Hi there everyone, hope you can hear clearly. We're just going to give guys one minute before we kick things off. So please just give us a sec. But you're looking at the cast today. So I'll run through the agenda in a few seconds. As you can see, this is what we've got today and a few people presenting the different bits. So intro is me. I'm talking right now. So I'm Tom, Lead Product Manager on the Product Discovery side of our business at Crown Feek. Then we will move on to the recent releases and roadmap update and the roadmap feature spotlights, which I'll have a part in. But then we've also got Yelena and Jan who lead our AI and EXO roadmaps respectively. So I'll run through what that means in just a sec. And then we've got Pavel on the tech stack modernization bit. So he leads our AI and cloud platform teams. So we'll be talking about the tech stack bit and how Crown Feek is evolving and how its technology can better serve you guys. And then the customer success use case from the wonderful Johan. So I won't give away any clues, but it's great to share compelling real life examples there. So looking forward to that. And then please stick around to the end because there's feedback from you guys. Let us know what you saw that you liked, what you want more of, less of, what you were expecting that you didn't see. And there's also a chance to get involved in some research topics. So please do flag which topics you'd like to be involved in. And we can put you on the list for those. So thanks very much for your time. All right, we'll kick things off. Thank you all for joining. Rodeman. Today you'll hear a couple of bits. You may hear mention of Fred Hopper and EXO. So just for those that don't know, we have two key products within our product discovery offering. We've got Fred Hopper, which focuses on the scalability and enterprise solutions for international customers. And then that works really closely with EXO on our more advanced nations use cases. You'll hear about both. I wanted to make sure you're all clear on that. Then we'll look at the focus areas. So the three pillars that we're looking at this year is testing and reporting. So you can have all of the tooling in the world. But if you can't measure the success, if you can't A-B test it, spot opportunities, come up with hypotheses to test, then it's very difficult to get the most out of those tools. So that's a big focus area. Accelerate Gen AI and AI use within the tool. So there's multiple areas we're looking at here. So you'll see quite a few of them today. We'll delve into those, but I'm very excited. Hopefully you'll see where we're incorporating those into our technology stack more and more over time. And then the final one is about connecting to a different e-com platform. So lots of people are changing. There's lots of new ways of connecting. So we're looking at different iPaaS vendors, integration platform as a service, and plugins where appropriate as well. So if you are thinking of moving, please do keep us in the loop because there could be something that you haven't been told of yet. Because we just didn't know it was relevant for you. So keep us in the loop and we'll make sure that you have all the information you need. So here's the product update. Quite a few items here. So I will just touch upon some. This is the recent release that you guys will have seen. So there's a lot going on here. I'll just run through the blue bits. But complete replatform to Java 21. Not many companies have done this. We just looked it up and I think only about 25% of people have managed to make this move. So this is really pushing our tech forward. And we've seen huge improvements over 25% performance improvements on quite a few setups. So we're really excited about that. That can help with your queries, response times, but also potentially dashboard responsiveness. And we're even looking into indexation times as well. So there's a lot of good news there. The blocking facet values as well will be rolled out. So if you guys have certain facets on certain pages that you want to show, on some pages and not on others, that is now possible. So you can set those up and block them. Even if they're in your data feed, you can hide them. The save and publish functionality. This is for anyone that's tried to save something with an item or a facet that doesn't exist at the time. So you've moved something to position one. That goes out of stock and is out of your data feed. You can now still save and publish that rule without causing any issues. So I know that that was a frustration for some before and that just makes your lives hopefully quite a bit easier. And preview on live is now the same as preview on test. So you guys should be able to see which result modifications have been applied there. And this amongst many other reasons, we've got security features, scalability, resilience, anything that we provide. We were noticing that people were getting these updates quite late because it was triggered from the customer side saying we would like to update now and merchandisers and the business in general was missing out. That's why we started the automatic upgrade plans. So you guys should see these upgrades coming in at regular intervals moving forward. And that will be quicker and quicker as time goes by. This