Well, hello and welcome back to the studio. It's great to be with you today and thank you for taking your time out. So we are having a mind stretching session where we're going to think globally as behoves all strategic leaders, but also act locally. And so a topic we've covered many times before, uh, where we look at cross border international, but remembering that our customers tend to live in one place, do one thing at a time. And obviously the thing is hopefully, uh, buying from us now in order to cover the topic, it's great to have with you in the studio today, uh, Imran from Crownpeak. And that's obviously me, but there's no substitute for seeing the actual real person. So, uh, Imran, looking remarkably similar to your LinkedIn photo. Welcome to the studio. Um, so tell everybody, uh, who you are, uh, and more importantly, a bit about Crownpeak just to get our bearings before we dive in. Yeah, sure. Ian. So thanks for having me. And thank you for the compliment. Uh, so Imran Choudhary so I'm VP of Product Marketing here at Crownpeak so very quickly because I won't go War and Peace Crown peak digital experience platform. So what does that mean. There's three pillars. So the first one is content management. How do we ensure the right content to the right people? Very much a topic for globalization and thinking local as you say. Uh, we need to personalize the content and how we engage with those people in their region. The second one is accessibility. So all digital content should be accessible to everybody. And as you are aware, um, in Europe there are various regulations to make sure that that becomes the case. So that's the second pillar. And the third pillar is product discovery. So many folks who will have join in this webinar today will know a Fred Hopper, our core search and merchandizing platform. And that's primarily the trading tools that we are using for e-commerce sites, brands so that they can trade successfully online. And the focus for today. Wonderful. Well, you did mention Fred Hopper and, uh, obviously a company that, uh, I don't know enough for 20 years. So it's great to see that it's, uh, continuing to, um, develop and involve within the Crown Peak family. But, um, we are, uh, going to bring some facts and research to bear here. So first off, uh, we are tabling. It's like a card game. Uh, I'll see you our top 1000, uh, Europe report, where, um, you tell me the other day that out of our top 50 retailers, um, 22% of those already your clients and a third of the top 25. So I'll let everyone else do the mental maths on that. But you've also brought some research to the table. -Um. What's yours? -Yeah. So, um, our research was commissioned last year. So obviously with the economic times we were going through, it was a case of trying to understand from the market, you know, where were e-commerce teams Merchandizers really focusing their time and effort. So what we saw very much so the global expansion was a big driver. But also we wanted to understand, you know, what were they looking to change. So we've gone through that trend of everyone going kind of headless, maybe Replatforming you know what was next? We wanted to understand what was that kind of next wave. And that's where we focused in on product discovery. So this is where we've done some research with regards to 300 odd merchants globally to really understand what they were focusing on, what were their challenges and what -they saw with their opportunities. -Interesting. So, um, today we're going to, uh, dig into just a couple of those challenges because, uh, uh, time is is limited somewhat. Um, but, uh, how should we approach this then? Because there are a couple that that you pulled out. I mean, we normally looking at, uh, the sort of global activity you have, the first challenge is to be seen by the customer. So make your products at least visible in the country. The next thing is to make them available in the country. So you have all these shoes of where's the stock, um, transport, uh, shipping, getting to and from with Brexit and other uh inhibitors to trade. So, um, physically having the product there are all of, if you like, the behind the scenes activities, um, that just get you to the point of being able to sell something. Um, but maybe let's pick up the story there and look at the, you know, what is what is the question we're trying to answer, -um, today. -So you're right, in a lot of these things that you mentioned, though, are kind of hygiene factors there just by in part, you have to do those things to be able to trade into those regions. So there's the next kind of evolution of that. So you're you're in that region, you're trading great. But how do you optimize? How do you really excel and drive that success in that region. So going. From having something to sellable to actually doing the selling to actually. -Set the. -Yeah, good. Distinction. And that was the thing that came out of this study. So 87% of those that we interviewed did highlight that product discovery was kind of their next key focus. It was their opportunity but also their challenge for that international success. -For sure. For sure. -I think, um, you