Personalizing the customer experience in e-commerce (with Camille Aâssila)

Camille, I'm really happy to have you on the podcast today. We're going to talk about quite a few dense topics, particularly around customer experience and personalization, as well as AI, which goes hand in hand with all of that. So I'm really looking forward to asking you my questions, but before we get started, if you don't mind, I'd like to introduce you. I'll give you a quick introduction and then we'll fill in the details. So, you're a Product and Customer Marketing Assistant at Nosto, a personalization platform that's well known among e-merchants, which allows them to adapt their customer journey, their product, their merchandising, and their campaigns in real time. Basically, it's the entire customer experience that is optimized in a smart and scalable way. Is that okay as an introduction? That's fine, perfect. What would you add? To introduce myself briefly, I'm Camille Aâsila, and I'm Head of Product Customer Marketing at Nosto, as you said. I've been working in e-commerce for a little over 10 years now. I started out at a web agency specializing in e-commerce and digital marketing called Axom. They really got me started, gave me a taste for this world, and above all, gave me a taste for well-crafted strategies for customer experiences. Then I continued my career at PrestaShop, an e-commerce platform that is well known in the French ecosystem, mainly in marketing roles. And after a few years, I joined Nosto. I started on the client side, where I supervised our customer success manager teams. These are the teams that support our clients in their personalization strategy. After a few years in that role, my interest in marketing eventually led me to my current role, which combines marketing, product, and customer experience. That's great and really interesting. I see that you're already very well established in the ecosystem with Axome, which we know well. And then there's PrestaShop, whose CEO Eric was on the podcast a few weeks ago. I suggest we start with a little innovation on the podcast, so it's a true or false question. So, quick question, short answer. The first one: the majority of e-commerce brands don't yet have a solid personalization strategy. I would say true, and studies show that there is a real disconnect between the perception that e-commerce brands have of the experience they provide and what consumers actually feel. There is a real gap between what brands think they are providing and what consumers are actually experiencing, so true. Interesting. You lose the ability to personalize a site without using personal data. True, absolutely. You can use what's called browsing data or behavioral data on the site, which is ultimately anonymous data. This can be the pages visited, the type of pages visited, the type of products viewed, the time spent on a particular type of page, and so on. And all of this can help us personalize the customer journey. Okay, great. You also have the user, the type of browser, things like that. Yes, absolutely, absolutely. Okay, interesting. In five years, all e-commerce journeys will be driven by AI. Well, I would say that's false, just to add a slight nuance. In five years, I believe all journeys will be powered by AI, but driven by humans. That's an important distinction. Fourth, personalizing emails is more effective than personalizing your website. False, false, because in fact, they are equally important. One cannot exist without the other. Personalization really needs to be applied at different touchpoints to be consistent and truly effective. An AI-generated product page can be better than a handwritten one. I can't answer that. You'll have to cut it out. I didn't really say anything about that, about the writing, the quality of the writing. Okay. In any case, perhaps what we can say is that it's easier for a human to be assisted by Alibaba to write and check. Do you want me to answer that anyway? Yeah, go ahead, we can test it otherwise because there are 10 of them. Yeah, oh yeah, you told me, yes, no, but go ahead, let's do it. What did you say? Repeat the question. So, an AI-generated product page can be better than a handwritten one. So true and false, I think that entirely generated by AI without human control, once again, I don't think it's, it won't be the most optimized product page there is. On the other hand, I think that AI really assists humans in this area and has value to bring to the creation of product pages, particularly in terms of productivity, but also in terms of content quality, especially SEO. The order in which products are displayed can have a direct impact on the average order value. Absolutely true. Products that are displayed at the top of a category page, for example, will tend to be clicked on more and purchased more as well. Merchandising is only used to organize categories and filters. Well, that's partly true, but merchandising is mainly what guides the user through the entire product catalog, especially since many e-merchants have very extensive catalogs. So it guides users to the products that are most relevant to them, but it also allows e-retailers to guide them to the products they want them to buy. Personalized merchandising is reserved for big brands. False. With today's technology, personalized merchandising is really accessible even to the smallest brands and teams, and that's where it will be of great interest for automating a lot of things. Personalizing search results is as important as personalizing recommendations. Absolutely. Once again, I think that personalization strategy really needs to be applied at every touchpoint, including site search and recommendation blocks. And finally, post-purchase excellence is just as important as the moment of purchase. Absolutely, the post-purchase experience will also ensure good customer lifetime value, ensure that consumers and customers return to the site, and therefore improve the brand's profitability and long-term viability, so it's very important. Great. Listen, I suggest we get to the heart of the matter. Before we get into the specifics, I'd like to take a step back and get your overall view of the market and these