use cases you’ve talked about, like one, you talked about an example of search disambiguation and search terms around like there was when a Beyonce launched a fashion thing when you were at Zalando and you’re like, “What’s going on? And how can we help sell more products with this knowledge?” Can you talk a little bit about that search use case and maybe one or two other use cases?
Katariina:
Yeah, so the search use case was really fun. That was actually the low-hanging fruit that we tackled in Zalando. And it started from our search experts being confused why suddenly Beyonce is this really popular but very low-performing or badly performing search term in Zalando, and that’s because Beyonce launched Ivy Park brand, and our customers couldn’t really remember what that brand was called, but they certainly could remember that it’s a brand by Beyonce. So obviously humans are being humans, they’re writing Beyonce into our search. And then the search experts were like, “Oh my God, we need to now do an override for all the combinations that we can find for Beyonce and for all the languages, all the 20 plus web shops that we are running,” because we’re a European e-commerce site, so every country has their own e-commerce site. And then we were like, “Well actually with semantic web you could disambiguate this.”
Katariina:
So we just do one connection of Beyonce to Ivy Park, and every time your search hits Beyonce, it will return you all the Ivy Park products. So we can do this also with not doing this manual overrides, but actually adding that knowledge of Beyonce design, Ivy Park launch, Ivy Park is associated with Ivy Park, if you can express it that easily. And then we used that to argue, and obviously at that point, the manual overrides were already done, so we never actually tackled Beyonce, we just used that as an argumentation of you could do this a little bit simpler. And then we tackled the top 20 of the low-performing search terms that were most popularly used by our customers. And I remember us estimating if we solve these, we will make a seven-figure sum, even after returns.
Katariina:
And then we started tackling them, one of those search terms, winter. So winter was performing badly, and what we did is we explain with the semantic web layer with the knowledge graph. We explained that winter means find me all products that are filled with down feather, warmly padded, made out of wool, using some of these trademark warming technologies, they’re like tens of those trademark technology, every brand has their own, and then just basically connecting winter to product trades.
Katariina:
And that helped us to then find all the products, and then suddenly, through A-B testing of having a variant of searches performing in the traditional manner, search is augmented with knowledge graph information, we could see that the B variant was performing so much better and our customers were clicking happily and interacting happily with the search results for any winter-related search. And this we did for a few of these terms. And I think our team didn’t cost as much as we brought in money, so it was really on the plus side for the company as well.
Larry:
Nice. As you were talking about that, you reminded me of a talk you did at the Knowledge Graph Conference a couple of years ago about the recommendation engine that you built and how that’s… I think always because you were talking about, okay, winter, down, those things are related. And you talked about how you ensconced the subject matter expertise of the interior designers at IKEA to make better recommendations. That seems like another classic use case.
Katariina:
Yes, definitely. Recommendations is the other. I would say search is one recommendation, and the third one is info box. So for sustainability, if there’s a certain fabric with a certification, how they have these certifications, then you could serve an info box, saying this is what it means, certification, this is why we’re calling it sustainable. But the recommendation is really interesting. That’s also quite simple to solve, because for IKEA it was all about… IKEA has a very nicely honed process of how they’re displaying things in the warehouse, or sorry, in the stores. And so every interior designer who’s ever worked in IKEA knows these rules, and there’s not so many of them. So they have them in their minds. There’s a sofa or a couch that you can pair up with a throw or a decorative cushion to add more comfort or with a lint roller to get all the hair out, or to maintain it, to maintain its nice surface. And those rules we can put into a knowledge graph.
Katariina:
So we can say every time any sofa can be matched with any decorative cushion. And then in addition, we would add rules of prioritized matching the same styles, or prioritized matching the same price level. So these kind of things we could add to that. But that’s a handful of rules, and that’s like one designer, one IKEA domain expert just putting those connections. Before that, they would manually do that. So they would manually go through their tens of thousands of products that match them with each other. But with knowledge graphs, you can abstract that. You can say any sofa with any cushion plus these rules, and then that’s it. And it seems like a no-brainer, and it’s like, well, you could do it in Python as well, but then your logic of a sofa and cushion go well together is locked in Python. You cannot reuse it.
Katariina:
So what we had is we had our computer vision friends who are helping customers to find the right products when they are in the store and they just use an app to look at the things. They were like, “Oh my god, we can reuse this because this will make our algorithm of understanding what’s in the picture much more performant because we can already narrow it down to next to your sofa, you will most likely find these product, instead of going through the entire product catalog.” They know that, “Okay, this thing we can reuse.” So that’s the benefit of a semantic layer, that’s the benefit of a knowledge graph, that somebody is putting some smart stuff into the semantic layer. It’s not locked away in somebody’s Python code, and then a different use cases can reuse that same information.