know, before we dive in, the broader context here is, of course, that, uh, there's an arms race at the moment to have more and more things on your site so you not only have your own products that you bought, consignment, dropship, third party, you then probably opened a marketplace. You have the feeds coming in. So the whole world seems to be stocking 200% of the available product at any one time. So within this, given that the screen sizes and attention spans haven't moved, it's all about getting the best products to make the customer happy. Convert optimize margin within that attention frame rather than just say, well, you know, here's a pitchfork, go and dig through it yourself until you give up the -will to live. -Yeah, if you remember, because it was at the event, I think about a year or two ago where I spoke, you remember the paradox of choice, I think was my topic for for the presentation. Don't get me wrong, I get dropship, I get it, I get the idea of having a big catalog. But you're quite right. We have got this situation just from a product catalog perspective, where we arguably have too much product and we're not sorting it based on someone's intent or someone's preferences or what's happening in that region. So if you remember and unfortunately, it's still kind of the case, you know, my example at EIR was let's go type in men's white t shirt. How many did I get back. You know, people were kind of shouting out 50, 100, if you remember, and it was like $2,000. And for me, it just blocks your brain. There's too many products for me to then go through and make a choice. It overwhelmed my brain and therefore I don't make any choice. So. So we do have a challenge. You just pick up on that first thing, because the first, uh, topic we have, uh, obviously is search. So why don't you talk us through, um, this part first. This is where the customer is politely telling you I'm interested in, um, thing X or thing Y. Talk us through, um, the findings here. Yeah. So these are some additional findings as you see. So what was interesting is that the consumer does want to find their product fast. Okay. So there is a high importance in terms of getting them to the right product as fast as possible. There are consumers that for sure, um, would prefer good search and good search results over and above of a retailer. So it's highly important to them, especially when they know exactly what they want. PVH So Tommy Hilfiger, Calvin Klein, you know they've got that quite they've got that right, especially across all the different regions. And they've already seen that those that are using search of five times higher to convert. So there is a huge importance on search. And as we've said before, everyone kind of gets all of the the foundations in place to trade in all these different regions. But then it's that next iteration of optimizing for that region, be it based on, let's say, beyond just language, but the various things in terms of cultural, cultural context, synonyms, variations, all of those things you need to start optimizing. And it becomes quite hard because a lot of that requires a lot of manual effort. Traditionally, yes. But I think interestingly as well, though, I see an analog here to staff training. So if in your store you have staff who can understand the nuance and detail of a question and know the product really well and know the use cases so that they can reply to you and say, I'm. I think the thing you should look at is X or Y. It really is bringing that level of, you know, you call it elevated search. I really like that uh, uh, that phrase, which unfortunately I can't steal now because, uh, we have it, uh, that is yours here. But, um, I like that idea that in a way, you're elevating that to. I don't say concierge, but definitely a more thoughtful level. Um, between the customer and the product and their intention. Um, and you had an example here, which, um, I had to read about three times before I fully understood it. So, um, -what's going on here? -Yeah. So it's interesting, as you say, the seller in that store in that region knows their market, knows their customer knows their product. And as you quite rightly saying, we're trying to bring that level of intimacy and concierge like that. Thank you. And I might steal that um, on online, you know, and it's quite hard if you've got a centralized merchandizing team that isn't in region. So this is talking about complexity and elevated search. And so if you think about Google, Google is really set the benchmark in terms of everyone uses Google. And we're starting to write natural language sentences to find what we're looking for. And consumers now expect that in their online shopping. So we are seeing this trend of 4 to 5 keywords being used in a search query. This one's quite an interesting one. So this is mango, uh, who's doing this extremely well. Um, not only just in terms of the complexity of the search term, but we're also seeing different languages being used within the same search query. So, you know, you can read it here. I'm going to test you. Vestido do you know what Vestido is in Spanish? You're a very cultural person. Surely you