Webinar: Promo Forecasting 2.0: How AI Helps Retailers Plan, Execute, and Analyze Promotions

Awesome. Well, let's get started. So, welcome, as I said, to the webinar. Today, we're going to cover promo forecasting 2.0 and really talk about how AI can help retailers uh plan, execute, and ultimately analyze the results of of promotions that they do in their business. And so, we've got a lot of content to cover. Um, but before we we jump into sort of the details, I'm going to go over introductions. I'm going to introduce us and then I'll talk a little bit about our business, um, what Leafio does. Um, and then we will we will dive deeply into promotion. So, uh, I am Brad Mitchell. I'm the vice president of North America for Leafio and have a background in in retail and supply chain technology. And alongside me today is Anna Irma. on a uh sort of leads our our business um really around the world and just has an immense amount of experience working with clients to achieve uh

tremendous results across across the retail space and Anna can say more about herself as well u uh when we get into her slides. >> Yeah. Thank you for the hi everyone. >> Yeah. Yeah. Great. Let's talk a little bit about Leafio. So, uh, not to have a whole history lesson, but Leafio has been around for over 15 years and, um, we've worked with hundreds of clients around the world in the retail space implementing solutions and, uh, delivering value. Um, we work with all kinds of companies from sort of small to midsize retailers all the way through sort of global retailers with hundreds or even thousands of stores. and uh and we use the experience that we've gained across the these uh these projects to really, you know, continue to drive value and and really bring thought leadership and knowledge to new projects that we do. So, that's a bit about us. In terms of the clients that we serve,

we serve retailers, as I said. Uh we serve retailers, you know, there's kind of our a lot of the logos of of folks that we've worked with over the years, but um big into grocery and gas station convenience. We do a lot in that space with retail. We do a lot in um you know, DIY, even CPG manufacturers, health and beauty, pharma, so on and so forth. And so here's kind of a smattering of some of the folks that we work with. Um yeah, in terms of what we offer for our uh our clients and people, you know, retailers around the world, we've got this unified platform that really manages the end toend supply chain and merchandising process for the retail environment. And so there's a lot going on in this slide. I love this slide because it really shows off all of our powerful functionality. Today we're going to be focused on sort of the bottom left of this slide which is promotions. Um but I wanted to touch on you know some of the

other aspects that we deal with um across the platform. Starting with assortment up top in the top right here. So we have an assortment performance module which uh really helps us answer the question for our clients what should we be selling and where should we sell it right? So one of the most critical questions that we have to ask ourselves as a business is are we selling the right items? And so assortment uses datadriven you know AI recommendations to to alert our customers that hey we should we should change assortment um or the range for a store or cluster of stores so on and so forth. Then once we kind of know what we're selling, our shelf efficiency module here on the right hand side answers the question, well, how do we visually merchandise this to delight our clients and to achieve our goals in the actual physical store in terms of sell through, revenue, profit, these things. Um there's a big compliance aspect to that as well to make sure that you know our planagrams are actually executed and and the stores

are getting updated to to match the strategy that we defined for the assortment. So that's sort of the merchandising side. And then here on the left hand side um you know inventory optimization then answers the question for us. We know what we're selling. We know how we're merchandizing it in the store. But how much inventory do we need to keep both at the store level and at the central warehouse level to make sure that you know our inventory and our supply chain is aligned with market demand. Right? The market is changing all the time and it's very critical for our clients that you know we sort of align our replenishment strategy uh to that demand. So this module uses machine learning powered demand forecasting. Um we automate replenishment to the extent that we can with our clients in many cases in the sort of 99% um automation to uh to really manage replenishment and um you know where it makes sense we've got a really big sort of fresh uh goods or ultrar goods um

component to that as well. And this is where promotions play a role right so we've kind of answered these three questions. What do we sell? How do we merchandise it in the store? How do we replenish it to make sure we're aligned to demand? And you know, one of the complex components of of all of this, right, is really managing promotions which are becoming more and more prominent across the retail space. So today we're going to dive deeply in this. Um uh we're going to first talk about the key challenges in promotions. We'll talk about how machine learning and AI can really be you know a gamecher in terms of managing executing uh these promotions. We'll talk about improving the forecast accuracy. How do we accurately forecast promotions and then you as critically uh we need to be able to measure the financial results from those as well. Are we getting the right you know results out of our promotions that we set out to from a strategical level. We'll talk about Agentic AI and how we can boost efficiency, right?

