Hello folks, afternoon wherever you are in the world. Thanks for joining us today on this Accelerate for Better Peak. We're going to talk through some strategies, some tips to hopefully get us all ready for Peak. And thank you all to the customers that are attending today. Of course, this will be recorded and shared afterwards. There'll also be some key takeaways that we'll also email. In the meantime, as always, there's a Q&A section. So if you've got any questions during the course of today's session, feel free to post those. And I'll look to try and raise those questions during the course of a session as is appropriate. Now, before I hand over to Andreas, who's our expert here, and many of our customers will know him, I want to set the stage on, as you would expect, two kind of stats in readiness for Peak this year. No surprise, the first one, you know, e-commerce continues to grow. This year, we're hitting about 6.8 trillion in terms of global e-commerce in 2025. So we've seen a huge increase there of about 8.8, 8.5% over last year. Okay, so good for our industry. But equally, we all know, it's getting harder and harder and more and more competitive. And it's in that vein where, you know, we all are very aware that Peak is pretty, pretty important. It's kind of break on May, generally, representing about 32% of our entire annual revenues that we see in e-commerce across the year. So we understand the importance of it. And what we have Black Friday and Cyber Monday, very fast approaching. If you're not prepared already, be preparing now. Now is the time. And that's why we scheduled today's session. Okay, we're here to try and tackle today, some of the practical things that we should be thinking about as we start that preparation, especially in the context of ThreadHopper, and how to find repeatable strategies to try and make it the best success that you can. And the most optimised platform that you can in readiness for this year. So on that note, I'm going to hand over to Andreas, who's our expert. And as I say, many will know of him. And I've seen the content already today. And there's a lot of very good tips and suggestions as we start with that preparation. And good luck to you. Thank you, Andreas. Thanks, Imran. Yeah, hello also from my side. Welcome to the webinar. I'm Andreas. I look after experience consulting. So that's essentially the function in our business that helps you, our customers, prepare, optimise, understand. So essentially my role is to understand your goals, your shoppers, well, our software too, bring it all together. And hopefully at the end have something meaningful, something positive, something with actually bringing money onto the table. So in that function, I've been working with our customer base, well, for a good number of years now. And the peak season is, of course, always the busiest time for me and for the wider business. So there are lots of questions and perhaps not entirely surprisingly, many similar questions from many of you and our other retailers. And today in the session, we're trying to look at some of the common patterns that we're doing. So there are lots of different things that we tend to see every year and maybe share some insights, some ideas, some inspiration, what we can do to basically to make your peak better by taking these challenges into account and looking at some of the possibilities you've got within the Fretiber system to help you with those. So there are five topics we want to cover today. So the first bit is mainly around rankings. So we'll have a look at the composition of ranking cocktails and how that ideally changes during the peak season. We'll have to have a look at stock fragmentation, which if you're anywhere near fashion products is perhaps one of your biggest worry the moment you go into any sort of promotional period. We'll also have to have a look at different business goals and how to conflict with each other. So this is how we can perhaps balance the promotions and bestsellers and new products or whatever else the business goal might be. And then for points four and five, we'll have to have a look at the user experience. Namely, what do we do with facets? And what do we do with search when we come to the sales period? So let's start with the ranking cocktails. And for all five areas, you'll see that as that's basically this type of pattern, we'll look at four areas. So who's this relevant for? What's the challenge in the first place? What do you need to do or to have in your data, for example, to work around it? And then we'll look at the approaches themselves. So here for the ranking cocktail, basically any retailer that has any meaningful promotional periods. So I suppose that's anyone in the B2C space and possibly a good number of B2B retailers as well if you are running promotions or other short-term events with basically a significant change in user behavior. And the big challenge there is that product performance is not stable. So for all of our ranking cocktails, we can essentially see them as a prediction, a predictive model. And the main input for them, if you're aware of your own ranking cocktail composition, is almost certainly historic sales performance, which is brilliant in most of the year. But then the moment you go into a sales period, there might be a challenge there. Because let's say on the