 Hello, everyone. Welcome to our second Spring's Live series. Thank you for joining us today. I am Paulo Sousa Jr., course lead for SC3x, Supply Chain Dynamics, which is part of the MITx MicroMasters and Supply Chain Management program from MIT. I am happy to be co-hosting this live event today with my colleague Miguel Rodriguez Garcia, course lead for SC1x, Supply Chain Fundamentals. Hello, Miguel. Hi, Paulo. Hi, everyone. Thank you for the introduction. I'm really happy to be here with you and with our guest speaker and excited to bring the industry perspective to our MicroMasters, which is what we do in these live events. So today, we will talk about inventory management. And to do that, we are going to follow this same agenda as we did in the first live event. So first, our guest speaker will give us a presentation that will last around 25 minutes. And after that, we'll have some time at the end to answer some questions from you guys, from everyone in the audience. So that will last probably around 15 minutes. And we encourage you to use the Q&A feature in Zoom. Please try to avoid the chat, because it's really hard for us to follow the questions. So you have the Q&A feature in Zoom for that. And Paulo and I will be channeling all those questions at the end to our guest speaker. And just before going back to Paulo and to our guest speaker, I just want to mention that this event is part of the MITx MicroMasters Program Supply Chain Management, a program that we developed here at the Center for Transportation and Logistics in MIT. And as well as supply chain fundamentals and supply chain dynamics, which are both courses host in this event, the MicroMasters program includes other courses, a total of five. And some of them are currently open by now. So don't hesitate to check them out. We'll be posting the link for our program in the chat in case you guys are interested in any of the other courses that we have. So now back to you, Paulo, so you can introduce our guest speaker. Thank you so much, Miguel. So today we are honored to have Rafael Anunis joining us. Rafael is the Technology Innovation and Data Strategy Director at CMA CGM. Before joining the company, she worked for more than 10 years in different supply chain roles and consumer goods. And as a consultant, supporting customers and deploying digital solutions. She holds a bachelor's degree in naval engineering from the University of Sao Paulo. And a master's degree in supply chain management from MIT. By the way, Rafael was part of the first MIT supply chain management blended cohort. Before being admitted to the blended cohorts, she successfully completed all courses from the MicroMasters program. And as some of you may already know, one among many other benefits from earning the MicroMasters program credential is that you become eligible to apply to the supply chain management blended masters program at MIT and also to other universities around the world. All right, so welcome, Rafael. The floor is yours. Thank you so much, Paulo. And thank you, Miguel as well. Thanks everybody for being here. It's super excited to be here with like the this big cohort. I've seen a lot of people chatting from different parts of the world. I think this is really like the energy that we had in the MicroMasters and also at MIT. So again, excited to be here and be able to share a little bit of knowledge with the group. Let me get started. Share my screen with you guys there with me. And Paulo and Miguel, you let me know if everybody can see it well. Yes, I can see your screen. Yeah, it works perfectly. Awesome. Yeah. Thanks again for the introduction, Paulo, just again a little bit about myself and just some some things to to find myself or aid myself. I don't know what is the what is the word as Paulo mentioned? I was part of the first cohort in the supply chain master's blended program. So I completed my MicroMasters in 2017. And I apply for the blended program at MIT. A little bit of my background, again, as Paulo mentioned, I had a career in supply chain for consumer goods and for retail, worked as an independent consultant. I'm originally from Brazil, came for came to the US for the MicroMasters and then for the masters, actually. And then I've been in CMA CGM, a container shipping company leading technology and innovation since four years ago. I'm based in the US and the case I'm going to present is back for my consultant times and again, excited to to share the knowledge with you guys. So moving to the title. So I'm sharing about inventory management and how to align the company strategies with the right policies. Again, this is just a case that that I worked when I was a consultant back in Brazil. I think before we get started, I would love to know a little bit about you guys and your company's supply chain. And Lisa, if you can allow to the poll, I would love to know. What is your company's supply chain strategy? So I have three options here. Is your company following the lowest cost possible for a given service level? Or the highest service possible for a given cost? What is restricting your company's strategy? And you can, of course, be not sure. I think sometimes it's a little bit hard to navigate big corporations and like the changes and strategies. So I'd love to hear from you guys. And Paula and Lisa, let me know once it's. Yeah, they're familiar with like the timing for pulling. Let me know when it's a good time for us to close. Yeah, they're still voting. OK, so by now we have almost one hundred and fifty. So I think we're good to go, Lisa. And here are the results. Can you see the results? Yes, I can see the results. Yeah, it seems so it seems like the majority of the folks, it's almost a tide, but the majority of the folks have the highest service possible for a given