We've got a lot of promotions across a lot of stores. If you're on this webinar, you probably have this. And so, you know, how do we do this at scale and scale it across our business effectively? And lastly, Honor will walk us through some of the lessons learned um across our extensive client work and uh we'll even show sort of a case study with some results related to inventory and promotion specifically. Cool. Now, we're in the content. So, let's get started. Um, we're going to start very high level, right? Why are promotions so hard to manage? Um, the reality is, you know, this is a necessary evil for for, you know, clients that we serve, right? Promotions are a critical part of our strategy that we have to execute on, but they're very challenging and and complex to manage. Um, you know, we're growing the percentage of promotional items, it seems like, every year that we have out in our store and retail space. And so, you know, it's just it's just becoming more and more of a burden and ultimately

consumes a lot of time for our resources in order to execute those. It's not enough to set up the right promotion with the right price. We also have to make sure our, you know, replenishment is updated accordingly. We've got to make sure our planagrams are updated accordingly. Even, you know, small things down to like price tags, right? There's a lot of little things that have to happen in order to make sure a promotion is effective. And uh and then obviously we need to be able to measure the the results in terms of of how how they went. We can't stop, right? Uh our customers love promotions, right? This drives loyalty in our business. Um you know, they they come to know and love it. It's a major part of our strategy as a business. And so it sort of becomes this addiction that snowballs a bit. And then we find ourselves, you know, as um as category managers or or planners, etc. really just wow I'm focused so much on the promotional element of my business uh that I can't add value in all the other areas and so uh we're

going to talk about how to how to sort of address that today and I will turn it over to Anna to go a little bit deeper and talk about some of the more um deep challenges and and really how machine learning and AI can help. Yeah, thank you Brad for starting this uh presentation and talking about our company and introducing our company and highlighting one of the most painful areas in the area of the supply chain and inventory management as promotion campaigns because we are talking to the customers on the daily basis to the existing customers our existing customers potential customers and we definitely see that promotion management uh in The area of the inventory management and forecasting is one of the most challenging um challenging areas to be to be managed because uh even if like the orders are automated, the replenishment is automated still like in the majority of cases the promotion is done promotion

forecasting promotion orders are done manually and uh that really complicates this process because in terms of the promotion it is in general pretty like challenging to make the forecasting but when we are talking about the promotion uh it's even more challenging. So why does this happened is first of all that the price elasticity is not linear because we don't know if we when we decrease the price we don't know like how uh like how many times the sales would jump um jump up it can be like three times five times sometimes even 50 times but sometimes it can be zero. So it's really important to find like this point on which we get the maximum effect from the price decrease. uh and um the promo is widely advertised by the retailers sometimes even like some television some media they are like publishing newspapers uh and that's why it's very uh important for

the retailer not to go to the out of stocks because in this way like in the case when there is a promotion campaign that's been advertised it's a very very high risk to uh to lose the loyalty of the customer if he comes to the stores and he won't find the item he's looking for. >> Uh and in terms of the promotion, it's like much much more risky than in terms of like the regular item. Um and very sensitive to in terms of the loyalty. Uh so that's why that's the reason why demand planners they are like being afraid of this out of stocks. So they are going into the um like ensuring the highest availability level and making sure that uh there will be like goods on the shelves. So that's why like in the majority of cases the promo forecasts and promo orders they are like overdone and as a result we have the post-promotional overstocks. And here the one of the challenges is that it's