first day of the sale, the user behavior changes so drastically that you might not be able to infer much from the previous data, just because of how quickly user behavior changes. And to make matters worse, depending on what type of promotion you run, you might even have some products that simply don't have any data whatsoever. So this could be special buys that you really just get for, let's say, back Friday. It could be flash sales that just by their nature wouldn't really have any any historic data or anything else along these lines. That's basically a short term offer with very little historic information you can rely on. Good news here is that as long as you've got some product performance data, so for example, if you connected Fretiber to Google Analytics or Adobe Analytics or the Fretiber Insights system, then we can probably pull enough data to make sense of it, even if you perhaps can't rely on historic data as much. So just to illustrate the challenge a little bit more, let's have a quick look at this simplified example, admittedly, of just two products. And if we were to just look at how a brand new cocktail would act. So let's say for product A, the one in purple here. So that's just the typical pattern for product with a, let's say, heavy markdown or some other heavy promotional activity. And the rest of the graph is essentially just the timeline. I've picked weeks, but depending on your business, it might be different. It could be days, could be, well, it could be something else. And an artificial product performance score. So basically in short, the higher the better. Now, if you were to look at, for example, the point in time three here, if we wanted to predict how well a product does, if we were to just look at historic data and if we make that time frame too long, we might draw the conclusion that product B must be better than product A because historically it's been doing better. But that was before the promotion started. So if we rely on very long time frames, we are bound to make wrong decisions. So we need to find something that's basically, that's more accurate in the timely sense that it's looking at a shorter period of time. But at the same time, we need to balance that with the amount of data that we've got available. So if we just look at, for example, a single day of sales for most of the year, at least, that wouldn't be enough to get any sort of meaningful information. So we need to find a good balance. And the second thing that you might see in the pattern of your products is the sensitivity to price changes. So for example, if you are a retailer selling a mix of own brand and branded products, you'll find that the same sort of discount, let's say 20% markdown will have very different effects on your own brand products versus maybe sought after third party brands who perhaps don't normally discount anything at all. So then a 20% markdown would have a quite significant change in user behavior as an effect. While perhaps a 20% markdown on your own products could possibly not be, or is possibly not visible as much. So that's just something to keep in mind when we look at the data points and the time frames we look at. So what can we do about this? I've already alluded to it. So we need to look at the time frames. And our challenge is to get more data, more information, more signal in a shorter time frame. That means we can play with two things. We can play with the time frame itself. So make it shorter. And we can also play with the data point itself. Usually my recommendation is if we need to go to a shorter time frame, we should step away from sales numbers simply because even in the best of sales and the best of promotions, there simply aren't all that many product sales. So instead, we should look at something that simply generates more data points in the same time frame. Page views. So the number of product detail views is probably the best candidate there. Or if you want to have a bit of a hybrid ads to basket is usually also a good choice. So with ads to basket, you still have, let's say, a factor of 10 more data than if you would just look at sales. But you don't normally have the challenge of, for example, a heavily biased number of page views simply because of inbound marketing activity. If you're maybe sending loads of traffic to a single PDP as a marketing activity. Then again, you might draw some wrong conclusions. But in short, a mix of page views and ads to basket is probably something you want to look into, especially at the beginning of a sales when you've got that changeover. And then a second trick that we can use here is to change the strategy over time. So we shouldn't think of it as the Black Friday or promotional period or peak ranking cocktail. We probably should think about different stages in that period. So perhaps something like this, where we basically have one strategy for that very beginning of the promotional activity. So exactly that time when we can't rely on historic data. And that will be different than the bulk of the promotional period. I've even split it here into an initial and a final phase. In practice, depending on your business, you might be able to combine the two. And then at the very end, we have the same sort of challenges in the beginning, just in a way the opposite. Now the products might not be marked down anymore. But