cost. It seems like cost is really restricting supply chain from companies. I think this is interesting. But a lot of a lot of you guys are also saying that you have like the lowest possible possible, but it's a service driven supply chain, which is pretty interesting. And again, for those that are not sure, I do encourage you guys after this presentation and throughout this they are weak to think about that. And then in case you're not sure, try to find out what is really driven your driving your company's supply chain for you to be aligned with your country policies, with your distribution policies or any any planning that you you might do in the future. Thanks for the poll, guys, and thanks for the engagement. Let's get started. So just to give you a little bit of context. So again, I was a consultant leading the supply chain practice in a boutique consulting company. And we have this customer that was a cash and carry wholesale. They have around 30 stories in low income and very sensitive to price areas around 10,000 sqs and two major distribution centers in the metropolitan area of Sao Paulo, which is a big city. Just to give a little bit of context, I know there are folks from different countries, different geographies here. I'm not sure if all the geographies will have similar chains, but we're talking about a place where people go to either buying book or get big discounted items for their small business. So it wouldn't be uncommon to see folks that are on a restaurant to go to the chain and also families that are seeking to save money. So for those that thought about the supply chain strategy and eventually the inventory strategy of your company, think about a chain like that that is very, very driven for price, where their customers are very, very driven for price. What should be the strategy? So let's move on. A little bit about the problem here. So this call was we were helping the replenishment team that was placing the orders for the distribution centers to add inventory in the stores. And the recommendations could also the procurement team in that chain to place the order for the suppliers. And those suppliers, those vendors, they would deliver products in the distribution centers that would then deliver to stores. And those suppliers could also deliver straight to stores. So there are two major teams that could influence the inventory. So we started by one small scope here, but also influencing what the other team would do. So what were you trying to solve for this customer? They had a very high stockout perception for discounted items. So for those that are familiar with retail chains, if you go to the store and then you don't have a product, you have like your empty shelves, customers not happy and like they store managers are not happy. Some of the key items in the shelves were not there. So there's, and when I say perception, it's because they didn't have a KPI for that, which is again, it might be common in some places. Some places are, of course, way more mature and there are plenty of KPI to monitor stockouts, but just something to think about. The other thing is that was a very low adherence in the replacement solution. So there was an ERP that this replacement team here could actually place the orders, but they were not necessarily like following because a lot of times they didn't have the right products to actually send to the stores, right? Because this team here eventually didn't have this procured. And there was no fixed procurement portfolio, this is extremely important. This team here, like the procurement team, they were extremely opportunistic and they will buy items according to the price. So if they would see an opportunity, they would go there and buy like in bulk that specific item for that specific week or month. And for the customers, as I say, the customers were very, very sensitive to price, right? The customers that would visit the store, they wouldn't necessarily be very loyal to brands. That was not the point. If you would be loyal to brands, you'd go to another chain that was all about pricing. So what we did, and that's what I'm gonna go through in the presentation, we worked on a solution that was like segmenting the portfolio in kind of main two different items, right? And we did different strategies for forecasting and replacement management for those two different category of items, right? That's kind of what the solution was about. The way we did it, it was through data analytics and analytics of stakeholders, the way I'm calling here, really like going and understanding what were the dynamic of those teams here? How did they work? What was the thing that was important? What was the thing that were important for day store managers here, right? When I say analytics here of stakeholders, I know it's not necessarily something usual, but just something to have in mind. And then we also worked on different, again, different modeling for forecast and replenishment for those two different categories. And in that case, we deliver something simple, an MVP tool that a company could use and pilot the strategies that we were proposing, right? So it was important to actually like connect them to end with some action, things that they could actually like implement. So let me just walk you through how it worked, right? This whole thing. So a few insights that we had when we were like in this problem discovering, talking to people, but also like going deeper into the data that we had. So when we went through this data, we got a few insights, right? So we analyzed the seasonality, daily seasonality, weekly size of seasonality. We did some visual analysis. We did some indexing, testings. Sometimes we needed to adjust certain things to the stockouts because there were a lot of stockouts in some items. We