not always visible on the side of the retail business. Not every retail business is evaluating and seeing these overstocks and calculating these overstocks and the cost of these overstocks. >> So, uh what makes it like even more complicated during the uh promotion campaigns especially like we are now I'm pretty sure that retailers are getting ready for Christmas promo and like retail is like overloaded right now. uh and uh for example if we having like martini ai that is has like its regular layout on the shelf it has like additional promote lay promotion layout for example in the form of some tree or like near the cashier desk and so on. So one SKU might have like several uh several layouts and it's very complicated to manage it from the side of the inventory both in the beginning when we need to ensure that we will refill all these additional promotional uh promotion uh promo layouts and when

we finish promotion campaigns because all these additional layouts they are potential overstocks. uh and uh what makes it like more complicated that one SKU can participate at the same time in different are the different campaigns which are overlapping and what uh makes it like even more challenging that they can be with different promotional mechanics. uh recently we notice a very high increase in the grocery retail business, very high increase of the ultrar category. So food to go category and of course like this category is also on the promotion campaign and for example we need to make it like some fresh goods. It's very close to the expiration date and we need to wisely approach this promotion campaign. But fresh category in general it's a very very sensitive and complicated category to be managed and especially it's like over complicated with the promotional uh and uh all these points they make the

promo planning very exhausting and very uh very long in terms of the process especially when it's the import products when we need to import for the promotion because many departments are involved in that we need uh to do it on the operational level to change the price tags. We need to make the additional layouts and so on. So it lasts a long for a long time. It's not always in time when we planned it to be and it complicates it even like more than uh than the regular punishment. So, uh, >> let me let me I I I I failed to say at the beginning of the meeting, too. Um, you know, we're going to go over a lot of great content, but please, if you have questions, put them in the question box for us because we'll have some time at the end of the webinar, >> um, to answer questions. So, so please enter those and, uh, we'll review those at the end. And and I think from a challenge standpoint for me, you know, looking at your slides there and you kind of talking about it, I'm I'm a supply chain guy, you know, and so I think so much about inventory and the impact of inventory and and as I'm

talking to retailers on a daily basis, what I see with promotions is that many of our our our clients and people we're talking to in the space, they're so focused on the the outcome from like a topline standpoint and and you know, launching the promotion, making sure we have plenty of stock, whatever. But we're often we often miss this very critical component which is how do we make sure we wind down the promotion in the right way with the right inventory etc because those overstocks can really take away massively from the financial result that we get on the on the top line of the promotion and uh I think this is where people really struggle struggle >> yeah for sure for sure so if we are decreasing like the margin for the promotion campaign whether like we are doing it on our cost or on the vendor cost still it's important to make sure that we squeeze like the maximum from this promotion campaign. Um so with all these uh like incoming uh like challenges so what can be really helpful here is ML and AI which can be like

really game changers here because this uh uh this technology they can work with their like high amount of the data with like data um uh different like data um data deviation s and so on with the complex data. They can perfectly do that and uh they really help to uh overcome this data overload. They really help to streamline promotional planning and ensure the uh flawless promotion execution. uh so if we're talking about if we do compare the ML model with a traditional statistical models for sure we understand that a model is the one which will be like a little bit more complicated in terms of the implementation maybe uh uh in terms of the price a higher a little bit in terms of the price but the result that we get in terms of the uh forecast accuracy in

terms of the availability of the model to adjust to different changes in the data and what is like one of the most important point to consider the complex interdependences. Uh it really brings amazing results in terms of the forecasting and the accuracy of the forecasting which help retail business uh to make the promo efficient for them and to like squeeze the maximum of it and to get the necessary effect the desired effect from the promotions. Uh when we're talking about the data uh in general like the data is a very big struggle not only in promotion but in terms of the inventory like the accuracy of the data the availability of the data but we need to understand of course that shoutout. So if we don't have the proper data in the very beginning we won't have the accurate forecast. Of course the model it is like self-educated. It collects the data but the clear the more