update, for example, is mid rollout. So you guys should all be seeing these very soon. Additionally, there was a big request to look at the new insights dashboard. So there was lots going on here. As I said, the focus is on insights for one of our main pillars. This went out in early Feb. It's part of a continuous improvement plan, though. So you will see more and more happening over the year and beyond. But had a huge UI uplift. There's a clearer dashboard. You have trend analysis now is the big one. So over any metric in any line item in any reports, if there's rankings or result or item campaigns or searches, whatever it is, you can see how those are trending over time. Maybe, for example, your conversion has dropped off on a certain ranking. It could be that you notice that by the trend over time. You want to find out where that's come from. If it's from click through rates or if it's from after it's added to the basket, are we losing them in the basket? Because then the remedy may be a little bit different. For example, if it's through click through rates, you could then go have a look at the front end, see that maybe the merchandising is stale and it's been the same for a while. Or maybe higher price items are coming through. So you can change the strategy and then go and test it with A-B testing, which has also added a facelift. So in here, you can see this trend line over any metric again in A-B testing. So this is great because if you see that conversions are winning one, but click through rates are higher on the other, let's see if that's consistent because maybe it's not. Maybe one's winning on the weekend and one's winning in the week. Or before and after payday, you're seeing a very different trend. And this will help you analyse that because it could be that A-1 overall, but really B is better for the one week after payday. So eking out all of those final optimisations in there. And we've got the roadmap. So here I've already gone through the green bits, or if I haven't, it's a clue. It's coming up on the next slide. But we've got lots of other bits here as well. Just quickly run through the now and next column and then later we'll probably do next time. But the customer alerting and tracking, this is going to be run through on the XO roadmap because we're combining a lot of things together. So I'll leave that one. But we've got the connection between Fredhopper and XO. So if you are using both of our tools, you can now seamlessly switch between them. We already have the shared pipelines for all of the data. This now enables you on the front end to have a combined view as well. And you'll see more and more and more of our deliverables will be for both sides. So that's exciting. We're really getting a lot of good cross-pollination in there. Make sure we get the best of both. We've got rule-based product tagging in here as well. If this is one you guys are interested in, then please do let us know. There's lots of things that this can do. So for example, if you have data that's hidden in a different attribute, so it's in the description, you need to get that out. Or if you just want to tag a thousand products with a certain tag, this can help there. Or if you want to map any data to something else that's neater for a facet. Or if there's any rule-based stuff, anything that came live within the last month, let's tag that as new in. These are all of our Black Friday products. Whatever it might be, this can potentially help with all of those things. So please do raise those with us and we can get you on the list there. And then the custom alerting bit, again, I'll let Jan run through that because that is the initial testing is happening on the other side and then we're porting that over. On the feature spotlight bit, I'm sticking very much on the insights trend today. So we've got multi-site and filtering. So at the moment, if you have 10 different websites, you may have 10 different views. The idea here is that we can combine them all into one so you get a global view. You'd still have the ability to filter to say only show one country or one, however you split your data, but you can also get the global view. And not just the filter, also splitting out on those different countries. So you can compare France to Germany, to the Netherlands, to wherever, and even cluster. So you could say, let's see the Nordics. That's the initial one here that we're really excited about. But then this also unlocks, because we have now built the ability to split and filter. It's also then in a follow-up project, we can add all these other things that I've listed. So we could do per category rather than just on the ranking. Maybe the men's tops ranking is working really well, but it's working for t-shirts, but not for polo shirts. And this will help you identify those and you can compare them easily. It could also be for anything that you provide in the tracker. So if that's important, please let us know and we can make sure that that's prioritized in the roadmap. Like if they're a new or returning customer or which search pass they've hit, or if they've engaged with a certain product type, let's separate those out and see if that's changed anything. If they've clicked on a certain facet and then we've seen that that's affected their whole journey afterwards, or