issues. Because it's true that there's a lot of talk about personalization and customer experience today. But my question is about the one-to-one journey. Do you think it's more of an expectation, a real consumer expectation, or is it just a marketing thing? No, no, I don't think it's just a buzzword today or an expectation of marketers. I think there's a real expectation on the consumer side, and in fact it's due to all these digital giants such as Amazon, Netflix, and Google, which have completely redefined the standards of personalization by providing experiences that are ultra-fluid, ultra-relevant, and personalized. As a result, consumers now expect the same level of excellence on the various platforms they interact with, including e-commerce sites. This means that e-commerce retailers need to address these issues in order to provide relevant experiences, but often with far fewer resources. That's where it can get complicated. I'm not the one saying this; studies show it very clearly, as I mentioned in the true or false section, including a study conducted by Nosto which shows that 99% of brands think they deliver a highly relevant experience, when in fact 69% of consumers surveyed believe that this is not the case. And so this gap between perception and actual consumer experience creates a kind of blind spot that prevents brands from identifying points of friction and therefore from offering truly personalized experiences. And what would you say are some of the main obstacles to realigning these perceptions? Well, first of all, you need to be aware of them and be able to analyze the friction points on the site so that you can then say, “OK, I need to take action. I need to provide experiences that address very specific strategic points.” And do you think, for example, that it's more of a data problem when brands analyze their customer journeys, or how is it that the person in charge of this today has a disconnect between their perception of things and what the consumer is reporting, an input or output problem? There are quite a few things that can explain why personalized strategies are difficult to implement today. First, there is data fragmentation. Often, teams will use different tools, and the data ends up siloed. This means that the tools don't communicate with each other, the data isn't unified at all, and behind the scenes, this creates results that aren't particularly consistent. It creates journeys that are as fragmented as the data itself and therefore not as optimized as they could be. Okay, interesting. So, in your opinion, what does a good personalized experience look like in an ideal world? In my opinion, a good personalized experience is one that is invisible to the user, but which still responds to the context and the consumer's intention. So the idea is to be able to understand their expectations at a given time and provide a journey with personalized elements that respond precisely to that context and that moment. So yeah, we hear a lot about personalization in all sorts of contexts today. What do you think makes a good personalized experience? Good personalization isn't just putting a first name in an email or showing similar products. Above all, it's about offering an experience that truly meets the user's expectations, especially in a specific context and in real time. In fact, it should be invisible, but it should be useful. The user shouldn't even realize that they're on a personalized journey; it should be completely transparent to them. They should feel that the site understands what they're looking for and what they need, but without actually seeing it. And above all, a good personalization strategy should really be involved throughout the entire customer journey. This applies both during the awareness phase, when the user discovers the brand, and during the consideration phase, when they are really interested in the products, comparing them and the offers, and so on, right up to the purchase and, of course, loyalty. So what's important to remember is that good personalization creates value on both sides. It creates value first and foremost for the customer, since they will have a super smooth and simple experience with the brand, and of course it creates performance for the e-commerce brand, which will see an increase in conversions, average order value, and customer lifetime value. Okay, so yeah, do you have an example of a brand that does this well? A good example would be a brand like Casio, the Japanese brand well known for its watches. The problem with Casio is that they have a lot of retailers. So it's in their best interest to provide a flawless experience on their website to ultimately promote direct sales. And like many brands, they also have a very extensive catalog with a lot of skew, a lot of new product entries, and so on. This makes it very complicated for teams to manage merchandising, for example, and show the right product to the right person at the right time, which becomes very tedious. So they have implemented a whole host of strategies, particularly around product discovery, to make this easier. For example, they have implemented a personalized search function on their website that takes into account all product attributes, such as the shape of the watch, the color, the range, and the product name, so that searches are highly relevant and can also be customized according to the individual preferences of each visitor. The idea is to show results that are already relevant and then make them ultra-relevant through personalization. Beyond that, they didn't stop at product search because on-site product search is one of the pillars of product discovery. But beyond that, there are also product recommendations that can guide the user throughout their journey on the different pages. So they will be able to implement cross-selling, but intelligent cross-selling. They have also implemented FOMO messages on their product page based on the visitor segment. For example, if visitors are a little hesitant, or at least appear hesitant according to artificial intelligence, we can show them messages that create a sense of urgency, such as how many units are left in stock or how many times the product has been purchased