should know that one. I'm just going to say if it didn't have Spanish written underneath it, uh, I would have struggled a bit of that I might have thought is a new style of dress. But, uh, tell me what? What does it mean? Yeah. So it's dress well with floral pattern. So not only have you got the complexity of understanding floral and pattern, but then you've got dress in a very different language. This is just really difficult for traditional e-commerce or traditional search to be able to understand, because you're picking up multiple keywords there, but also you've got that language piece there, which is also quite complex to understand unless you've got the right technology and if you're doing it traditionally, it's going to be hard. It's going to be very hard. But kindly, uh, you've broken down, uh, the sort of anatomy, uh, of searches, although this was, um, interesting as well. Um, and I let people, uh, squint and read this, uh, which is actually a good time, uh, to remind our long suffering, uh, uh, people in the studio. Um, the deck will be available to you later. So if you are, uh, squinting or thinking, oh, I need to copy that, uh. Feel free. Uh, we will send it to you, um, once the transmission is concluded. Um, but I was interested, um, at the bottom, uh, in where you have this, where merchants spend their time, but then, uh, talking about the long tail. And I think what's interesting is we're all familiar with long tails gone forever. In this case, it's quite interesting to see how. How. Close to the short tail. It is so in reality, it's as if most of the effort is on the top sellers. But then, you know, 80% of your products are not getting the love they need despite you having bought them, photographed them, described them, and attributed them so. So tell me about, you know, how we work this long tail, um, in an effective way, especially because the way the previous, um, search was structured is sounding to me a lot more conversational, a lot more like chatting to an AI assistant, chatting to your friends so I can see the search is changing. What's the actual doing behind the scenes to make the most of -that? -Yeah. So just very quickly we spend time on the one word and the two word kind of phrases, because that's where the highest volume is of our search queries. So it makes complete sense to spend the time there. Okay. You what where we're finding the challenge is this, this, this, this variance of the unique terms. So these really descriptive terms and that could be the sentence before it could be kind of as you say, natural language. Or it could be, you know, just very, very product descriptive. So red Nick men's running shoes I know exactly what I want and I'm ready to purchase. Yeah. -So you just need to. -Add your shoe sizing and then you're um. Oh plus cheap plus in stock plus next day delivery. Perfect query, -full intention there. -But you've got huge high intent there. But because the search query is so unique and so descriptive, traditional search just can't pick up on it or gives irrelevant results or even zero search results. But the volume of those very unique queries is so low that either it's not picked up on, or the time and effort to optimize the results accordingly. It just doesn't pay off because of the manual effort it takes. So this is where I really dips -in to try and take the bird. -Talk me through that and how it works from the merchandizers perspective. Because, um, I'm old enough to remember in the noughties creating spreadsheets where I had, uh, input lists of every brand input list of every category, every color from our color palettes. And then we did this big combinatorial production thing that said, for every brand name plus color plus product, plus keyword plus cheap plus whatever, create a basic answer to that search string. And we created millions of these. We're very pleased with ourselves. Massive redundancy. You know, 95% were never used, but they were there waiting just in case. So that was the brute force old man with spreadsheet approach. Um, are you going to tell me now that I have to do nothing other than look handsome, and that I is going to do all of this for me? What's the actual what's the lived experience of a modern merchandizer trying to make the most of this long tail and AI capability? Yeah. So there's no, um, let's say stopping of that. Some of that is still very useful. And obviously product attribution is hugely important. Okay. But what we're looking is to try and fill that gap, as it were, that long tail, which is about 20% of all search queries. So we're looking at user activity against products, against those product attributes, against those product imagery and against associated products against it too. Geez, that was a bit long, but we're looking at all of that activity. And because we're using AI, we can kind of mine that very quickly and therefore automate that optimization. So, you know, we've got a situation where we, you know, give you an example actually. So let's say I'm on a grocery store and I'm typing in, say, oyster Bay wine. Okay. Now that grocer might not actually sell oyster Bay wine. Um, but actually you don't want