again, because of that, we might not be able to use historic information as much simply because they're back to normal now. And we might again draw the wrong conclusions. If there's a highly price sensitive product that's now gone back to a higher price point. So again, we might draw the wrong conclusions. And then at that period, typically you also have new season products or slightly different types of products that you want to sell. So then we also have to balance newness with the best sellers or historic information. So basically in short, we probably need another transitional period to go back to that regular trading, that BAU ranking that you use for the rest of the year. Now for each of these phases, I've highlighted the attributes that you probably want to include with a higher weight during that period. If you look at the very beginning, so the first few days, yes, you want to use the attribute with the most information in the shortest time possible. So in other words, so in other words, you probably want to use page views over the last, let's say between one to three days. Possibly at the basket, if you've got that. And if you've got enough volume of, well, not of sales, but of at the basket. And depending on how your business is structured and how your promotions are structured, you might already have to include fragmentation of stock. Now that's something that we'll come to in just a minute. But if you're a fashion retailer, you'll be familiar with the challenges of having most of your sizes sold out for some key products. And yeah, because that's not a good experience, we'll have to include that. Especially as we go further into the sales period, when the challenges of fragmented stock becomes even more visible. So yeah, beginning of the promotion, use a short timeframe. And basically that's pretty much all you can do at that point, unless you've got some really specific data points, for example, for big bets that you as a business have maybe bought into specifically, special buys, that sort of thing. Then if we go into the main period of the promotion. So a few days away from the start. By this time now, you should have had some sales and some revenue numbers that we can hopefully rely on. Again, we need to use a short timeframe. So here we're probably talking about something like seven days for the sales numbers, which is usually the sort of timeframe or the short timeframe. We can rely on to actually have numbers for a good portion of your catalog. We still want to include the short term data points like ads to basket and page views simply because we're still in an early stage where the user behaviors might still change the offers that you put in there and the promotional messaging that you send might still change and especially might change the user behavior. your default, your standard ranking cocktail that you use for the rest of the year. Perhaps with the main difference being that fragmentation is hopefully weighed a bit more highly simply because there will be more fragmentation of your sizes and of your stock. And typically at this point also I wouldn't include newness, which might be the other big difference to your default rankings that you might use for the rest of the year. That's of course assuming that you don't really release any new products or at least not that many during this period. And then, as I hinted to earlier, at the end of the promotion, when you go back to basically regular trading, you essentially replicate the same sort of approach that we've used in the very first days of the promotional activity. activity. So again, short term data points, page views as basket will be your best friend there. Fragmentation definitely because you might have a few products left or few sizes left of products that haven't sold through during the promotional period. If you do have long term products sales performance data, and long term here is something like 30 to 90 days, you might also want to include that again assuming that the promotion isn't that long. So that's why you could look at the product performance effectively before the promotion started. So in a way, historic data can be useful again. But there are plenty of caveats there. So the duration of the promotion is one thing. Change in season is another thing. So if we, for example, after the holiday day peak, if we were to include 90 days of sales performance data, we're basically looking at autumn when it will already be winter at that point, and basically pretty much after Christmas. So this might not be that interesting anymore. But for any other promotions, yeah, it's a good one to include. And if you do have new products dropped during that time, newness, or at least some way of handling new products is something you want to include. Now, all of these ranking rules, you can apply them broadly. So you wouldn't need to, for example, change everything in a result modification, but you can just apply that as effectively an override to your BAU, to your default navigation ranking, for example. Let's have a look at the second key element of ranking cocktails. And that's the stock fragmentation. So we alluded to earlier, this is mainly for, or it's absolutely key for fashion retailers, because you do have your multiple sizes and possibly other variations like colors, but it doesn't mean that's not relevant