in the last 24 hours. That's the kind of strategy they used, and the result is that they also sorted their automated category page, where they were able to push products with a very good click-through rate and a very good conversion rate to the top of the category pages and, above all, ensure that products that are out of stock are a little more limited in terms of visibility, so they appear lower down on the category page. So what does that allow? It allows the user to be directly exposed to trending products that are available for purchase on the category page. So the result for a brand like Casio is that product discovery is ultra-optimized wherever products can be found, so they are highlighted in recommendation blocks, search results, and category pages, and it also saves time for the operational teams who have automated all of this. I believe they have 26% of their recommendation block representing 26% of the site's sales. So that shows the relevance of recommendation blocks, where for a visit to be considered a sale on a recommendation block, the visitor has clicked on a product in a recommendation block, arrived on the product page, and then added it to their cart and converted within 30 minutes. So it's not a huge attribution either. So that brings me back to the value of it all. It's also important to note that, like most e-commerce brands today, Casio gets the majority of its traffic from mobile devices. So when you have a huge catalog on mobile, it's essential to have an experience that is ultra-streamlined in terms of product discovery. We can't afford to let the user go back to the menu every time or do another search at the top. So this is done through these famous recommendation blocks where I allow the user to very easily find similar products on a product page, so they can go from one product page to another directly without going back to the search box or the menu box. Okay, so recommendation blocks are mainly found on product pages. So they can be placed on different pages, and it's the logic behind them that will be interesting. On a product page, we might have blocks of alternative products, showing similar watches, for example. On the other hand, on a category page, for example, we're not going to talk about artificial intelligence at all, but at the bottom of a category page, we can just have a global browsing history. So going back to the last products is super simple, but in terms of user experience, it's actually expected. And on mobile, it's essential to be able to navigate from product page to product page as much as possible without having to go through a whole proactive search phase, so to speak, and limit the number of clicks. This makes perfect sense, especially on mobile. For example, to understand this better, let's say you and I both have an iPhone 16, we both live in Bastille in Paris, and we log in at around the same time. We will potentially get the same recommendations or see how it works, you see, in terms of the recommendations. So it depends on the logic that is used. For example, if we want to show bestsellers on a page, such as a home page, we'll tend to want to show the bestsellers and a variety of products that can be found in the store. So, if the bestsellers are geolocated and we live in the same place, we'll find the same thing. If, for example, we're not in the same place, we'll potentially see other bestsellers. Okay. However, where it will be different is that the goal of personalization is not just One to F You personalization, it's also One to One personalization. So you said that we share certain things in common, but on the other hand, you certainly have your own affinities, and I have mine too. We're not necessarily looking for the same thing, and as soon as we arrive on the site, we send clear signals to the algorithm, which is then able to determine that Joseph is interested in a certain type of color, a certain type of material, and so on, while Camille has different preferences. So that's where, when we have personalized recommendation blocks or when we do a search, it will be taken into account to push you more towards red products, for example, while I will potentially see other types of products. So we don't have the same experience everywhere on the site, right? That's really interesting. Yes, it's still really powerful in terms of immediacy. And so now, if we continue to explore this idea of personalization a little further, let's take two user contexts as an example, which may have common attributes. Couldn't that sometimes be a little creepy? A little intrusive, for example? I don't know if you see what I mean, I don't necessarily have any good examples in mind, but I'll visit a page on a website at one point and then come back a little later. It says exactly the same thing to me. You were here 3 days and 2 hours ago, you scrolled through the product. How far can you go without it seeming intrusive to the consumer? Well-done personalization is invisible by nature. That means the user shouldn't even realize that the experience is personalized. They should just feel that the site understands what they want, what they need, and what they're looking for. So let's take a simple example of a user who is looking for a dress on the site. Thanks to personalization, we'll be able to know that this user has an affinity for a certain size. When I say affinity, I just mean that we'll realize that, for example, I'm more likely to go for products available in size M. So, behind the scenes, when I search for “floral dress” in the site's search bar, the first results will be floral dresses, which is the relevance of the results, but they will be floral dresses available in size M. As a user, I didn't see that, I don't see it. However, based on my experience, I can feel it because when I click on the results, it's smooth, and I arrive at product pages where my size is available. On the other hand, I'll notice if I click and the products aren't available in my size every time, and then I'll think, “Damn, the experience isn't great.” So that's the idea behind personalization: what you have to keep in mind is that it has to be invisible, so there's no solicitation, you don't know about it, it