that's when all the conversions come through is after clicking on that. All of these different filters that we can apply at the moment, we're considering lots and we'd love your help with prioritizing those. So if any of those stood out or if you've got any other examples, then please do back them here. But obviously the outcome we're pushing for is more visibility on opportunities and to understand what's under and overperforming so you guys can take action. And over to Jan. Thank you, Tom. So basically if you're not familiar with XO and more familiar with Fredoper and you know that we are passionate about helping your shoppers finding what they need quickly and intuitively, but discovery doesn't end with search and PLPs. And that's where XO and that's where XO and our advanced recommendation and personalization engine plays a key role helping shoppers discovering the right products even when they are not exactly sure what they are looking for. So before we dive into the next roadmap session, slide, sorry. I want to quickly work through what we recently released in terms of implement and features. So here you have the XO and so here you have the XO Chrome extension. Now on manifest V3. So we have this Chrome extension. For those who don't know, this tool that helps merchandisers understand what tools are being triggered on their recommendation widgets directly from their website. And basically we have the XO and basically we've updated it to Google's latest Chrome framework, the manifest V3 to be technical. And yeah, this was basically a required change from Google. But we've answered it retains the same capabilities you had before and it's now more stable and future proof. So if you or your teams use the Chrome extension to debug and then you can use the Chrome extension to debug and analyze widget behavior, it's now running on the most secure, compliant and reliable foundations. The second is about the monitoring and I will talk about that more and more. So we recently released monitoring regarding the items streaming API. By the way, this is a common component shared with the author. when it comes to send items to file proper or an or XO. And this is basically a part of our monitoring and alerting suite. And here we covered the catalog integration. So with the streaming API, you can use where issues are basically a common pain point when product doesn't show up in a recommendation. And it's often hard to know why. And we've launched this new monitoring tool that gives you answers to questions like, are your items syncing correctly? Are any issues detected? Which product are affected? And how can you fix them? Basically. So this is helpful, not just during the onboarding, but also over time and helping both of our customers and partners, which is our customers, detect the same. And we're not going to use it. And we've got to know how to do it. And until now, it's, it was more like something quite painful when it comes to replicate. And very complex, and widget within the environment is not available in the console. So you can, you can use it. And other platform improvements, making some more usable and efficient. The console you have the product So you have the product. And so on the page, you have the product. And so on the page, you, we, we, we, um, sorted, uh, alphabetically the attributes for an item. So basically you have access to the catalog, the items you, you send to XO and for each item, you know, have the, um, sorted, um, um, attributes, making them easier to, to scan. Uh, also you can see the, uh, last update timestamp for each item, whether it came from a catalog change, uh, in the pipeline or an automatic enrichment, like the tags, dynamic attributes, and so on. Um, and finally, we, we, we, we've reworked, uh, two of our core APIs, which under tracking and, um, item integration by rewriting them, uh, in a rust. And if you don't know what is rust, um, uh, it's basically trusted by industry leaders like Google, um, Microsoft, Amazon, and meta because it delivers better performance with, uh, modern safety guarantees. So again, it's ideal for, um, um, building secure, scalable and future proof systems. So it's a bit technical, but behind the scenes, um, XO, uh, is becoming more reliable, efficient, and even more scalable. Okay. Now, uh, let's take a look on what's coming next on the roadmap. Um, I'm going to cover the two first column, uh, here on this session. Um, so now and next is basically what we, uh, want to work on this year. Um, um, now is mainly about ensuring stability with this, uh, monitoring and alerting, uh, suite. Um, and we also have, um, some automation to, uh, help you keep your widget, uh, configuration up to date. Uh, so again, this monitoring and alerting suite, we already, um, um, um, released the monitoring for the recommendation part. Um, recently we, we released the monitoring and alerting parts. for the tracking integration. So you are able to see and get, uh, uh, when you're tracking integration, um, is, is, uh, uh, any issues. Uh, and we just released the, um, the first part of the item integration monitoring, because there is multiple parts. There is the streaming, uh, API, as I mentioned, but we also have a batch API, uh, that allows you to send a lot of items, uh, uh, uh, in, in once. Um, and we also have