really understood the procurement strategy that was not necessarily something documented, but again, interviewing stakeholders. And we also did some testings in running some regressions for the discounted items, what would be the increase in sales for those discounted items, right? I think some insights that were relevant for this analysis were, first, we decided to split the SQ segmentation into two different categories. And that was really key for this. We split those items in what we called replaceable and non-replaceable. So this first assumption that there was no brand loyalty at all was a little bit misleading. There were some brands and some key items that customers would really go to that place and look for specifically if they were discounted, right? So there were just a few, but let's say a milk for that specific brand was really in customer would go there for that. Tomato sauce for some specific brand customers would go there for that. Everything else could be replaceable, like the brands that were not necessarily like with high loyalty, it would be more price-driven. And those non-replaceable items, again, higher brand loyalty and sensitivity to promotions is specifically if there were discounted, people would go to these stores for that. Another thing that we found about like the weekends, that this seasonality is that the weekdays would be like relatively the same, but in the weekends, we had like a very high seasonality, right? So people would really let go there by in the weekend. Yeah, and just a few insights when we kind of start digging into a problem and what are the things that help us with that? So what we decided to do in terms of like modeling for this scenario after understanding this, the complexity. So the first thing is the forecasting, right? We do not start a replenishment strategy without a forecasting. So we split our items in those two categories. So for the non-replaceable item, those that couldn't necessarily be missing in the store and they were very sensitive to price, we had weekly forecasting by SQ in store, so very specific. It did include the promotion effect, the increasing factor from those regression analysis, what would change for this demand, for the demand of this product if I discounted the next amount of like currency, local currency. So for the replacement, the replaceable items, this was really like key. Traditionally, companies do forecasting and replenishment like SQ by SQ, but we decided to go by family, specifically because there was no brand loyalty, right? And there was no promotion in those items so there was no need to worry about that. So instead of forecasting a rise for that specific brand, that specific size or that specific SQ, we would do for the whole category, right? And we needed to standardize some units, but this whole idea gave us something that was extremely important, which is less variability in the demand. If you guys are following the course as well, if you have less variability in the demand, you'd need less safety stock. Again, that was key because we could lower the safety stock for the same service level, therefore lowering the cost with less demand variability. So again, we're forecasting all those replacement items just by the category. And again, advantage is here, less complexity, instead of having to forecast like a thousand items, we would forecast like 200 items. And the lower variability, again, the pooling effect, if I'm forecasting for everything separately versus like the whole thing, one variability of the night and that kind of lights the other one. So something important here. In terms of the replenishment strategy for those two different categories, one would be really like weekly by SQ, very high level target service level and alerts for stockouts because those products couldn't be missing. For the replaceable items, we would have a weekly, but by family. And I would just need to have in a particular store, one or two SQs from that specific family, right? And the leverages stock and variety across these stores, we could have kind of the same, I didn't necessarily need to have the same SQ in the same store, right? And in terms of stock visibility, I did have like, again, a much higher control. And again, why it was important to remember when we talked about what was really important for those stakeholders, there was a very high perception of like products missing, but the key products missing, not all the products missing, right? So this one we really needed to kind of analyze a little bit closer, right? And we would give therefore a focus in the non-replaceable items and we will avoid where we're calling fake stockouts, which is like a product that is missing in the plan, but the reality is for the consumer doesn't quite matter because they could replace for other items. So we would focus on something that is really like non-replaceable and that the customers would go there for buying that specific brand. In terms of like the way we implemented this, we specifically did a pilot with a tool. And again, it was a simple Excel-based tool like connected to an SQL server and giving some simple analytics in that specifically on the non-replaceable items. So we did pilot with one category in dry goods. And we have like the new segmentation visibility. Non-replaceable items would be on the SQ level and the replacement items would be in the category level, right? So the stockout would be a stockout alert for the SQ in certain items and for the family in other items. High stock levels the same, right? I would not, the point was not being one SQ by SQ for items that were easily replaceable by the consumers. In terms of what it brought us results for this pilot, right? And again, I'm presenting here a work that we've done like towards six months that