for other retailers. So I've had B2B retailers. So I've had B2B retailers with technical catalogs complain about the same challenge. That's perhaps not so much the T-shirt sizes, but perhaps screw sizes or radiator sizes. The challenge is the same though. If one of those sizes is not available, then the product probably won't sell. And if your catalog has any sort of these variations, typically that means if you have any sort of these variations, you can't sell it. So if you have any sort of these variations in the Diffredible Data Model as variations, then this is a challenge that's relevant for you. And what exactly happens here is that for the Diffredible Ranking Cocktails, as you might know, we're basically ranking products. So products in the sense of one tile in the grid. And each of these products represents, of course, multiple sizes, possibly multiple colors, depending on your data model. And the availability is, of course, for each SKU. So basically each image, each tile in the grid represents a number of different SKUs, which may or may not be available. And if too many of them become unavailable, then the historic information we have got from a product might not be reliable anymore, especially if they sold out really quickly, which of course is a common pattern that we see in heavy promotional activities. So we need to make sure that we compensate for it. Otherwise, as you've probably all seen before yourself, that if you look at the top of the list and you see the products that you really like and you click into them, go to the detail page only then to find out that they're actually not available in your size. It's probably one of the worst experiences for a shopper, that they spot something they like only then to find out that they can't have it. And so it's maybe not entirely surprising that whenever we run AB tests around drinking cocktails, the stock availability, fragmentation, whatever you want to call it, tends to be the number one driver for any sort of performance and results in an AB test. So as long as whatever data you use to model this is reliable and accurate, you should be able to see an improvement, especially if fragmentation is a challenge for you. So what can we do about that? There are many different ways. That's the short answer. But they all center around the idea of essentially telling the system, the Fredipa system, how well a product would suit typically the average user. There's a slightly different approach if we go down a full-on personalization route. I will leave this out for today simply because it would be a bit more specific to each retailer. So if you've got any questions around that, we can look at that separately. But yeah, for the majority of retailers, we basically have to model how available is a product for a given user, and in this case, an average user. As I said, there are different approaches. So there's, I guess, the most basic approach would be to just say products available. Yes, yes or no. So for example, if one variant is available, then we consider it as available. Or the other extreme is we only considered available if it's available in all sizes. Hmm. That's perhaps a little bit black and white. So some retailers have gone a bit further and maybe defined core sizes. So perhaps if a size, small, medium, and large, if they're all available, then the products flagged as are mostly available. Or some retailers use percentages of sizes available. So let's say if the product initially came in 10 sizes and two of them are available today, then maybe they get a, give it a score of 20%. But that doesn't take into account basically how important each size is. Which brings us to what you see on the screen here, which is my favorite approach. And the idea behind it. And the idea behind it is essentially that we're creating a score depending on the importance of each of the sizes that's still available. So in a basic example, for example, if we knew that only a size 8 of this product's available, then perhaps we would give it a score of 20%. If we knew that all sizes from 7 to 11 are available, then it would get a score of 100%. And so on. And so on. And these size curves, if you talk to your buying team, they almost certainly have these available in-house. Because that's exactly how they're buying the number of SKUs for, basically for each style. And yeah, we can feed that into a little enrichment. So you can do that on your side, of course. Or you can talk to us if we can do it on our side. But the logic is always the same. So there's essentially a mapping list of weights for each size. Typically, this needs to be done per category. Because of course, the size 8 in men's shoes has got a very different importance than the size 8 in a woman's dress. So you'd want to split that into, basically, the different key categories. And then for each product, you just add up the weights for all of the sizes that are available. So it's a relatively straightforward logic. It generates a score that's reasonably intuitive to interpret. So if you've got a score of, let's say, 20%, that means that an average user's got a 20% probability of finding their size. And that score, you can just stick into the ranking cocktail and mix it with the popularity of a product. And it tends to give you a good mix of highly relevant products and those that are actually