other, um, APIs to handle the, the management, the activation of the catalog, et cetera. So all those, um, endpoints, you will be able to monitor them and, and get alerts, uh, regarding those. Um, so it's, it's coming soon. Um, the next part, uh, it's about smarter tools, basically, for merchandisers and global teams. Um, first, we, we are introducing a preview UI for the recommendation widgets. Um, and this is a, there is a dedicated slide, um, for it. So I'm going to skip it for, for now. Then we, we are launching the new version for our recommendation API. Uh, it will be faster, more reliable and built on a shared infrastructure with Fredoper. This common data pipeline. We're getting the tracking and the items. Um, and that means better integration and quicker response time. Um, um, we are also improving internalization. Um, if your business operates across multiple countries, uh, or teams, um, it, this will really help you. Uh, you'll be able to control who, uh, candidate recommendations per market, um, and also track, um, who made which changes. Uh, so there's no more accidental, um, overrides across teams. And finally, we, uh, uh, really exciting part, um, is, uh, the, the Gen AI usage, uh, especially regarding the data enrichment, um, as, um, as Tom said, it's basically, uh, about, um, uh, accelerating the, the taking of our, um, of the, the items to make it, to make them available, uh, in the merchandising, uh, regarding stuff like materials, the compatibility, that is really, uh, difficult to, and, and very, um, uh, waste of time to, to, uh, to do manually. And this is where the AI can, uh, really help you. Um, so as a summary, I would say, um, now is where X-Tro is evolving to be more reliable, more efficient, more intelligent with a proactive monitoring and everything suit and then some stability. Now, while upcoming innovations like, uh, the preview UI for recommendation and new recommendation, API based on, um, API based on share core, uh, foundations with Fredoper, internationalization features and AI tagging will streamline, um, workflows next. And looking ahead, AI driven merchandising, um, and also analytics and, um, improve UI with further, um, automate and optimize recommendation management, driving, uh, efficiency and better customer experiences. So if we can, um, move on now, um, we can dive a bit more on the preview UI. Yeah. Thanks. Um, so what's the idea, um, for those who don't know really the, what, what is the Nixo widget? It's basically, um, uh, a configuration that retrieves, um, something to recommend. It can be a product and banners or whatever. So items. And, um, our customers, uh, usually have different widgets depending on where they need to expose recommendation. to their shoppers can be on their website and different, uh, specific touch points on their website, homepage, uh, list page, product page, et cetera, basket. It can be on other, um, uh, uh, also, um, um, um, um, support. Um, and basically sometimes they, um, because the, the widgets brings a lot of flexibility and a lot of complexity, if you want, uh, to orchestrate a lot of specific cases, um, um, um, it can be quite hard, um, if you just have to, uh, to, to, to create a new touch point based on something already existing, it can be very, uh, frustrating to basically to, uh, to replicate all this, uh, configuration. And this is where the, um, this is where the, um, um, the, um, the preview. Sorry. I was not there. Um, this is where the, um, the preview can help basically, um, um, about the, um, uh, the, the, the, the, the, the, the way to, to, to, to pre-visualize the, how they, they can, uh, test and simulate all the different, um, rules they did, um, again, because it's complex and it's, it's difficult to, to do that before they, they go, uh, production. Um, and this is where we, um, we basically, uh, uh, uh, hand there, um, testing all the different, uh, rules, uh, the different contexts they can, they can reach, uh, the different profiles because you can have multiple activities, um, and it's, uh, because it's dynamic, it's, uh, um, often difficult to, uh, to test all the different conditions you could, uh, you could meet. So, um, this simulation, uh, is a, is a new, uh, uh, feature that will help, uh, in testing all the touch points, uh, for the different use case. And I think this is on you, uh, Yelena, uh, speaking about more AI stuff. Thank you. Thanks, Jan. So, hi, everyone. Uh, just a quick additional intro into me. Uh, I'm Yelena. I started in January replacing the previous product manager for AI. Um, so yeah. Hi, everyone. Happy to be here. Excited to bring AI, uh, improvements. So next slide, please. Yes. So I just want to recap on, uh, a product feature that I believe was introduced in the last product of these hours, which is, uh, all about AI search being upgraded with larger language models. Um, we want to leverage, uh, the LMS in order to increase the, uh, increase accuracy of our AI search, um, through understanding the nuance in the shopper input, uh, white matters, uh, because it will increase, uh, product relevance and product ranking. So next slide, please. Uh, so yes, as I mentioned, uh, this was already in progress, uh, or introduced the last product office hours. So we are pretty much there just finalizing some documentation. So