00:00:00 Introduction
00:01:04 SMCP’s supply chain context
00:02:50 Replenishment before Lokad
00:05:11 Why the problem requires a quantitative approach
00:07:18 Building trust in Lokad’s recommendations
00:08:50 How operational routines changed
00:10:37 Keeping optimization robust beyond the go-live
00:13:26 Expanding and improving the solution
00:16:25 Results after one year in production
00:19:40 Short-term and long-term impact
00:22:57 Advice for similar projects
00:25:20 Q&A: Demand and supply planning
00:26:51 Q&A: E-commerce and omnichannel stock
00:29:12 Conclusion
Summary
Carole Thomazeau (Sandro), Ines Muguruza (SMCP), Joannes Vermorel (Lokad), and host Tristan Oualid discuss SMCP’s transition from manual replenishment rules to daily, automated stock decisions. They examine team adoption, stock rebalancing, e-commerce allocation, measurable results, and continuous improvement beyond the go-live. A practical look at building trust in algorithmic recommendations and improving omnichannel stock positioning at scale.
Full Transcript
Tristan Oualid: Welcome everyone. So I’m Tristan Oualid. I’m lead scientist at Lokad and it’s my pleasure to today be moderating this round table. So with us today we have Carole Thomazeau who is director of business planning and supply at Sandro. We also have Ines Muguruza who is supply chain and transformation director at SMCP. And we have Joannes Vermorel founder and CEO of Lokad. So this round table will be about the collaboration between Lokad and SMCP around automated replenishment and stock rebalancing for SMCP’s omnichannel operations. So this project has been live in production for more than one year which gives us enough perspective to discuss about what has changed and how the replenishment is organized today. I will ask a few questions to our panel and then we can take also question from the audience at the end. So maybe let’s start by setting the context. So in could you briefly introduce yourself and tell us a bit more about the supply at SMCP?
Ines Muguruza: Hello everyone. So as Tristan said I am SMCP’s Supply Chain and transformation director and before diving into our project I would like to say a few words about SMCP. We are a fashion brand we are a fashion group with four brands: Sandro, Maje, Claudie Pierlot and Fursac. We are organized around three regions, Europe, Asia and America. Each region has its own logistics platform, their own teams for buying and allocation and we have around 1,200 stores. This project started within a global initiative to reduce inventory. We started by reducing the quantities we buy. So when you buy less allocation becomes crucial because every piece you buy has to be at the right place. So we realized that we needed some help to improve our allocation and we met local and since then we have been working together starting with Sandro Europe more than 400 points of sale collection around 850 to 900 references and you know multiplied by the sizes a lot of SKUs and that’s the context where we started and maybe Carole then will tell us more about it.
Tristan Oualid: Yeah, great. We have the context. So maybe if we go back at the beginning Carole can you explain us how you managed replenishment before Lokad and what triggered the need to have a new solution?
Carole Thomazeau: Yes, of course. So before replenishment was based on stock coverage with a target stock level based on past sales and not at all with the forecast for the coming weeks. The system remains static unless users pushed manually some change in the system or anticipated some future events and the relevance of the action relied strongly on the human perception and manual adjustment. Second point all the stock rebalancing was very time-consuming and done only at the end of the season with big quantities. Shortage management because we are buying less and less as Ines explained was done only once or twice a week and with limited criteria and all our bigger stores were protected while I think we were starving the smaller ones because the smaller ones with the smallest assortment have less opportunity to shift the demand to other products. And last point we book quite important inventory for our e-commerce at the safety buffer even you see from store is alive and this doesn’t allow us to sell in full price in brick and mortar in summary I think we do the right secure decision but not the most relevant because it was impossible to take to take into account all the criteria to have the good decision because human intelligence cannot do it alone and with the buying decrease volume shortage management and stock rebalancing became critical and we needed to automate them and to do it more frequently.
Tristan Oualid: Yeah, I think it’s very interesting and that’s something that we see often in retail where you are at the same time very strong business expertise from the teams but thousands upon thousands of decisions to make and which calls for an automated solution. So Joannes in simple terms why is this type of problem particularly relevant for a quantitative approach?
Joannes Vermorel: So SMCP is a large company and they have decades of experience but only so many people to deal with what amounts to about a million decisions or more a day. You know even if you decide to do nothing about a product store somewhere it’s already a decision. So we are talking about like a million decisions a day that’s you know the order of magnitude and if SMCP could afford to have one person spending two hours on every single case this would be beautiful but the reality is you’re not anywhere close and the approach of Lokad is to create numerical recipes, a modelization but that really embraces the vision the understanding and the collective experience of the clients so that it’s not fundamentally like a system that behaves like a black box doing weird you know voodoo in the background but something that is like the purest expression of the expertise and what people would do manually if they were granted unlimited amount of time which means to compose a numerical recipe that really mimics in a way the way the people think strategically about their brand and that’s the work of the supply chain scientists that we have at Lokad and that’s where we are quite different is that this is not an out-of-the-box AI. It’s something that is built together with a client.