available in most sizes. If there's a fringe size missing, it doesn't really penalize it as much as with perhaps some of the other methods. Which, yeah, makes it, I think, an all-round sensible approach to use for any retailer with the stock fragmentation as a challenge. If you wanted to, you could take this further, of course. So I've used the same sort of idea with a good number of retailers in the past. So, for example, we had a jeans or a retailer famous for their jeans and denim products selling in the Netherlands versus in Japan. And then, of course, say, a size large might have a very different importance for the Netherlands than for Japan. Or if you've got distinct user groups on the side. So maybe you've got a plus size section. Then again, you might want to create several scores. So one for the typical categories, the average categories, and a separate score for the plus size categories. Just with the same logic, just with the same logic, just different weights. And, yeah, this is something I'd really recommend doing. So that solves the availability at the top of the list. And if you wanted to, you could even take this one step further. And use a little trick that I always like to try out. Because all of this is, of course, only relevant if we don't know the user's size. The moment they have self-selected by using a facet, for example, the fragmentation is not really all that relevant anymore. Or perhaps we could even argue that we might want to do the opposite. So we might want to invert the fragmentation score to promote those products that are highly fragmented. So only really available in a small number of sizes. And what the user's selected just happens to be one of them. So this could help you, for example, sell through the remaining units of an older product. This could be good for, well, both for your warehouse. You don't have single units of really odd fringe sizes sitting around. Could also be useful for the shopper. Because if there's only one unit left, chances are it will be discounted quite heavily. And maybe there's only one unit left because it was such a popular product. So there's something you could try. And in Fretable, you'd effectively just create another ranking rule. And you just specify with the trigger that if the user selected a size, and that's what you see on the screen here, then use a slightly different cocktail, either with the fragmentation taken out or with the fragmentation inverted. So that's a highly fragmented product gets a higher score. And yeah, so with one different data point, you can model all of these scenarios. Let's have a look at the next area, which is still related to rankings, but we're now going to think about result modifications as a possible way of solving the data. of solving the challenge. And here we are, we're essentially thinking about what do we do if our goals perhaps are a bit more black and white. And that's something that happens a lot during the peak season, when we have to deal with very distinct groups of users. So perhaps if you've got a set of bargain hunters as a persona, and maybe somebody else looking for gifts, or perhaps even just trending products that have been added new to the side, ranking cocktail might not be the best solution, because it will try to find an average product. So one that's okay for one persona, and okay for the other one, but perhaps not as great as if we were to just look at the best product for one group versus the other. And again, I'm not going to touch on the personalization route here, which is another way of solving this. I'm just going to look at the result modifications for something that you can do today without having to think about perhaps changing your implementation or adding some of the other features that we offer. So let's keep it simple for now. Let's look at what we can do with basic minimal adjustments. And the good news there is that we typically don't really need to add any data points. As you can see here, markdowns, discount percentages, that's probably something you've got already in your data feed to Fretbo. So we can just reuse that. So here's an example of what we could do. So if we know that there's one user group who is most likely interested in new in-products, maybe one other user group interested in markdowns, and a third one in just bestsellers and user group in this is probably in the broadest sense. This could also be internal stakeholders. So maybe if your brand marketing team insists on making sure you still include some new in-products, it's the same solution. Yeah. And we're essentially using result modifications here to create essentially different strategies on the same page, but we're not mixing them. We're really, we're keeping it a bit more black and white. So we have one really new in-products on position three in this case here, which of course, because it's dynamic rule, it will update itself if there's something else new in. We have two products that are heavily marked down. Again, they're dynamic, so you don't have to worry about adjusting it if they, for example, sell out. Normally, I would also include extra filter criteria. So what you, what you don't want to have is, for example, a fragmented stock in there. So any of these groups and especially marked down ones would have some restrictions on the, for example, fragmentation