I want to invite you to that if you're interested in, uh, uh, either upgrading to LLM powered AI search version or implementing AI search into your website, please get in touch with us through our CSMs, um, to get started on, uh, implementation process. The next, I just, uh, before I go to the actual spotlight of today, which is conversational search, I want to just, um, have a quick look of the AI, uh, roadmap. So this roadmap, um, is in fact, uh, applicable to both XO and Fred Hopper. Um, but in the column now, apart from LLMIS search, uh, we are looking at the Fed Hopper based, uh, improvements, uh, AI search simplified onboarding is meant to deliver. Well, I think it's self-explanatory, but, um, it's an internal tool to help us, um, deliver AI search to your, shoppers quicker, um, um, we are supporting, uh, Fred Hopper team as well in gen, leveraging gen, gen AI in, um, automatic, um, sorry, in the automatic generations of synonyms. Um, and then later we will look at, uh, further upgrading LLM AI search for other languages other than English, um, tagging, uh, as well for Fred Hopper that Jan just mentioned on the exercise. And we are also looking at solving some suggest issues, um, and upgrading it, uh, with AI, uh, solutions. So in the spotlight, I want to talk to you about conversational experience. So in this slide, thank you, Tom, uh, who is just pressing on, uh, uh, not there, the video. Um, this is also a recap. Uh, I'm sure some of not all have, uh, seen this demo before. Um, what is it? It's basically a virtual shopping assistant that, uh, unlike traditional chatbots interacts with shoppers in a natural language and, uh, in our context of Fred Hopper and merchandising, it actually leverages the merchandising strategy to, uh, deliver the relevant search results to your customers. Um, why it matters, uh, because we have found from our, uh, shoppers that we talked to that 40% of them would in fact prefer to use conversational search when they're looking for the products they want to buy. Um, and from the market analysis, some companies reported, some companies reported a potent increase from 7% in revenue. So in the next slide, we, I want to talk to you as well about, uh, potential use cases that we're looking at, uh, before I do that in this video, uh, you can see the other side of conversational experience, basically something that that would be visible to you, to your merchandisers, who would then control, um, how, um, how, let's call it a chatbot interacts with your customers, what language it's using a language it shouldn't be using. So we are looking at, uh, the use cases from the shopper perspective and also from the merchandising perspective. Um, so for the first iteration or MVP, we identified, uh, the high priority, uh, use cases should be product search, which you've seen in the demo, basically delivering the relevant user, um, product results, um, product results, and then also, uh, simply leveraging the LLM to provide some styling advice to the shoppers. In the future, following that, we would look at the image search by uploading images to deliver, uh, similar, uh, products. Um, also look into product detail and product comparison. All right. So, um, yes, I also want to let you know that we have in fact started working on the MVP and in the next slide, um, we also want to hear your thoughts. So having heard all of this, do you agree with the use cases that we identified? Um, do you think that this solution is something that would benefit your business? Um, do you have any feedback basically and want to talk to us about it? Um, and more interestingly, are you maybe interested in partnering with us in building the first iteration or the MVP? So please do talk to us and let us know if you're interested. Tell me later on, uh, share a poll where you can register your interest and we'll get in touch with you. Thank you. And that, that was all for me. Amazing. Thanks, Jelena. And over to Pavel, who heads up our cloud and AI platforms. Thank you, Zerom. Um, typically operating behind the scenes, our cloud platform team is excited to share with you. Some of our recent developments. Um, to do that, we'll take a small step back in time to peak 2024, which ran smoothly and we received positive feedback from multiple customers, which was music to our ears, especially knowing that we have achieved a couple of major milestones in infrastructure modernization ahead of that busy season. First, our cloud operations team, a very talented set of individuals. They deserve a special kudos really. They had opted to do this. We adopted this refresh platform and took care of the scaling ahead of time, as well as, um, proactively preventing, um, number of incidents by being on high alert throughout the peak, um, and utilizing the reworked internal monitoring system that we provided. So what have we done? Um, without boring you with all the small technical details, I'll summarize a couple of key points and I will shed light on one. We updated the multiple, uh, component elements to the latest stable versions. We increased use of the cloud native features, providing things like auto healing and auto scaling, for example. Um, all very much targeted to provide stability and ensure performance of our solution. One of the highlights I particularly enjoyed was