to be serving up results which are, say, um, oyster sauce because you don't want to be drinking that in copious amounts. But what we can see from there. So what we can see though, is people that have typed in such unique terms before. Yes, maybe in a very short period of time they got zero search results. Okay. But then we start to see what they go on and click after that point. So it might be other wines, but they would associate as very similar to the oyster Bay wine. Yeah. Or the brand oyster Bay, making those associations of that activity to the products that they clicked on and then actually purchased, we can start to make some justification of, okay, you typed in oyster Bay wine, um, we don't sell it, but here are some suggestions. But those suggestions have data behind it to justify that they are relevant and they make sense to be promoting to that customer. And if I am hearing this correctly, um, that also helps the zero um, session history question. So let's say you come to my wine shop and I've never seen you before. You just land. No history, no Brasi. You type in oyster Bay, I don't sell it. I can then pull in from the collective AI brain that. Oh, you don't have it. Have a punt at this. -Correct. -Custom. Correct. And we've got kind of accelerators, but we've built so for key retail verticals, we've already built a lot of those accelerators so that you don't have to start from scratch and kind of grow your own brain, as it were. Um, the accelerators never understand those products, those queries in those key verticals. So you've got kind of that fast start. Um, fascinating. Look, before we leave, um, search, uh, little company that we, uh, we know. Well, um, pretty little thing, um, they had some results with you. Do you want to just tell us quickly what, uh, what the -learnings were there? -Yeah. So we all know pretty little thing for sure. Um, to your point of, you know, do I stop what I did before? Um, no. There are still some key tactics that you take as a merchandizer that you want to be able to do. So, you know, simple things around drag and drop, for example, you kind of still want to retain control of that. That's not going away. But what we're looking to do is kind of automate a lot of the the manual tasks that are a big burden or a missed opportunities, like we said, with the zero search results. So what's interesting with Pretty Little Thing is if you used terms such as, you know, smart jacket, many will know pretty little thing. Um, the idea of selling smart jacket is probably not one that immediately comes to mind, but actually, um, and it's generally not in the product attributes either. But what we actually started to see was people, that is one example, started to searching for those types of things and the products that were coming back with traditional kind of jackets. They wouldn't be in the context of smart. And then very quickly, what we started to see was how the product assortment or the search results started to change to products which were actually smart in nature. So it's a good example and visual, um, available on our site. Um, -as a case reference download. -Lovely. Well, look, I got a little bit excited there. And click the slide in advance because, uh, we have, uh, drifted now from the moment of me asking a question. Uh, the, um, the site is, if you like, responding to my request. So we're coming in now to the the answer, if you like, to the question which is the merchandizing recommendations. So tell us what your research found, uh, about the, uh, -the focus areas here. -Yeah, sure. So out of those 300 merchants, we kind of wanted to understand where their focus areas and what we started to draw out was control was a big, big, big ask, okay. Everyone wants to see that AI automation and improvement and efficiencies all very good things. But we all like to talk about in here. But they still wanted to retain control. But they also wanted to add more sophistication, which I know we're going to talk to the other area. The other area was visual merchandizing. So, um, how do we create online stores to be more engaging and have a better experience? Just like to your analogy of in-store, we spend so much time and effort visually merchandise in the store. How can we do that online without having to, let's say, manually drag and drop and curate every category and every kind of product line on our site. How can we do that in a more efficient way and 1 to 1 personalization? It's a little bit like the old age of mobile. You know, it's the it's the big topic which keeps we still seem to, you know, be challenged with trying to get that really right. But it is still a core area of how do we drive relevant 1 to 1 personalization. And it becomes even more difficult, um, when you're going into national. I mean, what amused me about this is, uh, you know, this is a real retailer survey, because when you offer retailers a list of ten things, they must have 90 plus percent of the award, all ten things at the same time, whether or not there's a there's no like I'll put them in orders. I want absolutely everything. So that's had the ring of truth and the balance of um, you know, control