clear data we have at the very beginning the better results we will have in terms of the forecast accuracy. So uh what I want to say here is it is important to start collecting the data right now. Start gathering the data right now. So when you will come to the point when you would like to automate this promotion management promotional forecasting you will have enough data to do it. efficiently. So, uh there's like uh of course we take um like much data for the ML model to make the forecasting but this is just some of the examples of the data that we are using such as uh type of the promotion campaign, duration of the promotion, campaign, promo mechanics, discount level, price before and during promotion canibilization sales uh additional layout location and so on. So there are many many factors and just like some of the data that needs to be uh analyzed and considered but uh uh we inspire you to collect this

data and start having this data in order to have the more accurate forecast in future. >> Uh >> yeah obviously obviously with uh with machine learning like part of the big benefit here is that you know the model is adjusting itself and really learning about your business over time. So um you know even if we don't have a tremendous amount of data to start right we can expect that you know as we go forward through the weeks months and and any you know many many years even um the model is going to get a lot better at really understanding your past promotions the impact they had on your sales and and improve that forecast accuracy. >> Yeah. Yeah. For sure. For sure. For sure. Uh one uh more important uh uh thing to consider is that u when we are talking about promotion campaign not only promotion campaigns but in general like the forecasting but it's even more critical for the promo campaigns that we need to uh consider different weight of different factor while calculating the

forecast and it's impossible to do that with or very hard to do that with the statistical model but for the ML It's pretty simple task and uh this is the area that it perfectly deal perfectly deals with. Um so uh when we are talking about uh the promotion and inventory management in general, one of the top questions from our customer what is the forecast accuracy? uh and um it is very important uh not just to look at the forecast accuracy but to be able to like analyze the forecast accuracy and measure not only forecast accuracy. So uh first of all what influence the forecast accuracy is the assessment horizon. So uh day week quarter year it's the aggregation level. So whether it is done on the level of the company on the level of the SKU on the level of the SKU location of course the higher aggregation level we have the more accurate the forecast will be uh so

what are the methods that we are evaluating the forecasting accuracy in lithior it's their percentage error it's the Maya bias w map arms yes or there are different methods that you will see in the system and that we calculate uh But one of our top methods is and one of our favorite one. We think the most useful one is W mapper because this one allows us to evaluate the uh like the forecast in terms of in terms of the money in terms of like the budget. So uh now but when we are talking regarding the um like measurement and understanding the effect it is very important not only to look on the side of the forecast but even sometimes more important to look on the financial metrics because the forecast accuracy can be on the level of 100% but still we can have like very bad results in terms of the financial metrics. So that's why

we highly recommend to take a look at uh such financial metrics as sales the um the comparison of the sales plan with the fact the amount of the overstocks uh like the dynamic of the overstock and calculate that in money the inventory to nowhere in days uh and availability and out of stock. Of course, these are just like some of the financial metrics we have in the possibility in the BI block to take a look like deeper and uh have like very high level of the granularity of the data. Uh but this like financial metrics are like one of the most important to take a look at while when we evaluating the efficiency of the promo campaign. >> Yeah, you made you made a critical point there I think Anna which is you know obviously we're focused on the forecast accuracy. you're tirelessly working to make sure that we are pro, you know, forecasting promotions to a more accurate degree such that we have better projections for, you know, inventory,

replenishment, sales, etc. But it's not just about the accuracy, right? It's, you know, that that ties into the financial metrics for sure, but it's also about the execution, right? Am I replenishing, you know, at the right time? Am I selling out at the right time? again talking about those overstocks um you know that's going to impact the financial metrics that you have just as much as you know the forecasting accuracy so we need a good signal we need good execution and I think if we can do both of those things really well we're going to have tremendous you know boost in terms of results that we're getting out of this this hard work that we're doing in managing these promotions >> yeah for sure and when we have uh all these like data and when we were planning promotion campaign not in the excel files or like in different like majority of system. But when we have like precise like system for making a planning, promotion and controlling the execution of course it like increases the uh uh increases the speed and increase the operational excellence of that.