the evolution of our auto scaling. We had the ability to adjust capacity in response to the change in demand for a while. This works well for gradually changes, uh, changing traffic. However, however, we noticed that some customers have regularly occurring spikes in request rates. Our answer to that was to apply predictive auto scaling in conjunction with the load based one. As a result, we have stable performance throughout the day without any delay from a human intervention. Note that for special occasions like peak, for example, or big product launches and TV ads, it would still be a really necessary to collaborate with you. Um, so that we, um, can prepare everything for much less predictable, uh, promotional activity. The final thing to mention is the customer specific endpoints that we are releasing. Um, mostly interesting for customers who work with multiple threat operator environments at the time. Customer specific endpoints also provide better security, uh, through isolation. Um, uh, through isolation. It gives performance benefits due to reduced complexity under the hood. Um, and also opens the doors to enabling experimental or case specific features like single sign on keep alive connections and more. Please contact your customer success manager to learn more. Okay. What's next? Um, we're in motion to apply further performance improvements across the board. By modernization, the application stack on the application stack on the one hand and utilizing more performance resources on the other. We are hoping to provide even better experience overall. More to come in the next couple of months. There are some recurring topics, um, such as security of our platform, which we will relentlessly continue improving. Everything to make our platform smarter, stronger, and faster for our customers. Thank you. Uh, have had the pleasure to work with a few of our amazing customers on either federal or Excel, uh, for, um, already the past eight years. Uh, for this month's, uh, product office hours edition, uh, was invited to share a client use case with you. And I'm happy to be here after thinking about which use case to share. Uh, I take one that I don't come across often and therefore thought it may inspire you to, you know, also give more thought to, uh, to this specific, uh, topic, uh, sorry, next please. Uh, yes. Uh, did I pull up the wrong presentation here? Uh, not at all. Uh, I briefly like to prepare, uh, your mindset, uh, for this use case with a quick intro using the Netflix, uh, example. Um, because I'm guessing that many of you have a Netflix account and are familiar with the changing title images, which, which makes, uh, this a good example as a way of, uh, introduction of the, of the use case, uh, for many years, uh, the main goal of a Netflix, um, personalized system, uh, was to get the right titles in front of each member, uh, at the right time. So with a catalog of thousands of titles and diverse, uh, member base of a hundred million accounts, I'm just, you know, uh, uh, guessing, uh, recommending the titles that are just right for each member is, is crucial to, uh, to Netflix success. Uh, so in that perspective, this, this is very similar to, I'd say there is a comparison with the challenges that you, uh, you face as, uh, as merchandisers. Uh, next please. Um, but Netflix didn't stop just there with, uh, with trying to select the best shows for you. Uh, they also thought about, you know, why should the viewer, uh, care about any particular title they recommend, uh, what could they possibly show about a new and unfamiliar, unfamiliar title, uh, that will pick your, uh, interest. So how essentially, how could they convince you that a title is worth watching? And one method to get your interest and address this challenge is to consider the, uh, imagery used to portray the title. So if the image representing a title, uh, so if the image representing a title captures something compelling, then it acts as a gateway into the title and gives the user, uh, or user some like visual, like evidence, uh, for why that title might be, uh, might be good, uh, good for them. Uh, next please. So, so far the intro, um, so what are we looking at here? So if you look at setup one, it's working, it's selling, it's a successfully, a merchandise presentation of products in a physical store. Um, if so, if that works, then why at all change it to setup two for a seal event, which then also works and is, is, is, is selling and performing well. So these are two totally different ways of presenting the products, both work within the shoppers context. At first, this may seem like a trivial case, but is it really that trivial? So let's give it a bit more thought and think about how how this would be applicable in, in, in, in the online shopping experiences. Uh, the answer why they both work is because each presentation aligns with the customer's mindset, uh, the context of the shopping session, the premium experience for the new collection and a slightly more like, let's say, grab and go set up for the sale items. Both are completely different setups, but make complete sense in different situations. Uh, now with Threadhopper, it is possible to use context. Uh, to push the best product image, which is basically, um, a, um, uh, um, uh, the, this example is basically the