score. So that here, yeah, the, all of the sizes or most of the sizes available. What you also notice is that I haven't really defined anything for, for positions one, five, and six. And if we were to go down further, the list, there wouldn't be anything there either. So these are essentially just left us as is to be filled by the, the ranking cocktail. So this serves more of the, let's say the average user without highlighting any, any business criteria specifically, like data markdowns. So each of these, these, these manually or these dynamically defined groups can have its own ranking cocktail. So you could, for example, define a different approach for the, for the markdown ones and a different one for, for new in. In practice, you often don't really need that because adding the, the restrictions on, for example, needs to be marked down 30% or more, or needs to be marked down 50% or more. It usually does the trick already. Again, there might be a few different edge cases depending on your business, but typically you only need to define the groups. You need to define the, the criteria and you can do that relatively broadly. Broad here in this case also means that because this rule doesn't really, doesn't really mention anything about a category. You can apply the same rule to a good number of your category pages. So you don't have to, to, to, so you don't have to worry about maintaining, let's say one different result modification per category. So one little note that I'm sure our support teams will, will like to hear that I'll mention this. If you do apply this broadly across the entire catalog, just be careful that, that you're not applying it or that you're careful with, with pages that contain lots of items. So if it's tens or hundreds of thousands of products, then we just want to be reasonable with the number of these slots that we define. So if you were to, for example, define 100 slots for a page with 100,000 items, then Threadwell essentially needs to re-rank that page 100 times. So you probably want to test that in that case. It will normally handle it, but it's naturally more computationally expensive. But yeah, this is one way of how to deal with conflicting requirements, especially if it's about really, really different types of products and perhaps different personas that would really be interested in one and not the other, as opposed to a good balance of the two. So this covers the bit around product rankings. Now let's have a look at the other areas of the user experience, namely facets and then in a second also search. So for facets, basically the main thing that we can make sure that users can actually use discounts. Even if perhaps as a business, we don't really want to highlight promotions, if there is a promotion running and it's been messaged and you've actually have that bargain hunter shopper on your side, then we probably want to make it relatively easy for them to actually find markdown products. So in other words, this approach is relevant for pretty much everyone, perhaps except for the, maybe the luxury brands that perhaps want to have a slightly different nuanced approach to the, nuanced approach to the promotions. What you need for this is typically a attribute that models the discount percentage. Again, chances are that you'll probably have got this already. Or if you, if you're willing to put a bit of effort into it, you can, for example, use a calculation based on the current price or versus the previous price or something like that. But the idea tends to stay the same. So here, what we, what we should at the very least do is create a discount percentage filter. On the right hand side here, you can see an example of how this could look like. And you probably want to create the, the, the filter with a high priority so that it's quite visible. How visible depends again, a little bit on the retailer. So some would really want to have this at the very top if the, the entire site messaging is about promotions. Others might, might perhaps still have the, the navigation elements like categories first or brands and then the, the promotion. So that's a little bit up to you where, but I wouldn't do is, is collapse it and hide it and well, basically try to, to keep shoppers away from it because it's not the time for it really. Then in the rules itself, you probably want to use brackets like, like we've got here. Because what I've seen in the past is that if we were to use the, the default options, then you might end up with things like 30% to 49%. As a shopper, I always, whenever I see that, that's, that doesn't really help me because yes, 30% is perhaps my threshold of what I'm interested in. But I wouldn't say no to 50%. So in, in other words, you want to define lower boundaries for, for pretty much all of your, your rules. And you probably want to include zero in there. So, or sorry, anything above a zero. So just to show any sort of, of discount. And you can use the, the, the, the alias option in the, the facet definition to, to define that. And you probably also want to make sure that, that's the, the formatting of your, your numbers are, is right. So this particular example, the, the attributes uses, this integer numbers. So 25 stands for 25%. Depending on your data model, it could also be 0.25 to, to represent 25%. In which case, just use the, the alias option to