Uh so uh uh one of their also helpful things in terms of like the not only promotion in general but in general uh in terms of like usage of the system. So in Lifo we are implementing AI agent that will allow to deal with first of all maybe to interpret some of the um some of the result of the forecast of the promotional forecast and forecast in general because at Leo we have we specialize as Brett mentioned in the beginning not only on the promotional part but we take like wider we have like the uh management and automation of the whole inventory management process. Uh so Agentic AI can be really helpful in understanding how this forecast was calculated, what were parameters considered during this calculation, what influence this forecast and so on and also what we notice from the like execution point of view and uh like that

there can be like many mistakes in terms of the promotional planning. So the AI agent will help here to identify these mistakes and to help the demand planner to deal with these mistakes and to correct them with just like couple of buttons here with just a couple of clicks here. Uh so um um trying to like summarize that so we we've been working with the retailers for like more than 15 years in terms of the inventory management in terms of the promotional forecast and planning and when we implemented the uh in their promotional planning and promotional ML model to calculate the forecast we made some lesson learnings which we want to share with you. So first of all uh I will share that in the forecasting area and in the processing area. Uh so it is impossible to have the accurate forecast without clear understanding about the aim of this forecast. So we need to understand why do we need this forecast and what we're going to do with this

forecast. It is impossible. We talked about the data a lot and it is impossible to calculate the accurate forecast without the full and very prepared data. So uh we highly recommend to take uh care of your data uh like today and start gathering the data start like clearing the data. uh some models they can have like the same for forecasting accuracy in the initial stage but as we mentioned the many model is self-educated and of course we can see the higher increase in the forecast accuracy while the model is like self-trained and gather like more data and use like this more quality data for the forecasting um and also like two important parameters while talking while making and thinking about the forecast for about promotion planning is seasonality and price elasticity. So we need to consider that and we need to uh take them like um into calculation while planning our promotion campaigns. Uh

regarding the process area, it is um sometimes it is even more important to uh take a look at the financial outcomes than just like the promo forecasting and forecasting in general. And during the execution and running of the promotion campaign, it is imposs it is important to rely not only on the forecast but to rely on the actual promotional sales because we don't always know like how the promo will go and there can be like some adjustments during the running of the promotion campaign. So it's very important to consider that uh and uh their promo planning should in general be integrated into the general replenishment and inventory management. Uh so it is like a part of the replenishment like regular replenishment and forecasting process. So it's important not to divide promo and do it like manually for example like in different like Excel files and so on but to integrate that and to have it like

simply integrate it when the uh promo can directly like influence the readjustment of the order without need a man like for demand planner to think about that additional and to keep that in mind. Um it's very important to evaluate and to analyze promotion each and every promotion campaign and to uh make some lesson learnings uh because we should understand how this promo like influence the sales what are the results and make some make some like lessons learn some lessons for the future and for the future planning of the promo campaigns and of course we talked about that but it's important the precise finish of the campaign is as important as the precise beginning because when we if we have like the successful promo but after the promotion we have the high level of the overstocks we don't like calculate these overstocks we don't know how like how much this will influence our business in future