Larry:
Yeah. I want to go back to people stuff now because convincing developers that that’s a better way to do it, because any one developer tackling a project is just like, “I can do that in Python,” and boom, there’s brroom, they just do it. How do you convey to individual engineers and probably more germane to an organization like, “No, no, no, there’s a better way.” Have you had success with that, or…
Katariina:
By showing… So you get that low hanging fruit use case, and then you just wait for the other team to appear and be like, “Oh, we can reuse this information,” then point proven. I’ve argued my case, so you just wait for it to happen. It happened at Zalando. The things we did for search was then later used for contextual recommendation of like, “Oh, you’re looking at a blazer. Well, here are other businessy clothes that you can look at.” And in IKEA it was, “Okay, we are tackling this recommendation.” And then quite quickly another team found out about it.
Katariina:
And I’ve talked to friends who’ve also worked in bigger knowledge graph projects like Google, and they said that that’s what happens. And people come knocking to your door and be like, “You have that knowledge graph, you have that information. Can I reuse it?” So it seems to go organically like that. And then even if you tell business about it like this is going to happen, they don’t really believe you until they… yeah, they need to see to believe it.
Larry:
That’s right. And there’s more stories like that floating around now because of the rise of knowledge graphs. But one of the things that comes out, if you do a Google search for knowledge graph, you’re as likely to find information about label property graphs like Neo4j as you are to find RDF-stack kind of stuff. I’m going to ask you to pitch the benefits of a standards-based approach like RDF, not that… Neo4j is awesome. A lot of people get a lot of value out of it. But in terms of the kind of work you’re talking about, can you talk about why the RDF stack is sort of the best way to do this?
Katariina:
Well, I would say that RDF Stack and LPG’s labeled property graphs are just different stacks meant for different use cases. If you want to do graph algorithms and calculate path distances, the label property graphs are definitely the way to go. And we’ve seen that machine learning interacts really well with this network of data. So that’s why I think it’s currently so popular. Plus Neo4j has fantastic marketing department. They’re really good at pushing these things.
Katariina:
When it comes to RDF, first of all, it’s a standard by a research community. So it doesn’t really have a marketing machine, doesn’t really have anyone thinking about its brand. So that’s the first part. But then what RDF does really well is the schema. So the ontology level, the reasoning layers, all the things, the description logic side of things. And it’s just a different approach. It’s more about thinking about the form and having that conversation.
Katariina:
So if you want to realize the semantic layer together with domain experts and have those business rules and have those associations of sofa goes well with cushion, it’s best realized with the schema, with ontologies, and then some SKOS concept schemes like you would call the vocabularies or taxonomies. So you kind of do that best. So the search disambiguation, info boxes, and recommendation as we did in IKEA, I would realize with RDF and with that well-defined part, and then probably any big-data calculations you need to do where having things in graph is a benefit I would do with LPGs like Neo4j.
Larry:
Okay. That’s helpful because I think a lot of people are fairly new to this, so that’s a really good distinction between those. Hey, Katariina, I can’t believe we’re coming up on time already, but before we wrap, is there anything last, anything you want to revisit from the conversation or just make sure we share?
Katariina:
I think my message I guess out there, or what I want to put out there is that when it comes to e-commerce, knowledge graphs make a lot of sense. And if you’re a big e-commerce vendor, you shouldn’t overlook it. You should really try to solve it other way because I just see it as solved. I’ve done it now for eight years. Now I’m working on other things in my stealth mode startup more to be continued to more information later on. But the past eight years have taught me that this is really the way to go. This technology works. And yeah. At this point, there’s like… I’d love to hear the arguments, but I don’t think there are any arguments why you shouldn’t invest in knowledge graphs if you’re running an e-commerce platform.
Larry:
Nice. Well, I appreciate you being out in the future figuring that out. Now it’s like settled science, everybody should just do it this way, and there you go. So thanks. One very last thing, Katariina. If folks want to connect with you online or follow you, what’s the best way to find you online?
Katariina:
Yeah, LinkedIn would be the best one, and my handle on LinkedIn, which is like linkedin.com/in, I guess, my handle is K-A-T-S-I. So that’s Katsi. That’s what my family calls me, what everyone in Finland calls me, because that’s a Finnish nickname.
Larry:
Okay. We’re all honorary Finns now in the world. Well, thanks so much, Katariina. Always fun to talk with you.
Katariina:
Thank you, Larry. Thanks for having me.