just give it a, a pretty name. And yeah. So that's, that would be a very basic way of, of setting up your facet. Just getting a, a lower boundary, making sure it's displayed in a, in a pretty way. And making sure it's visible. And perhaps also making sure that it disappears after the, the promotion ends. Again, assuming that's, that you don't want to show it all the time. So use scheduling with a date and time trigger. If you've got that, it, it does help. It takes away some of the, the, the grant work that you might have to do otherwise, just before the, the start of the promotion. And finally, this brings us to search. So in, in search, there honestly, isn't too much different for the sales period than for the, the rest of the year with one exception. And that is, you will get some, some searches for promotional period, promotional terms. So you will have somebody looking for, for sales or, or discount, or perhaps even the, the name of your promotion, if you're given it a specific name. And you want to make sure that you get results for that. Depending on, and again, how your data model is defined and, and how these promotions are defined in, in your source data system. So, um, the, the, the, the solution will vary slightly, but in short, you want to make sure that's, that products that are on sale also have the information that, uh, this is on sale somewhere. And ideally this should be a piece of text. So not just a discount percentage, for example, which we can also use, but it's, um, it's much harder to define and it will require a small change request. So we can make it searchable. So the, the main options are essentially listed here on the left. Um, if your, if the word sale is for example, in, in the, the category tree. So let's say if you've got a dedicated sale category, then, um, you probably don't have to do much. Um, if you said for sale, you should only get sale products. You might have to create a few synonyms. Um, so just to, to cover all of the, the typical words like promotions and sale and, uh, and discount. Um, but that's a ones off. And you probably want to do this, uh, globally. So you wouldn't want to say, um, let's say shirts sale and, uh, bags sale and so on. You probably just want to say that sales the same as discount is the same as promotion and so on. Um, if you, um, if you, if you don't really have the word sale in your, in your catalog, then, um, yeah, we, we might have to think about a few different options there. So we could use, uh, really a brute force approach of just ignoring the word entirely. Um, unless you really mark down everything in your catalog, I probably wouldn't go down that route. Uh, so that's what I've, what I've marked down here. So I've got to, uh, uh, uh, to use a stop word. So that's essentially telling search system to ignore certain words. Um, it might be an alternative if you simply can't get the, uh, the, the search to work otherwise. Um, but perhaps the, the best option would be to, to also make sure that's. That whatever the, the source is for, for the information that has products on sale is pretty searchable. So, um, step one there is to, to basically check where in your product data can I see that a product's promotion. Um, so this, this, yeah, as I said, this could be the categories, this could be a dedicated promotion attribute. It could be like the, the badge that a product might have assigned to, um, or something similar. Um, make sure that's included in the search index. So in the list of searchable attributes and then enabled, uh, for actually the active search itself. Um, he can reach out to, to our customer success or support teams to, to double check some of these. And, um, um, yeah, uh, that's should ensure that you can actually search for the, for the terms. If you do have some, some cases where, when non-discounted products come back, uh, when you search for sale, you can also try the, the result notification that I've put as a screenshot here on the right hand side, which is, uh, actually blocking, uh, those products or variants, um, um, that have a zero percentage discount in any search for sale, but, uh, test us first on a test system just because different data, uh, structures might react slightly different to this approach. But yeah, basically in short, um, try to, to get the, the information as products on sale into the, the search index and anything else. So that's essentially work around, but we, yeah, we, we appreciate that. Uh, you might be quite short on time, but, uh, to get this sorted. So please feel free to, to reach out to, to double check what's the, the right approach. And these are the, uh, the, the five areas that I, I generally see most when preparing for peak. So here I was especially sure to summary, um, for rankings, just use a shorter timeframe, get more information at the same time. Um, um, and, and, and use that, uh, uh, avoid the, the fragmented product near the top. Um, so get, get an attribute that really bothers that make sure that, uh, if you do have different, uh, user personas on a site that you, that you perhaps, uh, really mix them, um, on the, the same page as opposed to just with a, uh, with a cocktail. And then help, uh, those, uh, those, those bargain hunters that, to try to find, uh, products with facets, um, and or searches, just make sure that's, that you actually get