it gave this customer as a small tool for one specific category of products in their store. I think the main point here and back to our poll at the beginning is that we had the inventory policies aligned with the company strategy. So again, what was the company strategy to really, really lower the cost of the products because they wanted to serve that consumer that was very price oriented but to have that specific products that the customers would go they're looking for when especially when they were discounted available for the customers at that specific store, right? So in terms of results, it of course, promoted the higher service level on the discounted items that would be looked a little bit differently and a better visibility about the real problem which was like the non-replaceable items stock out. So the company had like some fires of discounts and if that specific item was in discount and that specific week with like this higher brand loyalty that is really like what the problem of stock out was. And then again, the inventory and cost reduction here I needed the lower safe to stock level if I have less variability. Again, that pull effect that we started to forecast those products by category. And there was something here that enabled the teens to follow their current strategy because the procurement of that company that was like placing the orders for the vendors they were being very opportunistic. Sometimes they didn't necessarily need to buy that SQ for that specific brand. Let's say for milk, if you have like some vendors that could provide something for the lowest cost if there was in that category that the customers are not very loyal to the brand you could just go there and buy. And then if the price was not good and that specific month or week of cycle you wouldn't buy them and that was okay because you would have like all their items that could replace. So it enables the procurement to, they were already executing this but really like it was an overall strategy of inventory management that enables the procurement to execute that opportunistic approach in negotiating with vendors saying, okay this is what I can pay and then getting like this lowest price. And again, aligned with everybody else instead of having the replenishment team thinking they should like send the inventory for all the items for the stores based on their previous ERP system or recommendation but now they could do this only in some specific SQ user like the for the brand. And again, the policies aligned with the change strategy, right? It was a low cost chain for sure and low brand loyalty for most of the products, right? So again, back to our question in the beginning I do encourage you guys to think about what is your company supply chain strategies? And then how can you align not only the inventory planning here but even like the day-to-day interactions and like any decisions that you do in your supply chain to have that in mind. I think it will help a lot in the day-to-day decisions and what are the items that needs to be on shelf? What are the items that eventually it is okay? And what is really like important for the customer for the brand and for the consumer? So I hope you enjoy and it was a little bit insightful for you to connect the course material with a real case example. Excited to get some questions. Thank you, Rapala. Yeah, no, for sure. At least as you want to learn us, I'm pretty sure that they are connecting the dots right away because the content is great. And right now we are like right in the middle of the course so they are gonna start getting more and more into inventory management. So I think this overview of that connection between demand forecasting and inventory management was really, really great. So yeah, I'm gonna start and Paolo if that's okay with you with some of the questions. And so first, this one is actually mine because I'm really interested in the demand forecasting topic and then I'm gonna go right into the learners' questions but I'm curious, how did you segment the SKU between replaceables and non-replaceables in terms of like the KPIs that you use? Like how did you create that split? Because I think that's a big deal for supermarkets like in terms of like how to decide, oh, this is a replaceable item and this is not. You know, what kind of information data did you guys use during the project to make that split? Yeah, that's a great question, Miguel. It's interesting when I say like the analytics of the stakeholders, right? So I think it is a combination between we going through all the items that were promoted or discounted in the past months, but then presenting it to the stakeholders and specifically the procurement team and then kind of suggesting some items and aligning this portfolio with them. So it was a combination of us like looking through the data but also working with the teams to see, to validate what we saw in terms of strategies and taking that human expertise. So we had like folks there in procurement that were doing this for like 20 years and really understanding what was the customer's preference and what worked and whatnot and selecting that portfolio that they needed to discount. So I would say it was a combination of us like looking through the data and seeing, okay, that seems to be very sensitive to pricing and it's much more sensitive than the other ones in the category versus talking to the procurement team and they store us in and then confirming, yeah, yeah, yeah. This one we cannot not have because customers come here looking for it. Really interesting. So price sensitivity was one of the key KPIs that's really, really. Okay, Paulo, you wanna go ahead? Sure. Thank you so much Afaila for a great presentation, sharing your insights. And it's interesting because SC3x learners, they