analogy for, uh, for, uh, for, uh, for that, uh, within Threadhopper is the product image. So within the spectrum of cool automated or completely manual, uh, both are possible in, in Threadhopper. Um, this, uh, this, uh, use case is, uh, made possible, uh, with the support of our cat. So our cat is one of the H and M portfolio brands, uh, and all the three brands are cat other stories and cost make extensive use of this Threadhopper capability, uh, which is the topic of today's use case with the support again from, uh, from our cat. So, um, presenting our kids use case. One of the three, um, brands as mentioned, um, they are optimizing the shopping experience and driving conversion, uh, by managing product image presentation. Um, so yeah. So again, uh, a quick, uh, thank you. And now let's, uh, look at the use case itself and the examples. Yes. Next slide, please. Cool. Um, so you may have other systems to manage product images, uh, to a certain extent, but pushing images tailored to, and based on user context, I don't come across often. Um, and that's why I'd like to zoom in on, uh, on this Threadhopper capability that allows stores, uh, like the H and M portfolio brands to configure which image type, uh, to show on product listing pages, uh, under, you know, uh, which circumstances. And, uh, the impact could even be bigger than in physical stores because online, uh, especially on product, uh, listing pages, you typically only present one image, uh, for each product. Whereas in a physical store, you can have multiple, uh, multiple ways of displaying the same, uh, item within the view of the shop. So in a way there are strong, uh, similar similarities between the, uh, the challenges that Netflix, uh, faces and the one you face on a product listing page. Uh, not just are there many products to choose from and your merchandising decide to present the most relevant products to your shoppers, but also relevant could be with what image you present the products based on the context. So in this example, um, in this image, um, my hypothesis is that because the hat worn by the model makes the hat shape stand out more, uh, the model image is more likely to convert to a PDP visit. Uh, the challenge in this use case, uh, when do we show what type of image to present a product? Uh, next slide please. Um, another part of context could be, uh, or an element of context. It's the, it's the, it's the POP they are viewing. Um, and here is, I think a good example where the model images are preferred, uh, over product images to give the shopper the best experience. Uh, why that is, is because I think of, for three reasons here, a relatively low number of items, um, 39 in this case, new collection and full price. So basically offering a premium experience and skirts are just, you know, suitable products to use a model image for. Um, there are also types of products where you could argue with a model photo adds any value, uh, or even damage the shopping experience, like, um, uh, a model wearing socks or, uh, wearing a belt, for example. Next please. Uh, another example, in other cases, the use of a model image could be very functional, uh, like with this example, where it's hard to tell how this bag will actually look on you without any reference to the sides. So if you look at this image, you could think that it is a, a hand, uh, handbag, uh, next, uh, slide please. And using a model shot instantly shows the size in relation to, uh, to, to a person. Now, and of course the, the, the PDP itself. So after, you know, uh, shopper would click from the list, uh, listing page to the PDP, the PDP will offer all these different images, but to prevent the shopper from ignoring the product on the POP or getting an incorrect impression of the size of an item, which then leads to a brief moment of disappointment as soon as they visit the PDP, uh, changing the image on the POP, uh, for this item would, um, improve the shopper's, uh, experience is the hypothesis. Yeah. Next please. Uh, one more example in this example, archive chooses to show product images, um, in an as functional way as possible because scrolling 200 items is, um, is more user-friendly I'd say using product images, uh, also the visitors of this category aren't, you know, babies necessarily like the parents and they're shopping for baby outfits, not only looking at the design, but, uh, certainly also at the, uh, the, uh, the function, the functionality of the, of the items. And also maybe because they're just, uh, not enough, uh, uh, uh, baby models. So maybe the, you know, the, the model, uh, the model, uh, images wouldn't be available, uh, either here, but this, I, I, I think you can agree, uh, uh, it makes, it makes, it makes a lot of sense where the models, model pictures make, uh, images make a lot of sense on the, on the, uh, the skirts category. Uh, next please. Um, and yeah, if the item contains multiple products, this would be mentioned in the product name or description, of course, but, um, but using a, uh, an image, which also kind of like, uh, confirms, uh, uh, that the, uh, the product contains multiple items, uh, can also help with boosting conversion for that, uh, for that product. Uh, next please. And if your goal is to create your piece presenting a diverse range of products, uh, let's say if, uh, you