Retailer's Guide. Key to inventory optimization in retail.

Hello everyone. It's It's a pleasure to be speaking here today at the conference and I think there's going to be just a little change because I'm going to be uh delivering the presentation on in English and I hope it's going to be convenient for everyone. My name is Vlad Bezborody. I'm the head of international growth at Leithio and Leithio is a cloud solution provider for retail uh companies uh that optimize uh that optimize and automate supply chain. Um the topic that I'm going to speak on today is um about the inventory optimization because that's that's where we hold the majority of the experience and there are actually a lot of keys uh to optimize the inventories um and I'm going to touch this topic from the perspective of uh trade

promotion management. And uh the insights and findings that I'm going to share with you today actually come from the experience we've had over the past 10 years with more than 150 projects that were implemented in more than 15 countries with a team of roughly 200 people. And we had the opportunity to work with uh various retail companies ranging from um health and beauty, DIY, uh specialty retail, grocery to supermarkets. So uh we work with these companies on a daily basis and we have the opportunity to actually understand the pain points and the challenges they are facing. Um and we see a lot of companies coming to optimize and automate the supply chain um with pretty common

uh from pretty common um for pretty common reasons. And these reasons typically fall under one of the three buckets. The first type type of companies are looking to um optimize and improve their operations. The second type of companies are looking to um acquire support for the scaling efforts they're going through. So, in the growth phase. And the third type of companies actually want to make more money with what they already have. So, uh the way we're able to deliver this for these companies is through a range of software solutions that that starts from inventory optimization uh that actually ensures balance of inventories throughout every stage of the supply chain. Um then followed by the next step, which is shelf space optimization. We deliver

this by relying on the data to uh automatically generate planograms and help companies manage end-to-end merchandising process all within one solution. And then the third point um of our software platform is the promotion management. And this is very interesting topic because we see an increased demand uh from a lot of different retail companies to help them manage the forecast uh for the trade promotions. So, having received a large number of requests, we we've taken on the challenge and uh started exploring this topic in more details. And when when uh we looked at this process, we noticed that the forecasting forecasting part that companies that are looking to solve is not actually the only problem. And forecasting is

just the tip of the iceberg. For um what we've noticed that below the forecast, there is a vast majority of processes that are very complex and require cross-functional coordination between the departments and the suppliers. And there is a a a growing trend that every retail company is sees an increased number of items that are in in promo. And depending on the vertical and the location, the the number of items in promotion varies from 20% to 70%. When we looked at the trade promotion process, we wanted to compare it with the regular sales process to understand what are the differences and to develop a plan on how to work with it. So, we've looked at the regular sales.

They typically start with a assortment planning that moves on to the negotiation phase with the suppliers followed by merchandising, purchasing, logistics, finance. Well, this is all pretty common, right? And this is done on the annual basis. Then we we've looked at the promotional sales and the process turned out to be pretty similar but a little bit more complex. That started with the assortment management, selecting the right SKUs, selecting the right mechanics for the promotion with further negotiations, merchandising planning, purchasing, logistics, human resource planning. So, everything looks pretty similar. But there was a dramatic difference with the speed and the um uh the frequency because the uh the sales cycle uh for regular sales is done on the annual basis, while the promotional is done on a weekly basis or bi-weekly basis, which pull puts a tremendous

strength on the operations. So, having noticed that, we decided to to to take a step back and understand how this process fits into overall picture. So, there is there's a promotional strategy that retail companies are following. And then once the promotional strategy is uh is set, there are four phases for the promotion to be um to be completed. They are planning, preparation, execution, and promotion exit. For each and every stage, there are sub-processes that are in itself pretty complex and require uh involvement of a lot of different departments. So, for SKU selection, supplier selection, promo mechanics, forecasting, negotiations, and there is a large number of uh people from planners to marketing managers to purchasing managers, manufacturers, suppliers that are working simultaneously.

And they're working on hundreds or thousands of SKUs that are in promo simultaneously in parallel. This creates a very vast complexity. And uh we what we've realized is that the forecasting does not solve the problem. Something else is needed in order to create efficient promotion uh management process. So, um we've also noticed that retail companies are hostages to the situation because promotion helps to generate new traffic. New traffic is needed. And if companies uh stop making promotion, there's not going to be customers are not going to be happy. What This is not a good scenario. So, um companies are forced to do more promo to drive more traffic and put more

constraints. Uh and this is done in an inefficient way because uh overall at the at the level of promotion process complexity, each and every stage must be very efficient. And if it's not efficient, the promotion is not efficient overall. So, we decided to take a deeper look at how retail companies are solving the problem of the forecasting. Now, and the way it typically happens is that there is a person who is responsible for the certain forecast. This is a planner. And there is a unique combination to how that person is working because that person relies on typically on three things. The first one would be some sort of a model that generates the forecast. That number is then compared against market situation and the gut feeling is used.

So, the intuition and the person then produces a plan, which in 50% cases or a little bit more is accurate. Well, at this level, it is very inefficient process because there are no uh there is no reliable way to have constant result uh with the forecast. And the forecast itself then translates into promotion, which affects operation. And uh if our sales increase three times, 10 times, or 50 times in some cases, uh this puts a tremendous tremendous pressure on the logistics and the operations. What if out of stock happens? What happens then? No sales happen, marketing money go to waste, and it loses the efficiency. So, um the uh people who are responsible for promo forecast play on the safe side.