>> um I wanted to uh show one of the examples of our customer it's uh they've been working with the inventory management with our inventory management uh module for years And uh when we developed the promotion model they were the one the first to to implement that. Uh so when we implemented that we got amazing results. Uh they have like really professional team on their side. So they really helped us to like implement that properly and to get this amazing results. So uh what we have is 56% of the post-promotional overstock decrease and at the same time we got the availability increase on 2%. So it was during the promotion it was 98 and 9% which is like a very high level for the promotion campaigns it and it's about the grocery retail business. So uh of course we need to uh take a look at uh these results not only from the

perspective of just like promotion campaign but we need to have a wider look on how it influenced the inventories in general. So like what is the influence of this post-promotional overstocks in general on the inventory what is the financial effect on that? So Brett do you want to share that? >> Yeah sure. So if we if we think about you know um the results she just shared right we've got a 50 plus% reduction in promotional overstocks um we've got increased availability in the case of Novvice uh you know up to 99% promotional availability so these these are dramatic in themselves right you think wow that's this is going to drive business results but what does it actually do across your business right um if if I save or if I you know reduce overstocks on promotions by 50% that equates to roughly 10% of inventory in total. Um the additional availability on promotions represents 30 basis points of sales. And we need to be able to think about this in terms of profit, right? How does this impact the bottom line of

my business? And in this case, you know, um I would ask everybody on the call like think about what this would mean, a 10% stock reduction, a boost in sales to the tune of 30 basis points. Uh what does it mean for the business? Is it 2% profit? Is it 20% profit? Etc. It's actually it actually equates to a 15% um profit impact. And so it's really dramatic especially as we see pro promotional element uh element of um businesses growing right we're doing more we're doing them more often. We're doing them seasonally across many SKUs etc. you know we're seeing more and more uh of an impact and that's going to ultimately impact that profit in a very very significant way. And then I think also in terms of results, you know, for Novas as an example, you know, they get these these tremendous financial results, right? More effective promotional planning yields this great profit impact, but also it's about managing it, right? How hard is it to manage it without a structured system in

place that connects with replenishment in our case across the platform connects with even planagramming and assortment. Um, it's very hard, right? We've got items that are on multiple promotions at once potentially. We're we're maybe managing dozens of promotions at once across all of our stores. And so just the, you know, the manual work and the communication that has to go in to make sure replenishment is right, you know, merchandising is right, all of these things. To the extent that we can automate this across the platform, not only are we saving tremendous amount or or you know, you know, impacting the net profit in a tremendous amount, but we're also saving an immense amount of time for our people to be able to focus on real value added task. Um, and so this is the other the other big winner here. So >> yeah, I want to say that wow, that's impressive like plus 15% of the net profit increase. That's like really impressive. That's huge. >> Yeah. by tweaking and improving something that we're already doing today and saving a bunch of time while doing

it. Um, so it's it's unprecedented, but it's also it's also available. And uh, you know, if you're on this webinar, you've probably you're probably managing promotions in some way, shape, or form, and this is what what drove you to to join us today. We thank you very much. But, you know, um, reach out to us. Let us know how you know how your business works and um, how how you manage promotions today. We'd love your feedback. We'd love to know how it works. We'd love to show you how in the Leafio platform we can uh we can add value to your business. And uh if you're thinking, I'm not sure if it'll work for me. Um let's have that conversation. Let's let's have that conversation. You can email honor or me directly after the webinar. Um we'll send out the webinar recording so you can reply to that as well. Um but get in touch with us and uh we will definitely talk about that uh as it relates to your business. Um well, thanks Anna. That was great. Let's go through a couple questions that came through during the webinar. >> Sure. >> Um let's see. So, oh the first one that

came up when you were on the slide of uh you know talking about the data required right to and you're kind of pushing people like hey start start collecting this data now because this is going to be important in the future. Um so the question relates to that. It says, "What's the minimum data quality and volume needed to start using AI around promotions?" Uh, and so maybe you can have a stab at answering that, but I think, you know, we need to think about what's the historical longevity of data that we need and then what are the types of data that we need to add to our data stack that probably most retailers aren't aren't uh using today. >> Yeah. So uh uh I would like to say that every business is different actually and every business has its own specific if it's even like the retail chain of like one convenience stores and the same like not the same but also convenience stores still it's like different business it's different countries and we are working globally so it really depends on each and every like specific of the business. So we are very uh like precisely