just has to be there, it has to be simple, transparent, but above all useful. Relevant. Yes. Okay, very interesting. Have you noticed any recurring mistakes in personalization, in terms of good and bad situations? Yes, I think one of the main mistakes is wanting to personalize everything without a clear goal. It's a bit like everything else. You need a goal, you can't just say I want to personalize for the sake of personalization because it will make for a great journey. No, you need to have a clear goal. The clear goal could be to increase the overall average order value on the site or to increase the average order value for a given segment of visitors. For example, my segment of visitors who are interested in a particular brand. I can see that they have a slightly low average order value, so how can I try to increase it? So you need to have a specific goal because if you don't have a specific goal, you end up personalizing a bit blindly and you can't really tell what's working. Another very important factor in setting specific goals is having access to insights, being able to analyze your data and identify where the problems lie and where you need to take action. So, on platforms such as Nosto, we provide access to business intelligence tools that make it very easy for e-merchants to see the performance of segments and products, and therefore to identify points of friction and say, ‘I'm going to put something in place’. For example, to be very specific, we can see from insights that visitors coming from paid traffic sources have an average order value that is too low to have a good R0S. At that point, we say to ourselves, "I'm going to have to work on that a little bit and create a personalized experience when these visitors arrive on the site. Maybe push more in all my merchandising for products that are a little bit more expensive or on which I have a slightly higher margin, only for these visitors who come from paid traffic sources, to improve my R0S. Okay. So that's the kind of thing we can put in place. It's basically data-driven personalization. Yeah, so okay, you have, let's say, objectives based on the attribution source. What else could you have? What are some of the most common use cases that you see? It's often going to be global. Here, we're getting into advanced strategies a little bit, but we're going to increase the overall conversion rate of the site, the average order value, and the average order value for the site as a whole. These are, of course, metrics that are closely monitored by our clients. But then, when we start to really look at personalization, that's where other teams come in because, for example, acquisition isn't necessarily managed by e-commerce. So the people who manage e-commerce internally need to be able to communicate with other teams to say, “Hey, your acquisition campaigns, I can help you with my tool to tailor the on-site experience for your traffic so that you get better results on acquisition, for example.” Interesting. And you provide context from the outset, so you have user context, perhaps with certain integrations with the CDP keyboard view, etc., you have a visitor who has already visited in the past, you can obtain data from other sources of attribution. Do you have other data that will provide even more context? Absolutely, there's all the data that Nosto already collects. Nosto is basically a commerce experience platform that collects, unifies, and enriches data. The type of data we use is customer data, product data, and content data. But in fact, this data can also come from different sources, as you mentioned, so it can come from a CRM, a CDP such as Klaviyo. With Klaviyo, we have a direct integration so that they can send us elements from their segment and we can also send them other elements. So what this typically allows us to do, as we were discussing at the beginning of this issue of data fragmentation, is to unify all the data and unify the tools. Backlavius is used more by the marketing teams, while Nosto is used more by the e-commerce teams. So we're going to unify the tools, and therefore unify the workflows and the data, and that will allow us to have highly advanced scenarios. To be very specific, that means we'll be able to personalize events, sorry, Klaviyo emails in a more advanced way, and we'll also be able to personalize the experience of visitors who arrive from a Klaviyo email at Nosto. Yeah, yeah, that too. Yeah, and then after that, there could be lots of other possibilities because, you see, when we have integrations like this, it also allows Nosto to send events that Klaviyo wouldn't necessarily pick up, such as abandoned searches on the site. So we can say, “Hey, this person abandoned their search,” knowing that when a user searches for something on the site, we estimate that this shows a strong intention. So all of those users behind that can trigger Klaviyo to send abandoned search emails rather than abandoned cart emails with smart product suggestions in them. It's a virtuous circle. Yeah, yeah, yeah. Really interesting. Now I'd like to talk a little bit about merchandising. You mentioned product discovery as being really important. We often say that you never get a second chance to make a good first impression, so I think we understand that it's important to arrive on the site and quickly realize that the journey is relevant and that it anticipates where you want to go on the site. Do you sometimes manage to introduce products to users who weren't even looking for them in the first place? You mentioned this a little earlier. Of course. Sure, if we always show them the products they've seen or shown a lot of interest in, it loses some of its appeal, so product recommendations are pretty good for this kind of use case, for introducing new products. In particular, they allow us to use algorithms to suggest products to each user that they haven't seen yet but that are similar to their interests. For example, if I'm interested in floral T-shirts, I'll go back to my pattern. I'm in the mood for flowers today. If I'm interested in that, I haven't converted yet, but we can show me similar products, so potentially tops with floral patterns that I haven't seen yet. So