Tristan Oualid: Very interesting. So I think the relevance of the quantitative approach is clear but as you say you also have to embark the teams to learn the solution. So in your case Carole, with Sandro how did you manage to progressively gain trust in Lokad recommendations? Carole Thomazeau: First of all, the project phase was very important and the fact that the Lokad data scientists also had a supply chain expertise made things and the discussion highly relevant and very complementary with our business expertise. It makes the opportunity to challenge the way we were working and we had a strong partnership and a strong support from the dedicated team. During the go-live phase, the teams learned to trust the tool because of major validation and testing phases where they understood that the way they are managing the allocation can be optimized. And we had a lot of KPIs and reports that explain the trade-offs of the tool. Everything is explainable and we can evaluate the decision of the allocation and third point we are doing weekly meetings because it’s always continuous discussion and improvement and it’s very collaborative and the fact that all our issues are taken into account straight away and solved very quickly make the trust bigger and bigger.
Tristan Oualid: Great. And so now that you are in production so we talked before the go-live if we if we look at how you are now using the solution what are the main, I would say, modifications in terms of routines in terms of daily management of all the replenishment decisions?
Carole Thomazeau: But the role of the teams has changed because they are now more focused on consistent checks and analysis rather than execution itself. They have time to look for improvement, new ideas and not just put in masses of manual inputs. We gained a lot of productivity. All the automatic operational reports are live every morning. So we have straight away all the answers to the questions we have from retail and it’s very easy. For stock rebalancing, the time for stock rebalancing has been divided by three and they are smaller and more frequent and this makes it easier for retail to execute and we have a better adherence to this request. Shortage management is done every week, every night and is automated and taking into account many criteria and is powerful and what else? Yes, last thing they are monitoring now a new KPI that is stockout rate that allows us to measure the impact of the actions, the quality of stock and to understand the performance.
Tristan Oualid: Okay. So we see that the solution like the go live of the solution is not the not the end. It’s really only the beginning. And so you keep steering it and adjusting it while the business context is also modified. So Joannes what keeps an optimization solution robust and useful in the long term beyond the go-live?
Joannes Vermorel: The technical term is called but it’s a little bit technical. So the bottom line is a numerical recipe is just that it optimizes a numerical target that is fundamentally static unless you have software that will write itself which is not a good idea, you know, not automatically. Don’t let an agent do that for you yet. So fundamentally you’re shooting at a target but the way this target thinks about your business is static and that’s exactly what the teams are doing when they say we do reviews we want to have continuous improvement it’s about how do we rethink what our business should be optimizing in the first place and at Lokad the idea is that what we have noticed more than a decade ago is that supply chains are always in flux you know the geopolitical landscape is shifting. You have especially in fashion even if we don’t put the macro concerns fashion itself is an ever-changing target. So you need to have people who just spend time to do high-level thinking and the key is those people need to be free of the mundane daily drudgery and execution otherwise nobody has even the time to think about that. So free the people from the daily drudgery then they can think look at KPIs that are on their own terms that’s also an important aspect of Lokad is that when we say we have like KPIs, dashboards, numbers, figures, etc. Those are typically numbers that have been vetted by the teams so that what Lokad measures is we measure things in the way the client company wants the things to be measured. And that’s important because otherwise it becomes like very much blackboxish because people don’t even understand what is being measured. And then all of that is reviewed continuously and the supply chain scientist is there to just keep delivering the updates and keep the complexity and maintainability of the solution under control at all times.
Tristan Oualid: Okay. So it’s really all about continuous improvement. So if we if we come back to the case of SMCP, Ines, how did the collaboration between Lokad and SMCP evolve after this go live having the solution in production and maybe if you if you could give the audience concrete examples of improvements or adjustments that you have made while the solution was running in production.
Ines Muguruza: Yes. The solution was first developed for Sandro and then we followed with Maje and Fursac this morning. We’re working on launching the same solution for the American platform in a few weeks and now we’re working with Claudie Pierlot. One thing I would like to say is that our first project was around eight months something like that and now since Maje, we have shortened by half. So this means that the solution even though it was custom-made for us we worked from scratch together. We have been able to reproduce and adapt to the other brands in a much faster way. That’s one very important point. The other thing is as you say we are in fashion. It’s a very lively industry and sometimes our collections or products perform well other times we miss something. So they are less fast selling and we have this kind of discussion with them which is at the strategic level saying okay last season we were in a shortage situation which is a good one. It means we’re selling very well. This one maybe it’s slower. So how can we work this strategy together to maybe push more or things like that. So we work together your team works with them on this kind of discussion which is a strategic discussion and we also work together on improvements on the platform itself like developing a new dashboard to explain something we were not able to explain before. We have adjusted the rebalancing strategy and we have also dealt with our Canada stock which is mutualized between Europe and Canada but two different business units. So we have been adding let’s say use cases and situations that we face that we hadn’t anticipated. So that’s the continuous improvement loop and continuous collaboration that we still have.