know, if it's a high level category page, like, you know, men's men's new products, for example, uh, that you can have many different products and to help, uh, you know, to make sure that the page doesn't look too messy or, uh, um, make it look more vivid, uh, you can, you know, use different image types to kind of, uh, design the page a little bit nicer than it would look with only, uh, product flat images or only model images. Uh, next please. Um, so here on the first product row in the new arrives category, we have different product types and it makes very much sense to use different images for the shoes compared to the, to the other, uh, three items. Um, so I'm, I'm now coming to, to the end of the, the presentation, uh, finally with, with all the developments in, in AI and the investments we're making in that area, I wouldn't be surprised in the few, if the future of personalized PLP experiences also, uh, it will surround product, not only around product relevancy, but also product presentation, um, powered with, uh, AI driving the, the images. Um, Um, currently already possible in Fred hopper is to set most preferred image and least preferred image in an automated manner or manually select, uh, the hero image of any product on any user location. And, uh, with the use of all the available trigger conditions, addressing different user context, it becomes, you know, quite a powerful, uh, feature. Uh, there is one caveat for, you know, all the, uh, the book people here in this call who, um, who, uh, I got, uh, enthusiastic about this. Um, there is a, there is a dependency on, uh, on the data model, uh, specifically around how your product, the varied attributes are structured. So, uh, if, if this is something that you would like to, uh, try, please, you know, when you, when you contact your TC or CSM, um, yeah, please make sure that, um, yeah, that you do like a technical assessment. Uh, and then, and there may be some development, uh, needed, uh, on your data model to, uh, to make a better fit for this, uh, for this future. Now, our cat is currently running, uh, various tests to measure the, the impact of, um, of image switching. We can't share a concrete results just yet, but the expectation is that optimization of the product presentation, uh, which is also still, uh, currently also still work in progress, uh, will have a positive impact on click through, right. From, uh, product listing pages to the PDP, uh, which will then eventually lead to a positive impact on, add to card and, and purchases, et cetera, et cetera. Um, and because image, the image management, uh, part is powered by the result modifications in FedOpper, um, because of that, it's also possible to AB test, uh, different images, uh, with the, um, uh, the native, uh, FedOpper AB testing capabilities. Um, then I'd like to end with one final, uh, comment that, uh, I understand our cat is of course a fashion retailer. And, and I appreciate that not all people in this school are working in fashion companies and hopefully, um, you know, you're able to translate this use case example to your own and apply to your own industry and your own, uh, catalog. Um, but, um, generally speaking, there are so many different type of images, uh, image presentations that you can think of. So within fashion, it's, it's typically, uh, around, you know, do I use a model shot? Do I use a product shot or maybe a detail shot? Um, but you know, uh, in any industry you can think of studio shots, detail shots, uh, usage, uh, shots, imagine a person jogging to promote the fitness tracker, uh, size reverend shots, bundle shots, uh, products with packaging shots, animated images, infographics, uh, product with accessories, you know, different color variants, et cetera, et cetera. So, um, so outside of this example, uh, specific to our cat, I think if you give it thought, I think you, you can think of, uh, uh, yeah, of, uh, uh, use cases for yourself where, you know, where, where this would, uh, could apply, uh, and you could improve the shopping experience, uh, in this manner. Um, yeah, that was it. Thank you so much for listening. Thank you, Johan. That was perfect. Uh, really interesting. I hope that everyone found it as interesting as I did. There's lots of speakers today, so thank you to all of them. One last thing before you guys go. I'm just about to put a poll live, um, where if you are interested to get involved in any of these things, then please do. Also, there's a Q&A box. So if there's any feedback about things you have seen, haven't seen, wanting to improve for next time, please let us know, and then we will take that on board and we'll respond to the questions that are in there in due course. Um, all right. I will push it live now, but it's basically, if you want to get involved in the insights, internationalization, insights, actionability, stuff to get more out of the insights tool, conversational experience or the conversational search piece, uh, personalization and AI and Fred Hopper, um, and the preview UI for XO that, that Yann was talking about. So you should see that popping up right about now. Um, additionally, as I say, there is a Q&A box. So post in there if, if you would like. I'll just give you guys a few minutes to fill that out before we close things off. Thank you very much for your time.