here, we're going to encourage my discovery, and we can really integrate this into different stages of the product discovery journey. It can be on the home page, the product page, the shopping cart page, or even in emails. In emails, for example, this can create opportunities in an abandoned cart email. I put a particular product in my cart, but I didn't convert. There are lots of reasons why I might not have converted, let's say. When I receive my abandoned cart reminder email, the simple email will summarize my cart and suggest that I return to the site to finalize my conversion, but there may be reasons why I didn't buy. So if we add a product recommendation block within that email, a recommendation block that can show me alternative products to the one I put in my cart, that becomes really smart because it encourages me to return to the site with another range of products that might interest me. So there's that, but there are also other, slightly more global, slightly more generic ways of inspiring people. We were talking about bestsellers, which is one example. So there's that, but there are also other, slightly more global and generic sources of inspiration. We were talking about bestsellers, which is one example, and look shops on product pages, of course. That's something that works well for discovering other products. I'm looking at a pair of jeans, and it shows me a top that goes well with the belt so that I can buy the whole look. So that's something that works well, works pretty well, and another element that I think is interesting to mention here is user-generated content, which you're familiar with, especially on the visual side. So it's all these visuals that showcase the product through real people, which can sometimes, and often, be more engaging and inspiring for users and visitors who see the product in context. And in these cases, it also allows you to inspire and introduce new products in a very natural way, sometimes even guiding the customer to the point of purchase. Yeah, great. And I guess you do b testing too. Absolutely, which is really important, and we'll talk about that maybe later, but of course, you always have to test. I'm not stating the obvious when I say that, but basically, NoSTO integrates, and that's why we integrate it directly into our platform. It allows us to test all elements of the site, but also to test campaigns and personalization strategies against other personalization strategies so that we can always optimize in line with this logic, but also to personalize, sorry, to test a personalized experience against a non-personalized experience. That way, we can really see the uplift from what I've put in place. Do I show a blog with personalized recommendations or do I put up a best-seller? What brings me the most conversions? The same goes for my search. Do I personalize my search? We were talking about Casio earlier, which highlights products with high click-through rates at the top of its category pages. Did that lead to an increase in revenue compared to when you didn't do anything? That's interesting to know. You can do incremental tests. Yes. Okay, great. And if we look at the product page itself, what would you say are the best elements to customize a little bit without overdoing it, so that it remains fairly readable and to the point, if I may say so. Product pages are really key pages because they're the type of page that attracts a lot of traffic. So you need to have strategies for these pages. Again, these strategies aren't necessarily very visible, but they can have a huge impact if they work well in relation to the traffic these pages represent. So, first of all, on the product side, it's very simple: you can add blocks of personalized recommendations, as we said, look shops, which are extremely effective, allowing you to offer a complete look based on the product being viewed. Beyond product recommendations and product discovery, we haven't talked about this much yet, but there's also content personalization. We can also display different content depending on the visitor. Typically, on a product page, we can have a small ribbon like Asos does, a small banner next to the product description where we display a message for new visitors. So, for example, new visitors arriving on the site will be shown that they can get 10% off their first order. But that's a message we don't want to show to someone who has already bought from us because, overall, it's going to generate frustration. Yeah, that's clear. So that's where it gets interesting. We say, “Well, no, not to those people, for example,” and then it's a subject you know well, so maybe in this case I'll show the program, something related to the feed program. So maybe someone who usually orders from us instead of 10% off, I'll remind them how many points they currently have or how many points they need to unlock another reward level, and that's where integrations can be interesting, especially with loyalty tools. And that's content personalization, which is also very relevant on product pages. Then there's the source of entry, which we talked about a little earlier, but a use case that has interested many of our customers is all the visitors who arrive on a product page from Google Shopping, for example. We are able to identify these visitors, put them in a segment, and offer them a different experience. These are visitors for whom we pay a certain cost, so we absolutely want to avoid bounce rates. They need to stay on our sites, and we need to convert them. So that means we get a click, they arrive, and we have a page for them, but that's it. So what we can do, for example, is when I arrive on a product page and the product is unfortunately not available in my size because it's out of stock, what we can do is offer alternative products that are similar to the product currently being viewed. So what we can do is offer alternative products, i.e., products that are similar to the product currently being viewed. These can be similar in terms of style, in terms of many things, and in terms of price as well. But here we're going to place the block really close to the product upstream at the top of the page to make sure it's visible. We have a few