analyzing business processes while the start of the project during the implementation to better understand your business to better understand the specific and to understand your goals of the business at this like at each and every point because some business want to scale some business want to increase the revenue and so on. So there are different aims on the different stages. So that's why it's very important in each particular case to analyze the situation based on that to provide the requirements for the data. Uh but for sure like as a like at least that uh like it cannot be like ideal in the majority of cases it's not ideal. That's why we of course try to be flexible. So there is of course the necessary amount of the data that needs to implement the system and some additional which we uh can for example be gathering during the implementation itself which is not critical for the very beginning but will be important when we like start scaling for example.

Regarding the period for which we need to have uh the data usually we require like one year but of course the more data you have uh the more like period of data you have it is it will be better because for the ML model like more data we like fit to this model the more like a precise like forecast we will have. >> Go ahead go ahead on. >> Yeah just wanted to say that like the second half of the question is regarding the ROI. Um I want to say that in general for the inventory we have the ROI so up to six month by on average but uh it can be also like two or three months. Uh but we will see the first results after like one or two promotion campaigns uh when the system will show itself in terms of the promotional planning. >> Yeah. And I think the ROI builds over time, right? Because the model is building over time. So you know as we get further into it we have more data, we have more intelligence that the model is training itself on. that's really allowing it to be more accurate. Um and

uh and so that that provides tremendous ROI over time as well. >> Yeah. >> Great. Um another question here. Well, yeah, we kind of answered that with the historical data uh that we need to get the best promo forecast accuracy. Another question, um if you are running a chain of stores, how would you customize the promotions for each store to match the demographics during a certain period? obviously except to festive times. So I think I think the question really relates to you know can are these c these promotions that we're driving and forecasting are they are they customized at the store level and uh or are they more aggregate than that? So you know can we really match the demand that we expect for the same promotion across a different set of stores that has very different market demand and different market patterns. >> Yeah. Uh so we analyze the demand patterns of this particular store for which we are planning the promotion campaign and of course like we are planning the promotion promo campaign on

the level of the store not on the level of the like retail chain in general. Of course we can plan like the general supplies and so on but still like the demand and the forecast like should on the level of the store. So we need to understand how each particular store behaves itself. So it is done like on the store level and there are like different settings different like uh algorithm in the system that allows to be like very sensitive and very like adjustable to the demand of like each particular store. >> Yeah. Yeah. I mean, so so it's definitely at the store level, right? And and you know, if you run a promotion on items, it's going to be really looking at the price elasticity and the changes of the stores specifically to make sure that we're accurate, uh, you know, accurately forecasting at the store level, uh, the store skew level. So definitely yes. Uh, one question just came up. Uh, let's see. Let's say I run a bill level promotion 10% off, you know, a bill value of $100, right? So

10% off if you spend a certain amount. How would the system identify it? Will it tag it as 10% discounts on all products or is there a different way of tagging these promotions? So yeah, uh this relates to I think promotional setup like how do we set up and and make sure we're capturing the right items um on a promotion that we've that we've run in the past or want to run in the future. I don't see this question. Uh is this >> in the Q&A panel? Um >> yeah, I think I think uh this is an interesting question about the bill. So it's like, hey, I spend a total amount of volume and I get a discount, right? How how would the system identify it? I think it would identify it like like you said in the former, right? It would tag each of the items on the bill as a 10% discount um for capturing the future and uh and the model would use that to drive the forecast going forward for

additional promotions related to those items. I think that's how it would work. Will the tool help to identify a skew level promotion versus a bill level promotion in the past? I want to say that a bill level promotion it's more about the loyalty system. So it's not from the side of the like inventory management because what matters for us in terms of like the supply chain is like the proper planning and to ensure the right level of the inventories uh for like this particular store. So that's why like on the level of the bill we need to understand that from the loyalty system which we have by the way also in lithium portfolio but from the side of the like replenishment we need to say we need to understand there like the store level and the promotion level but not like the bill level. >> Agreed. Another question here Anna how