and capability. So. Absolutely right. And you then talking a bit about the, um, trading KPIs just introduces it's too much to digest in one go. And then we move on to example. But I'll just tell us that the shape of this, uh, of this insight. Yeah. So those leaders, um, in kind of retail, as per your er, rankings, um, what we saw from that is those, those folks are really leading the marches in terms of trading and success, are using sophisticated merchandizing strategies. So they've gone beyond the traditional let's promote the bestseller to the top without thinking about have I got inventory or size spread there. Um, they've kind of started to move into okay, well, let's look at a row by row strategy where we're pulling in data, uh, different data sources, and we're waiting them. We call it ranking cocktails. So building these sophisticated strategies. So here is an example Nguyen. So I want to promote my newest in uh, if it's getting the highest views, the highest purchasers. But I want to make sure that if I'm giving it that prime spot at the top, um, but I actually have got stock availability and a good size spread. I love that cocktail though, and my favorite phrases. So let's just bring that to life then. So you've got, um, a mix of products. Uh, here, um, just talk us through what I'm looking at, uh, is a little bit like, um, the Racing Post's tips for a horse race, but, um, to just bring that to life for us. So, um. Yeah. So if you think about your product listing page just as a, as an example, you can have, but with, with ourselves, what we've started to build is row by row merchandizing. So as per the last slide, it's a case of build your merchandizing strategy per row if you wish. Or for the whole listing page. This example, you've got three merchandizing strategies split across three rows. Fundamentally what you're seeing those the strategies remain the same. But because we're pulling in data from each region, we can dynamically change what's actually being promoted based on that regional preferences. So it's a it's it's a very simple example in but Australia it's warmer right now versus let's say uh the UK a month or two ago. So the strategies remain the same. But what we're pulling is the different data from those regions and therefore changing the product assortment to be more in tune with, in this case, seasonality across those two different regions. And so, um, I'm assuming that, uh, for example, if there's um, a t shirt that, uh, is worn by a highly influencing, uh, social media star, the views will go up and that'll trigger the algorithm to weight the viewed products more highly. Uh, and so the cocktail, if you like, changes based on the data. Um, correct. But how real time is this? Is this sort of, um, 50 times per millisecond, once a day? What's the sort of responsiveness of of -the system? -Well, you. Need a decent amount of data to feel justified with what you're going to be changing. So, you know, we wouldn't say in real time personally, because you're not got the data to really justify, but kind of a 24 hour kind of window. You've got enough data there to really be able to have some, some evidence that you can make those changes because you want to make those changes automatically. You don't want all these pings to merchandise is asking them to change this, this and this. You want to do it automatically, but you want some decent -data behind it. -So you -answer. -Yeah. Anyone else have got oh it's it's it's like every millisecond it changes. We say -that I don't have some data. -Um, no. It's like you say, you know, we've been doing this for a while. Um, we we do our best to have justified statements and not just hype everything up. So I think, you know, being realistic, you want to make sure that the data is there to back up the changes that you're making. What's interesting is, um, Puma. So Puma are running. Are you you've moved to Adidas. That's controversial. -Okay. -Just apologize to all sneakerheads out there. Um, talk about Puma over Adidas. -Let's see whether the screen melts. -I don't think I can do that. I don't think I could do that. Let me change it. Then let me go to edit -because. -I've removed it, I removed it, -I could, I could, I. -Couldn't do that. But you know, Puma um, spoke recently about how they're changing their merchandise and their assortment changes based on local influencers. So Puma have some really big brand ambassadors in different regions, and you can already see the assortment changes in one region based on that influence or that brand ambassador at different periods of the year and trading times versus other regions. But you wouldn't want that change by that one brand ambassador in that region. That makes sense to then change the rest of all of your -Europeans. -That's a localized. Understanding. What the bounds are of, um, customization is important. Um, if you need to wash your mouth out or, uh, have a break, a deep breath, uh, let's go back to, uh, just to prove that other, uh, sneaker companies are available. Um, yeah. Even though they're no longer trendy, given a certain Prime minister's been wearing them. But, uh, for those who are