seconds to convince the visitor and prevent them from clicking on the previous arrow to continue browsing and searching on Google. Instead, we want them to continue browsing our site so that we can ensure the profitability of our campaigns and control our acquisition. Okay. And there, you raise a point that is quite important, which we may have glossed over a little in terms of the structure of the catalog, to ensure that we are able to further customize the way it is displayed. Do you have any best practices for this data structure? Yes, absolutely. This is a really important point because good merchandising always starts with excellent knowledge of the product catalog. So, the first thing to do is to feed as much information as possible into the tool we use, for example Anosto. When I say as much information as possible, I mean everything like categories, sizes, colors, prices, and not just that—there may be custom attributes, but we could also send the margin, for example, because when we send the margin, we can have really advanced merchandising strategies. So that's the first thing, and then you need to use a tool like Nosto, which will analyze all this information, but above all how the products interact with each other. So you can see that product A is often bought with product B, that when visitors view product C, they tend to buy product D, and so on. This creates what is called a product graph, which is a scoring system between products that provides a detailed understanding of the catalog, as well as a detailed understanding of the user and their affinities. This is where we can then achieve ultra-high-performance merchandising. It's all really very interesting. It gives me a lot of ideas, and we also talk a lot about cross-selling and upselling. It's a bit like the holy grail of the first purchase, as you mentioned earlier. It's true that acquisition costs are so high now that you really have to try to get the most out of your hard-won customers. So cross-selling and upselling fit perfectly into that strategy. So what are your best practices when it comes to these two topics? Yes, well, cross-selling is definitely very effective, especially on shopping cart pages. Typically, on the shopping cart page, I've just bought a pair of shoes, and I'll be recommended the cleaning product that goes with them. It's seamless, it's Usky. Spoiler alert, it's soap and water, you've got it at home. But you still need it. You need it. So in any case, it's expected, it's almost expected. I mean, it's like when you go to a physical store, it's expected. Yes, absolutely. We're going to offer the little additional sale at the end of the process. So we're simply replicating that online. It's seamless, it's expected, it's integrated. So it doesn't disrupt the process at all. It simply allows us to increase basket sizes and average order values. So that's pretty much the basics. The same goes for fashion stores: complementary sales on product pages, which we've already talked about, are also pretty basic, but they're a real help to the purchasing process. In other words, they're the basics, but if you don't have them, you're missing out on a real purchasing aid for visitors. But what is often underused is what we call post-purchase upsells, i.e., upsells right after the purchase. It's an offer that we can make right after the order confirmation page and before the thank you page, and this is mainly on Shopify. Okay, got it. And here we can be really creative. For example, I've just confirmed my order, and you can show me a product that I showed a lot of interest in while browsing, but that I didn't order. For example, we can offer me 5% off for 10 minutes. If I buy within 10 minutes, I get 5% off the product that caught my eye but that I didn't buy. It's super effective. Another use case, for example, would be in the consumer goods sector, such as beauty products. I've bought a cream and just finalized my order, and immediately afterwards I'm offered a pack of two creams for 15% off. That's extremely effective and useful, and it gives a sense of exclusivity. From the user's point of view, it gives a sense of exclusivity, like, “Hey, here's an offer just for me for 10 minutes,” and above all, it's extremely user-friendly because I don't have to re-enter my bank details. That's what's interesting: I've placed an order, I'm in a buying mood, so to speak. The hardest part was taking out my card. I'll take your paninis. Exactly, but I took out my card, I started buying, the hardest part is done. Now I just have to click on a button, I don't have to re-enter my bank details, and I can add an offer that I wouldn't have had access to otherwise. So that's super effective for boosting the average order value for brands, the post per chase. Yeah, that's really powerful. I've already fallen for it once or twice. It plays a big part in FOMO too. Basically, if you close the page, it's over. Exactly. That's pretty powerful. Great, I think we've covered quite a lot, and now, of course, we can't do this podcast without talking about AI. While preparing for this interview, we identified several types of AI, so I'd like to go over them with you to see how they relate to what you've presented. Which is the most relevant in this context? I think this is a subject that listeners don't necessarily have a lot of visibility on. So, we have predictive, visual, semantic, and generative AI. Could you quickly explain the differences between these types of AI and what they are used for in the customer journey? Of course, you've hit the nail on the head. Sometimes AI can be a bit of a mystery. We don't really know what's behind it, what it's used for, or how it works.behind it, what it's used for, how it works. So that's what we've tried to clarify at Nosto: to show how we use AI. We actually use the different types of AI you mentioned, which will serve the customer, or rather the end user, the end customer, but also e-merchants and operational teams. First, you mentioned predictive AI. Predictive AI allows us to analyze a vast amount of data, draw conclusions, and identify trends in order to predict future purchasing behavior. In very concrete terms, this will enable an e-commerce brand, for