are probably preparing for their metarium exam and they may recall the importance of aligning the supply chain strategy to the company's strategy. So great insights there. Thank you so much. And let me read a question from Parishak. It's related to non-replaceable products. So there's learning mentions that forecasting at SQ store level. The data is very sparse generally. How much forecast accuracy at SQ level was achieved? And if you can share which two and which forecasting model was used for that? Yeah, that's a great question. So again, this was a small pilot that we've done in that category of like a thousand of SQs. So the strategy to just go for this specific items in the store level enable us, and those are very like high volume items, right? So for those, we didn't have like as much sparse as if you would go to other items. I think the biggest challenge in terms of like data was to adjust for the stock outs. So to be honest, I don't remember for sure, but I know we did some adjustments in the stock out based on the, so we would compare the level of stocks versus the demand. And if there was a stock out, we would do some adjustments. That was done again in a sample and then we was too wise for that pilot was Excel plus then replicating in the database for like a large amount of items. And then to forecast those items, we had a simple forecasting model that again, it's a case, it's been a while. I think we started by having like some really simple moving average, but the key there would be a regression to adapt that volume to the discount that was applied. So maybe you guys can follow Miguel, you can help out, but there was like a model that would consider the past seasonalities of the weekend and the discount factor. Maybe some, I'll find a name. So, no, I mean, whole winters for sure can, I mean, take into account seasonality, if you consider your variables like the discounted price and all that, probably you need some kind of like more advanced model, I guess. But the basics for seasonality trend, all that I think the learners have seen them in with whole winters, yeah, but interesting. Okay, thank you. Thank you so much for the answer, Rafaela. So we have a lot of questions. I can tell the audience now that we are not gonna be able to answer them all so because people really enjoyed the presentation. So I'm gonna go straight into another question from Noel Lam. So this learner is actually asking us something really interesting related to the consolidation of SKUs. So the question is if an item is considered replaceable and you actually forecast and replenish by family, does that lead also to a consolidation in the procurement side? Because at the end of the day, the idea is, okay, now you have kind of SKUs that are exchangeable somehow. So did that also end up being a consolidation for example, suppliers in terms of reducing the number of suppliers to gain economies of scale? Like how that affected the procurement side? That's a great question. Yeah, I think the recommendation for procurement would be you have to buy X amount of items for that category. Do your best. And what was happening already was they were doing their best in buying only those where the price was, it was an opportunistic strategy. So they would have like meetings with vendors and suppliers all the time, but say if you don't achieve that specific price, I wouldn't buy for you. So you did enable them to say, to really execute the strategy that was already done, but align with the other policies, right? So we would say you need to buy, let's say X amount of units of that specific category and have at least two brands, two different types of SKUs to have a little bit of variety, right? So depending on like what they would find in terms of like lowest costs and more opportunistic procurement or better conditions of procurement, then they would buy it. So it seems like it definitely brought a lot of negotiation power to the procurement team. Like- It did, yeah, yeah. And again, I think it brought the negotiation power and enabled them to already use what they, because that was the way they operated anyways, right? But the stores and the other side were not necessarily aligned and they were looking for some items that were not in the procurement strategy anyways, because that was not, and the procurement team was really like the core of the company's strategy in terms of like really getting the lowest costs and enable the company to like grow. Interesting, interesting, thank you. Paulo, do you wanna go? Yeah, we have one more question here. So Steve is asking how to find out whether customers are loyal to a specific brand, how did you follow this approach in this case? Yeah, that's a great question. Some of the analysis that we've done and that was more like, again, there was a data analysis and then the stakeholder interviews and all that, but some of the analysis we've done that if one product was missing, that whole category would be down in those non-replaceable. For some products, if they were missing, life would go on, right? So in that specific category. So I think the point of like, you would see the movement of that whole category and see, okay, what was the excuse missing? Some of them they would really like lower the demand of the whole category. Some of them, as long as we had some stocks from others, it would be okay. So with that list and then the validation of like the stores and all the folks that would say, yeah, yeah, this is really something that makes sense, right? So it's kind of data analysis across validation. Thank you so much. We still have many, many questions here. Maybe we can take one more, Miguel. You want to take the next one? Yeah, maybe, I mean, and we have time, maybe like a couple more, like one me and one you until 