still, uh, fans of Adidas, I can see here that, uh, we have a UK site and a German site, and, uh, this looks like an IQ test, but I can see that the top left trader has moved to bottom right trader. -Uh, what's happening here? -Yeah, it's kind of spot the difference, -isn't it? -It is. Mark, you need your little mark opinion, but no, you're quite right. That's the immediate one. But you know that top left to top right. But as you said, as we saw earlier, we've got merchandizing strategies across these different regions. And what Adidas you can see is it's taken that local data that preferential data. And you can see how the assortment changes and you know, you've picked on the one which is obvious with the pink. But you can also see various nuances there with other different products in different color ranges and different price points. Okay. Um, so that is something that we also see, uh, prices a sensitive area, which is a common, um, weighting in our ranking because let's say Switzerland and Switzerland, they're less price sensitive versus let's say the UK. So again, understanding that shopping behavior, um, is really important as to how you sort, but then also get that click through rate in some of those conversions we all -want. -So let's then jump back into we've seen what the consumer sees. Thank you very much. But if I jump back with my merchandizers hat on, um, I do my work. Uh, I head home for the day, have a nice sleep, come back in the morning, and I see. Ah, no, my favorite trainer is now bottom right. Do I get any intelligence that says, um, you know, because that move could be according to your metrics. It could be simply because we're running out of size three or, um, you know, we're no longer price competitive compared to somebody who's got a special on. So how do I, as a merchandizer, understand the reason for shifts and then get told, mate, go and get another 10,000, go and do this, go into that, because there's this amount of revenue that you could have been taking. But I've had to throttle back because how communicative is it again for the person who's in control? I trust you, I know why you're doing it. How do you help me act better to sell more? -Yeah. -So as you would expect, there's insights, there's dashboards, there's overlays so that merchandise. And when you come back in the morning, as you say, um, or if you take your laptop home, maybe in the evening if you're behind targets and your KPIs, you will see these overlays, these insights, these analytics kind of showing you why that has been changed. So try and make it as you say, we are guided by our customers. So we try and make our solutions. So kind of work as to how the merchandizers want to work so that data is there, that insight is there. But you pick up on one piece there. We have a we do have a cultural trust thing going on. We as merchandizers want to fully control everything and do it all manually. But you raise a good point. You cannot move, especially when you're international at the rate in which things are trading, and be able to keep check on all of the data to then make all these changes manually. That's why you need to trust the data, the controls as per the merchandizing strategies. That's your your choice, your guiding. Um, they're the guardrails to make sure that whatever the AI or the automation is doing keeps within the guardrails of what you are trying to achieve as a merchandizer. But trust in the data to make sure that the right assortment is is kind of promoted for sure. So come on. We leave. Um, I answered the first one with my, um, algorithmic optimization hat on with this has to, uh, if I moved into as if I'm in a different store. So it's not like it's the same stuff jumbled up, there's sort of slice and dice. This looks like for the same question, you said, oh, here's a different sensibility. So tell us what's happening here and why these two sites are both so different but also internally coherent themselves. -What's going on here? -Yeah. So Hermes is a pretty cool example, and it's a visual one, which I know we might touch on today, but, you know, there are very different tastes. So in France it has a slightly more chic, um, couture kind of taste versus let's say, Netherlands, which is more casual. So this is where it's a more dramatic visual kind of effect as to how you can see different regions have in different taste, different preferences. And being able to, to, to, to meet those needs to meet those preferences in that region for sure. Lovely. Well, look, we've looked at the product listing pages, if you like the first polite answer, uh, to our question, um, but now if we go to the next level, which is when we're looking at individual products, um, we have the recommendation side. So these have always been tricky because in one way we're saying to someone, hey, you asked this, this is what I recommend. But also I'd like to confuse you with 600 other different options. And so I'm thrown once more into the ball pit of commerce floundering around. So just talk us through, um, a sort of a better approach. To recommendations on page and how these can be improved. So these are first of all. Yeah. So