example, to see which visitors have high potential, to create a segment of high-potential visitors. So basically those who are... Exactly, high-potential visitors, not bad. That's not bad. Basically, those who are close to making a purchase, and then once again trigger mechanisms that are appropriate. It could also be based on weather forecasts, for example, if it's going to be very cold in a week or two, I'll suggest a certain type of product because that's what sells best overall during these weather episodes. Interesting. That's one thing, for example, about predictive AI, which we then linked to semantics. Semantic AI will really help with site search. What it allows you to do is understand the user's real intention when they type in a query, even if it's complex. Today, users tend to use searches on e-commerce sites in the same way they use Google. So sometimes it can be extremely complex, but that's where semantic intelligence comes in. It allows us to understand the user's intention and therefore provide relevant results, even if the query is complex. And that's extremely important because a study we conducted showed that 70% of users go straight to the search bar when they arrive on a website, and 80% of them leave the site if the search results are not relevant or effective. So this shows the importance of intelligence like this to provide ultra-relevance. Then we have visual AI. I'll try to be brief, but visual AI is more of a lever that will help teams because it will allow them to imitate human vision. So it will be able to recognize within an image a style, a pattern, an emotion, a shape, or a flower. A flower, for example. And behind the scenes, for an e-merchant, I think it will be particularly interesting in the fashion sector, as it will enable you to do visual merchandising, i.e. show products that are visually similar without having to go through ultra-complex rules about whether the product is such and such, so show this or that. Yeah, yeah. It's done automatically by visual intelligence, which will match products that are similar or visually different according to the merchandiser's wishes. Okay. And finally, generative AI, which is what everyone is thinking about, I imagine. Exactly. It's the one that everyone has been talking about a lot recently. It's an AI that we've also integrated at Nosto to really streamline the work processes for retailers. So, for example, we use it to generate automatic lists of synonyms for searches. Right. So I think any e-merchant who has worked a bit on improving the relevance of their search results knows that it's extremely time-consuming to analyze all the pages, all the queries that return zero results, and then decide which synonyms to use, and so on. Now, artificial intelligence does that for them, suggesting a list of appropriate synonyms, and then they can accept or reject a suggestion and retain control. And it's really interesting because, if I understand correctly, the first three types of AI have been around for 10 years, and it's generative AI that's a little newer in this field. And from what you're saying, generative AI is used more on the merchant side to save time than on the consumer side. At least, that's the impression I get. Is that because generative AI isn't powerful enough today, or because we're not confident enough to let it generate results for users directly, or because it's more interesting for us as a company to control it a little and save time internally? I think there will always be control mechanisms; I think it's essential. After that, it's up to us and the developments we make. I think that we're also going to move towards a lot more things that will be, let's say, customer-facing, visitor-facing, especially with the development of technologies such as conversational technology, like JPT chat, or search, which is being redefined a bit. So I think we're clearly moving in that direction. And yeah, that was actually a question I've had in the back of my mind for quite some time. In the era of GPT chat, where, as you know, I saw another LinkedIn post this morning, you'll soon be able to integrate your Shopify checkout into JPT chat, meaning you can buy from JPT chat. My first question about this is whether you've already managed to structure your product pages and recommendation blocks so that they're optimized for generative search engines. I don't know if that's the right way to say it, but you know what I mean. What's your take on this? Well, one thing's for sure, these are topics that interest us. We have a team, we have a lab that's into everything related to artificial intelligence, so they're very interested in all these topics too, from an agency perspective. I mean, really all of these topics, so these are things we're thinking about today, of course. It's not something we offer directly, but of course these are elements that are at the forefront of our teams' minds. Yeah, and you mention AI and teams, so I have two questions about that. On the one hand, internally at Nosto, and more generally in companies in general, what are some of the barriers to adopting AI? I don't know, sometimes some of your customers are afraid of being replaced by a super AI merchandising agent that will do everything for them. Do you see behaviors like that today among your customers or even within your own company? I think that AI, especially with ChatGPT, was a bit complicated to use at first, but I think people are becoming less apprehensive because they're realizing that it can't really replace the human brain, at least not yet. However, it does provide a significant boost to productivity. So there you go, I think that at Nosto, at least, it's very uncomplicated. On the contrary, we're very encouraged to use everything. It would be a little strange for a tool that has been rooted in AI since the beginning. I mean, for us, it's not a new topic. It's just that Nostto was born, well, as soon as Nostto was born, it was integrated with AI. So at Nostto, at least, we're very encouraged to use all forms of AI that can improve our productivity. What do you use, for example? I use Chat GPT quite a lot. Yeah. Mainly because I write a lot of content in English, so it allows me