9.45, okay, yeah. Because yeah, we have a lot of questions. So I think we should give the audience like as many chances as we can to ask Rafaela some great questions. So Michael Wagner is asking something really specific, but I think really interesting in terms of like SKU, forecasting and segmentation. So how did you deal with different units, packages, sizes, when you did the whole product? Like in terms of forecasting and those exchangeable items? Yeah, that's a great question. So we standardize all the units. So let's say if I was dealing with rice, I would do everything in grams or kilograms versus the unit itself, right? So, and again, a recommendation for the procurement would be like you have to have X amount of kilos of data specific item, right? And you could even, for instance, if you're dealing with a vendor and they have like a lot of like inventory laps for one specific size, that could be something that could be used opportunistically. And but one thing to point out is that chain didn't have a lot of variety in terms of like sizes. And so again, I wouldn't necessarily recommend it for everything, but in data specific context, what the customers would be like, you know, small business, they will look for bulk items. It was a little bit of a ready standardize. You didn't see like 10 different ways of selling the same product, right? It was really like already very standard packaging. The sizes would vary, but it wouldn't be like you can buy something very small and you can buy like owns of something. So. No, it makes total sense. I mean, as you mentioned throughout the whole presentation, this strategy aligns like the inventory management strategy has to align with the business strategy. And it wouldn't work for probably a discount store that sells like really, really small and personalized like sizes or in packages. So yeah, no, thank you. Thank you. So Paulo, do you want to take the last one maybe? Yeah, let me take one more. And this one is from Atyab. So how do technological advancements such as the AI, machine learning and blockchain impact inventory management? And how can companies leverage this technologies to optimize their inventory levels and increase efficiency? Yeah, thanks for the question. So again, this was a case that we, I mean, we use very sophisticated techniques, I would say, but not necessarily the use of AI and machine learning. The way I've seen, I'm gonna answer this like my past experience, like looking through the trends in the market and companies that are emerging. And so one, so what I've seen in that sense is a lot of emerging solutions in what people are calling enterprise AI. And specifically for inventory management is instead of going for a deterministic approach, as I say, we have like a forecast, even if you consolidate this, but it's very deterministic, right? You have a forecast and then you go, you have a strategy for replenishment, you have like weekly orders and so on. Today there are models, and again, there are companies bringing solutions in that space where you really take into consideration like tons, tons, tons of data. So you have like how many customers were in that store, at that specific time, where the product was positioned in the shelf and then what happened in the procurement, like if you can track some behaviors that were on in the procurement side as well and you kind of take this all into consideration and you can maybe use AI to optimize for service, to optimize for profitability, to look for what is the best model that you can actually go and like to simplify the recommendations, right? What it happens though is that the models are not necessarily super easy explainable. So, but then what I've seen working is really like companies that could measure the results satisfactory, right? So are you using AI to go for the highest profitability than you follow up your profitability or to go for the highest service level and then you'll follow up and you see that model is actually helping you versus what do you have in terms of the deterministic approach? So it's just kind of a different way of like you interpreting the models and the results. It's not necessarily super interpretable for planners in general, it's a, I would say it's a big change management, but like especially for newcomers, I think that is a lot of advantages in using like more sophisticated tools like that. Again, I've seen in this enterprise AI market, a lot of emerging companies doing that that is again not a deterministic approach as I described, but really like taking into consideration tons of factors and forecasting, optimizing and planning for certain key KPIs. That's great insights. Thank you. Thank you so much for bringing this final like overview of what I don't know, the future could look like for inventory management consider all the new technologies that are coming. So I think with this, it's time to wrap it up. I mean, we could spend here the whole day probably because we still have many questions from our learners, but we wanna be respectful with everyone's time. So thank you so much, Rafaela. Thank you so much to everyone who decided to join us today. It's been a super insightful session, I think. And just before we say goodbye, I just want to remind the audience a couple of things. So first, I mentioned at the beginning that we still have other courses that open at this point for enrollment. So we encourage everyone to check them out. We posted the link in the chat earlier. And second one, that there will be another life event in this spring series, probably in a few weeks. So stay tuned because we'll let you know soon. And again, thank you so much, Rafaela. Thank you so much for joining us today. And just have a great week, everyone. Thank you so much, goodbye. Thank you guys, bye.