some, some options for them. Thank you. That's yeah. That's what I would recommend. Thank you, Andreas. And hopefully this has been, uh, helpful. I've got one question that's popped up actually, Andreas. Um, I've, I think you've covered it, but let's just double check. Okay. Um, so you mentioned adjusting ranking strategies multiple times during a promotion. What signals or data should I be paying closest attention to, um, and what to decide when to, to switch a strategy? So I think you might've covered it, but maybe let's just quickly go over it if that's okay. Um, so if I understand it, the question right is essentially, how do you go from, from this to this and so on? Um, there is a, I suppose that there could be scientific answer. Um, I'll skip that one. I'll go straight to the, to the one that I would actually use in practice. Um, it depends on the, the amount of data that you tend to have, uh, for each product on average. So, um, for example, well, on the first day, you, you definitely want to stay in this product. promotion or in this set of rules. And then, um, if you've, for example, got, uh, sales numbers or page views data one day, then, um, and, and let's say seven days is the next one. Uh, then you probably want to stay for, for something like three days in here before you go to the next. So just so that the, the seven days is mainly filled up with a promotional period. If you had three days as a, as a data point, then I would probably stay in this one for, let's say for two days and then move to the, to the three day data point. So essentially if, if the data is mostly from your promotional period, um, then it's fine to use. Okay. And I know we speak at length about, we could all benefit from more AB testing, but obviously this is a very short window of time, quite volatile potentially. What's your take on AB testing during this period? Um, yes, it's the answer. Uh, it, that, that, there's a few caveats said, of course, um, as you said, that the timeframe is much shorter and, and typically there's, there's always this desire to, to test individual changes. Uh, because when, when you're preparing for promotions, you, you prepare individual changes. Uh, but that's, that can be a bit of a, of a pitfall just because you won't ever find it or get enough data. Um, for example, test the, the first days of promotion rule for the dresses category. So, um, you, um, you need to find ways of, of how to combine these, uh, so that you can, you can still draw some conclusions. So I would, for example, say instead of testing the dresses category for this, just test it, uh, if it makes sense to, to have a, a different approach for first day of promotion versus, versus initial phase, uh, and do that across the catalog. And that, that hopefully gives you enough of a, of a timeframe or, um, sorry, not timeframe, uh, sample size to draw conclusions. Um, other biggest caveat is that, uh, any of these learnings are really only applicable to that particular timeframe. So anything you learn about the first day of promotion, you shouldn't really try to reuse that, for example, for the, the end of the promotion. Maybe to some extent for, for the switch back to, uh, uh, uh, business as usual. Uh, but again, with, with some caveats. So yeah, uh, at the very least it will help you for the next big promotion. Okay, cool. Thank you. And I guess that's it. We know that it's not business as usual. Um, it's a very busy time, a very volatile time. Um, so there's just two things I just want to draw folks' attention to. Um, hopefully, you've got some good advice and tips and some suggestions in terms of preparation for this year, but equally, we are here to support. So please reach out to your customer success manager. Um, if you want some more intense peak readiness with our team on a one-to-one basis, that is always an option and a service that we can provide. And the earlier you request it, the, the better. Um, because it is always a high demand request as we start to prepare. So don't leave it until the last minute, please. The second thing is please keep an eye out on email communication. Um, our support team, our CS team will be sending out, um, notes and communications in terms of what we are doing to prepare for this year and ensuring, uh, the stability and the scalability of our platforms because we need, we realize how to prepare for this year. And how important it is to keep it, keep the lights on and make sure that we are supporting your success. So please keep an eye out on those comms because they're there to keep you updated as to what we're doing. And on that note, because I've seen it too many times, um, please ensure that you have a record of all of the escalation paths within our business. Please ensure that we have all the correct contact details and escalations on your side. Um, these hygiene things, I know they can sometimes be, be like administration, but they're really important. And when something goes wrong, we're reliant on them. So please ensure that you are working with your CS team or CS customer manager to ensure that all the escalations are in place and all of the details are in place. And that way we can best support you. On that note, thank you, Andreas. And good luck everyone in their preparations. Uh, we'll close now and, um, appreciate all your time and energy. Thank you. Thank you.