to do an initial check on my writing before giving it to the content teams who are going to use ProFFRID, for example. It's not Chat GPT that writes for me, but it does allow me to have sentence structures that are often better than what I could have done myself in English. So for me, that's the main benefit, and then I use it for lots of other things, but it allows me to summarize long documents, for example, and things like that, so I can get through my work much faster. Yeah. You still have to reread quite a bit, though. Yes, yes. But still, it does a lot of work that can't be ignored. That's for Nosto. But you were talking about the client side, and I have a lot to say about that because I've worked on the client side, so I've also had clients that I've helped with their personalization strategy. And the first obstacle I often heard was this slight concern about not knowing how to control or master what AI will actually deliver, the results it will bring, and also not understanding how it works. Typically, for example, a retailer might want to personalize the products on their category page, so they want to show personalized products to Florian, for example, and push products that are red or floral for me, but they still have very important profitability targets, so they can't just say, “Alia, go ahead and show personalized products at the top of the category page.” It's too strategic to say, “OK, I'll leave it completely open.” What they can do is use control mechanisms to add a rule to their merchandising to say, “I also want you to take into account the products on which I have a better margin, for example, in this weighting that will allow you to sort the category pages.” What will work better afterwards, and what I always told them, is to test this famous rule with my margins added, or this rule where I basically let everything be controlled, and see what happens, what performs best in the end. And in some cases, it may be AI alone, in some cases it will be AI with controls, so filters, merchandising rules, and so on. For me, what's important and what I was trying to tell them is that good technology today must integrate AI, but it must ultimately be driven by humans, so it will actually integrate all the control mechanisms that allow it to be tailored to the specific strategy of an e-merchant. Yeah, absolutely. That's interesting. I had another point. Yeah, go ahead. I also think there's something we hear a lot, which is a bit of a fear of the black box. That is, we don't always know why a product has been recommended. I'm very into product discovery, but why was a product recommended, why is a particular product in a particular position in the search results, and so on. And sometimes you have to be able to explain this because other people in the company will say, “Hey, that's funny, I typed in that search and I saw that product there, I don't understand.” So it's important that the person in charge of merchandising is able to say, “OK, that's why.” Typically, at Nostos, we've implemented what we call “inside scoring.” So we show why a particular product was chosen, what was actually taken into consideration to promote that product in that position, whether it was merchandising rules, AI, and so on and so forth. So that's also important, to be able to explain these elements to the user. And then you can clearly understand the whys and wherefores of each recommendation that is made. How each recommendation is made and avoid the black box. Exactly, we are able to say what influenced it, it's the same thing, it's a scoring system. What influenced the weighting, in fact, the promotion of this product rather than another. And you can change the weighting if you want to. Okay, absolutely. We always review the merchandising rules or filters, and we can add lots of things. OK, great. OK, great. OK, great. Just a quick question to finish up. If we try to look ahead to the next few years, you said that at Nosto you are currently working on other slightly natural developments in AI, whether it be agents or conversational AI. Can you explain a little bit about what the big projects are at Nosto in particular to follow this trend, but also this revolution in generative AI and the impact of generative AI on consumer behavior, which is becoming much more commonplace, For example, having a chatbot that you communicate with in real time is no longer awkward when it's AI, whereas a few years or even a few months ago, talking to a bot was always a horrible experience, to be honest. So, I'm curious to know how you see this at Nosto and then your personal view on the impact on the ecosystem. I recently found myself searching for my vacation on Chat GPT, and I thought to myself, “This is really concrete, even when searching for hotels, asking it to find a hotel with certain criteria, etc.” There were some nice things, and it was pretty refined, but it's true that these are clearly new uses, even if they're not new in themselves. But Google with Alexa is really conversational, meaning that we're now used to, or at least starting to get used to, searching in a very conversational way. And so that's what we're going to have to take into account to redefine how products are searched for, and that's what we're currently studying at Nousto, how tomorrow's customer journeys might look by integrating this very conversational aspect. Can you give us a little spoiler, or is it too soon? It's too soon, but there are lots of great things in store. We can't wait, anyway, the teaser is very well done. And you, personally, in the e-commerce ecosystem in general, more from a customer experience point of view. What's your take on that? I think it touches on that a little bit, because, as I said, it's what I've also experienced on a personal level. I think that, going forward, we're going to have to be proactive in what we suggest to visitors. Not just wait for them to send us signals, but also be able to offer them a personalized experience right away. Okay, great. Well, Camille, we're right on time. That was a really interesting conversation. Thank you very much. Thank you. And then we'll see you for episode 2 once the new features are out. With great pleasure. Thank you.