hard Hi there. How hard is it to integrate the AI forecast into our ERP so that all departments use the same number, right? So, you know, I guess related to pro promotion specifically, you know, as we create the forecast, how hard is it to get that data back in our system so that you know, people on the merchandising side, for instance, you know, are are running off the same data that we're using uh from an inventory standpoint? >> Um, I can answer this. I mean, >> very relatively easy. Yeah. I mean on a nightly basis typically we're um you know exporting the forecast data at the skew level or at the aggregate level back to the ERP and uh you know putting that in your in in your database so you can use that however you want and in many cases yeah this this forecast will drive many other functions of the business. Uh so certainly very very easy to get everybody on the same page in terms of what we're forecasting. Uh, one more question here on I think for you. I'll let you drive this one. So, you have the slide which depicts the

importance of features for the ML forecast model and it says promo discounts are not so important in comparing the sum of previous sales. How does it work? >> Uh, for sure it is like very important. I think that it was like maybe some misinterpretation of the data on the slide because what I meant is that the level of like the discount is not linear with the influence on the sales because we should have we should go to this like point when we can have like the high increase in the sales because for example when we have like 5% decrease in the price it will influence I don't know like for example like two times in sale when we have like 10% it will influence I don't know three times but when we have like I don't know 15 it will influence like 10 times so we need to find like this point of course the price decrease is one of the most important points uh uh for the calculation of the promo forecast for a cast and that what

influence but what I wanted to say that it's like very hard uh to understand this point when it becomes like uh like critical increase high high increase in terms of the sales uh but you know it's a very interesting question so uh you can definitely contact us we can arrange a call out side of the session because of course we understand that we're limited in time here so we can talk about that during the meeting and we will be happy to so please contact us >> they said got it thanks yeah I think that was a bit of a misinterpretation because definitely the uh you know the pro the discount the promotion is one of the critical factors that we're looking at to really build the inflection point of of price and uh and uh sales, right? And and forecast. And so this is one of the great things that the the model looks at and does. >> I wanted to add here that right now we developed the uh modeling part in the promotion that allows to like enter the level of the discount and see like how

much it will in increase the sale like influence the sale. So we have this kind of modeling of their like the level of the discount on the uh increase in sales. >> Yeah, it's very exciting and something I think our our clients uh and people we talk to are asking about all the time. Well, how can I how can I project the level of impact that a promotion will have before I run the promotion? And and the answer is, you know, we sort of build this curve of price elasticity based on, you know, your past promotions. and uh we're going to be able to to let people um sort of see the optimal outcome of that inside the platform. And so another uh you know not only a you know great feature in terms of getting the numbers right but also in terms of time savings and really avoiding a lot of manual analysis and work to make sure we're driving towards the right promotion promotional settings ultimately. >> Yeah, for sure. For sure. >> Awesome. I think that is all the questions we have I believe. Let me go Yeah, that's everything. So, um,

yes, that's everything. So, thank you so much everyone for attending. We've got a lot of people around the world on the call and we very much appreciate your time and hopefully you learned a bit and hopefully um you're excited about the future of promotion management and really how AI can play a role in uh impacting your business in a massive way. As I said earlier, if you're interested in, you know, how this would work for your specific kind of business or you have further questions um about the uniqueness of the promotions that you run, please do not hesitate to reach out to us. We'd love to talk with you, get your feedback, um even show you some software, which I think is always fun. So, definitely reach out to us and uh yeah, thanks thanks everyone for time. Thanks on uh for the great presentation and we will see you guys next time. >> Yeah, thank you Brad for the company. I think it was very insightful for our visitors. Thank you everyone for joining us. Uh hope to see you during the meeting. Have everyone good day or evening where wherever you are in the world. >> Yes. Cheers. >> Bye-bye.