these are two examples of what we typically see um when customers come to us. So on the left hand side you've kind of got this visually similar, um, use case that you should be trying to promote. So I'm looking at this men's camp shirt. It's actually saying customers also liked. And as you can see it's a very different gender. And not even in the shirt category. You've also got you may also like which is again a blend of different genders and categories. So here is a is a classic one really. You should be trying to show something that is visually similar. Um, to help with that discovery. If I'm on this product page and it's just not right, maybe the price point, maybe the fit, maybe the color, but there's just something that's not getting me to add it to the basket. So you should be trying to automate visually similar there on the left. On the right. Um, this is a classic. I'm looking at the the long johns. You know, it makes sense, but you should be trying to sell me also the thermal top. But instead what you're actually showing me is that I think it's the Parker jacket. You know, this is this second use case we often see, which is shop the look. Um, both of these, -this is. -Also a classic. Case to you. And I don't know who the retailers are. And, um, you know, it's invidious to even name people, but this, to me, is a case of fire and forget. And then it's like a trebuchet of product, just catapulting things, hoping something sticks. So again, from a consumer experience, um, this totally lacks relevance and cohesion. Um, so how can this be -improved? -Yeah. -So you're. -Right. And it's what's also given a bad name to recommendations, but typically it's because we're relying on traditional e-commerce doing this or we're limited by, let's say, the strategy we can implement because of, again, technology. So what we can look to do and these are two examples very well timed. Um, so forever, forever new that, you know, what we did here is you can see visually similar there at the bottom, which we've automated. So we're looking not only at the product attributes, but we're actually looking at the image. So we're understanding the image and making those comparisons to be able to automate this visually similar kind of recommendation. Now if you try and do that manually it'll be -difficult. -Yeah. We say we're looking at this image. Um, it's not you. It's not me. Uh, is this the AI Pixies who are going through and looking at every product image, creating, um, a set of attributes, things, a floral dress on a path in a, you know, wildflower garden, etc. and then using that against all the other images, that's I assume you're -nodding. That's what's. -Happening. A combination of understanding the image and then also the product attributes, the product description and what other product information we have and making those comparisons. But the image and then also the activity that we've recorded. So you may previously have clicked on this linen cut out dress and then clicked on one of the others. So we're making a match between the image user activity and product data that we got, bringing those together with the AI Pixies, as you call it, and then putting forward these recommendations, knowing that we've got enough data to justify -that it is relevant. -And before we move on, because I think the point you made earlier about this balance between, you know, incredible intelligence and I still want to control it. Um, think about the control side. If I'm a sizable retailer, I may have bought a system that is automatically scanning all my product shots and creating, um, attributes and descriptions, either in the back end or for the website. You're looking at the pictures as well. Um, does your eye, while we're all asleep, chat to other eyes, past data back and forth. So don't worry your head. I'll do this. Or hey, thanks for doing that. Can I grab your attributes? What's the chat that goes on between eyes and my visibility of that, you know, while I'm doing other things? So this is another topic, probably for another webinar, but it's an issue that we've got in our industry a lot of, and it's something we feel quite philosophical about. A lot of the vendors in the market are all claiming that their AI is the best, and they're creating this walled garden where they're not sharing one algorithm with another. So give you an example. Um, see discount okay. See discount. Have an amazing data science team and building their. Own algorithms. Okay. And what they want to try and do is be able to share those algorithms with other vendors. Now we believe in an open ecosystem. So we have no closed walls. Yes, we build our own AI and our own library of algorithms and will continue to, but we're open to sharing those and also receiving from others. So we have a situation called algorithm orchestrator. I won't get into the detail because it's another webinar, but we are creating this area whereby. Our algorithms, third party algorithms and bring your own algorithms can all be brought into one place and managed in one place, so you can blend them to the ranking cocktail idea, blend them together, manage them a b test them because we