They tend to over um overstock in order to have enough product. What happens next is not very important. And very few companies actually analyze on what happens post promo. So, we are in a situation where we have a big problem, inefficient process, where we can't stop doing this process. What do we do? We've taken a look at how successful companies uh solve complex problems that involve large amounts of data, and how they extract value from them from that data. Uh companies like Facebook, Amazon, Google, Alibaba, and others have learned to work with large amounts of data, extract value, and radically improve efficiency, which puts them in leadership leadership

positions in the market, and allows them to penetrate uh other verticals, industries, um and uh develop successfully as a business. So, why don't we do the same for the retail? Well, in the perfect world, that would be uh a great scenario. So, theoretically, we would use the ERP system to to supply the machine learning and AI algorithms to produce the forecast. Well, the reality looks a little bit different because in the real world, there is an ERP that has a massive amounts of data. There are sales data, some promotional data, and other sorts of data. And uh when we start working on this on the on the on the forecast, it turns out that the data resides in different uh different sources, in email, Excel,

Word files, uh messengers. So, it's not structured data, and it doesn't hold value. If we use poor quality data, it produces poor quality result. So, the technologies of artificial intelligence and machine learning are becoming increasingly more available on the market, and they start producing very very good results. We just need to learn how to work with them. So, this creates another problem, uh which is the data problem. So, a small or medium-size retailer now is much closer to Google and Facebook than ever before with the access to these technologies. So, based on everything I've said, a perfect solution to this, the solution that we've developed, uh would look as following as follows. So, there's a promo strategy

followed by process of planning, preparation, execution, and access. The uh data, first we need to aggregate the data, to create the data, start gathering the data. And uh once that is done, we need to hold this data for all of the participants for uh every stage of the process to uh to be accessible in a single solution. So, we have the data problem, we start working on it by structuring the data. We have process problem uh and we solve that in a single point of contact. Um this will put a company in a position where planning is well coordinated with marketing, with logistics. So, when the planning is done, uh the system would use the data to generate the forecast that will source the um

the mechanics, algorithms, the uh uh negotiation conditions uh that will then translate into the preparation, the logistics, so everything as we know so far. During that process, if the company uses a single application, it can analyze the performance. Right? Because we have all of the data in a single uh in the single solution. And once the trade promotion is completed, we can then go back and analyze the plan and the fact and the efficiency and calculate the compensation. So, this is a perfect scenario. It is um very few companies have actually learned on their on on their own to uh to create this process. And the uh and um the formula that we have developed as a result to solving these requests from retail companies is that first, we start

with a quality data and we provide the algorithm on how to structure the data, what type of data is needed, and how to aggregate that this data. Because having having high-quality data does not does not solve the problem immediately. The models need to learn. Even if you have the um the data already gathered, you still need 6 months for the models to learn, and at least 12 months to go through your annual cycle and analyze um all of the trade promotions that are happening. The second piece is efficient uh AI and machine learning models that are then powered by a single single source of trade promotion process management. So, basically, end-to-end uh end-to-end process management. And our answer to uh this big big

challenge that increasing the amount of retail companies are facing is the uh our our promo management tool. Uh it Uh of course, the theory that I have just shared is very good, but in real life um if you have the promotion calendar available to all of the participants uh at any time to analyze the historic data, to prepare for the upcoming uh upcoming um promotions, you're able to have a very good coordination. Uh and for this process to be efficient, multi-level, multi-step process um should be powered by an internal uh internal flow, which ensures that each and every participant completes the task at at the point when it's needed. This will close the the need for an efficient process coordination between

departments. And uh having this data available to analyze on all of the previous promotions, the AI model can then predict all of the following promotions and um shorten the uh time you spend on forecast generation. And actually have a predictable result. So, you don't have to rely on the models that are built by planner who is not uh not sure whether it's right or not. Let the system do the the uh the hard job of the calculation. Um the good news is that the more trade promotions you have, the the more the models can evolve and increase the efficiency. When you work with machine learning, you're able to constantly improve the efficiency by testing theories. So, if you add more

data, you can get different results, and you can do this on the go. That is a huge advantage. So, within the solution, you get the possibility to have the continuous improvement process available. Um and uh this is another overview of the all of the promo campaigns in the list available. No spreadsheets, no words, uh no messengers, everything in a single source. So, what type of data is needed? What we see is that more than 80% of retail companies do not have the data ready for AI. Even though everyone is talking about AI, it's everywhere everywhere, it's going to solve every possible problem. Uh AI cannot do anything without the data. So, we urge and suggest start structuring and gathering the data as soon as possible. If you

start today, you will probably have the data available within the year. Once you have enough historic data for algorithms to analyze, you can start getting first results. So, if you start today most likely and you have no data at all, most likely you'll be able to uh have value from AI models within a year and a half or 2 years. So, the sooner you start, the better. And the Here's a quick example for what kind type of data is required. Uh promo type, duration, promo mechanics, location, additional layout, a lot of data. Uh for those who are interested to uh learn more about uh what information is needed, feel free to take a picture, download the sample list, and uh compare it against what you have right now. How far are you from uh using the models? Key takeaways.

Promo will continue to grow for every for the majority of retail companies. It's It's inevitable. Second, if you don't yet have efficient promo process, um and you're looking to improve it, first analyze if analyze if you have efficient replenishment process. It's better to start from the beginning and then move on to more complex things. Uh the third one, there's no silver bullet. There's no pill that you can buy that will serve solve the problem. You can only do this in in in a gradual uh methodical way and over time get to a more competitive position on the market by leveraging uh these technology. The model for successful and efficient promo is good data uh that is processed by AI and well-coordinated in end-to-end process management. Start

preparing the data in advance. And use continuous improvement to um to increase the competitiveness of your business. Um thank you very much. I'd like to say that we were invited to the conference uh uh by our partners who are offering who are also a solution provider here in the Croatia and in the in the region. And if you have any questions, you can either talk to our partners who are here to uh support uh or you can uh talk directly to us. Uh we're happy to answer your questions. Um thank you very much. Thank you.