know there's no silver bullet with one algorithm can do everything, so the way to optimize is to bring these all together. So we are philosophical about having this open ecosystem, but unfortunately it seems like the market needs to start being a bit more open versus claiming we're the -best. And you can only do it in our world. -Yeah. Well, look, I think I'm going to pick up on that suggestion of a separate webinar and a piece about that. Um, our dear listeners, if you have thoughts on this, drop me a note, Ian, at internet retailing. Net, especially if this is something that you are already doing, feel strongly about, feel strongly against, uh, let us know. And, um, Imran and I will pull something together because I think it's an interesting point, but let's not leave this topic. Uh, Imran, before looking at a case study. Um, so you, uh, you put a slide in here about Larry. So, um, just tell us what they got up to. -Yeah. So this is. -For me a great example of huge data, huge challenge and being able to automate at scale. So back to the international piece and the volume of effort that we all have and all the tasks we need to do. Yeah. So over half a million products in their catalog, um, they have over 10 million plus customers and they're operating over 20 different countries. You just cannot do 1 to 1 personalization without some form of AI. And we've had amazing success with them in driving that 1 to 1 personalization across all these countries, all these different types of customers, not only online because that's it's important, but also what's important is let's call it offline. So the marketing campaigns. So given that we are API first headless, whatever jargon we want to use, what we're ensuring is consistency of any 1 to 1 recommendation is not only online, but also in any of the follow up marketing activity. And as you can see there, the the stats that we're showing is kind of the improvement to click through rate improvement to repeat purchases. So it's a great example of managing the complexity of all this data product customers, but also the different preferences of all these different um, regions to manage daily. Yeah, great. Well, listen, um, time is, uh, not on our side, so we should maybe just get to some of the, uh, final thoughts. And we've covered a lot of these, um, at a high level, as we've journeyed from customer intent to the search results to the, uh, you know, the product, uh, the product page. Um, but I think these were, uh, some, some points that we've come out a lot, which is, you know, how you can be effective in a complex world. Yeah, and I know we're doing the follow up so folks can read them at their leisure. But fundamentally it's language is important. But languages. That's kind of hygiene. What you ideally need to try and do is find a way to dynamically change to the local preferences. That's been kind of what we've gone through today, but being able to do that at scale requires two things from our experience. One, I because we can't build lots and lots of merchandizers doing all of these manual tasks, you just can't do it. It's not possible. So you've got to leverage AI. The other element is, you know, we need to get away from, okay, we're going to do this in UK. Then we're going to do the exact same thing in France, then in DACH, then in US. Whatever. In all these different instances, there needs to be a central way to be able to do it. And that's what we do. So we build a single platform that allows Merchandizers to do this at scale, but effectively across all of these regions. But when you do that, you also need to make control. So access rights management of that merchandizing team. So that's how we all wrap it up into our platform. I think I went a bit salesy there. But you know, -I think I've been quite good today. -So I'll just say -evangelical evangelical. -You motion got to you, you got carried away. So that's a that's totally fine. Um, but in case anybody is thinking, um, I'd like to get a copy of the report, just a reminder that, uh, you will indeed get that copy. Um, but I would also like you to let us know, uh, what followup you'd like on the, uh, the AI, shared AI, collaborative AI, etc., etc.. Uh, sector AI, um, something we're definitely very interested in. So just before we, uh, head off, uh, Imran, um, anything else that we we haven't covered? -No, I meant. -That was what's next? You know, there's the follow up of the slides from today. So thanks for having us, Ian. And then also the report, which is free for folks to download and. -Hopefully be. -Insightful. So thank you again for -having us. -Absolute pleasure. Thank you for taking us through that journey. And a really nicely balanced position, particularly taking into account the, uh, sort of like that contradiction between control, uh, but also wanting scale and automation. I think that really is, uh, a topic of our day. So thank you for bringing that to life. Uh, thank you, everyone, for joining us. And, uh, give us your time today. We'll follow up with the emails, uh, straight away afterwards. Um, and we look forward to seeing you in the studio again very soon. Thanks, everyone. Thank you.