Tristan Oualid: Thank you. Very interesting. So we have talked about how the solution works how it keeps evolving after the go-live. Now if we focus on impact and like the consequences of having Lokad as an automated replenishment solution so after one year of production do you have any KPIs or results that you can share with us today about the impact of having Lokad?
Ines Muguruza: Yes, a lot. Maybe first I would like to talk about before talking about KPIs I would like and I will let Carole explain we have a different relationship with retail and with e-commerce. Well if you are in the in our world you know that retail is always complaining about not having the right stock and allocation always being the one not doing the job. So I think for your team it has been easier to talk to them. Maybe you can tell us a little bit more. Yes, I have a real example. The retail teams used to have some push orders on top of automatic replenishment just in case. It was an extra precaution of stock. And with Lokad we were able to measure the effectiveness of these pushes where the stock has been sold and where it is sleeping in store and at the end only 40% of the pushes were sold after five weeks. So we can show this to the boutiques, to the stores, and explain to them that it was less effective than expected. And they understood that Lokad can make better allocations because it has a global demand view of the inventory and that local adjustments blur the scenario. So it was very easy to convince them to stop and they stopped straight away and now they trust the tool and they are very convinced by the right allocation. So easier that I was thinking for me that’s the best result we have trust and dialogue and to talk more about maybe more precise KPIs that maybe you’re curious about for Sandro Men we reduced by 25% the stock that was reserved for the e-commerce and we didn’t observe any additional shortage this made this stock available for brick and mortar in the early phases of the season. What else? You were able to reduce like the shortage by three points of Sandro Women. Another very good example I think it’s for Maje, we have observed growth in revenue in smaller stores. These stores as we said at the beginning our team was not able to look at every store every day. So naturally we privileged the big ones the big stores with a lot of sales and nobody was looking at these small stores and I think that is a very relevant result to share as well.
Tristan Oualid: Okay, very good. And so maybe on a more global view, Joannes, what is the kind of impact that company can expect after having Lokad as a solution and maybe is there a difference in the short term and in the long term? I’m talking about maybe more than two years of production and so on.
Joannes Vermorel: The key impact is to transform the time spent by people on something that is like expendable to something where you capitalize where it’s accretive. You see how do you treat your employees? Do you spend one hour of an employee’s time as something that is consumable? You know, they are going to work one hour and then it’s gone. It’s gone. It’s only what it took for the supply chain for the retail network to not stop today. So you had to pay this one hour and then tomorrow you will have to repay. Nothing lasting, nothing accretive, no capitalization. So the transformation that Lokad wants to convey with our clients is to shift the human contribution towards something that is extremely capitalistic where people refine their expertise, they refine their understanding. And so imagine you know the long-term business is that all the people who have expertise develop their expertise refine the understanding and help in turn Lokad to adjust the numerical recipes with all of that being automated. So that’s the key because I believe that all the other metrics come from that you know down the way we can improve the stock but is it reducing the stock it can be good except that stock helps selling. So if a store that is empty will not sell. So fundamentally all the metrics are only valid up to a point and it requires high level human understanding and that’s what we want to cultivate. So the immediate effect of Lokad is to improve a series of short-term KPIs that were usually identified even before us such as for example small stores being starved okay improved but the long-term benefit is to have a much better more strategic understanding of what supply chain means for the for the brand for the retail network with a you know a very specific understanding not like a generic any companies you average of the market something that is purely specific to SMCP and that would be the same for any other clients that we would serve.
Tristan Oualid: Okay. Thank you. And to conclude just before we go back to the to the audience and potential question with the benefit of the experience that you have now and maybe for you Carole who implemented this project at the start if a peer were to ask you for a similar project some advice what would you tell them?
Carole Thomazeau: For me it was really a game changer and I will push for it. For all the reasons we said this before but I will sum them up. One it was quite a quick and easy project. And the design was customized for SMCP and able to improve week after week. Second point it was not so obvious at the beginning but the teams are very convinced by the tool and they don’t want to step back because as said their activities have changed with much higher value activities and for us it’s a new job and much more interesting I think what is very important it’s we are not losing the control and it was a fear at the beginning of the project is not a black box everything is explainable measurable and we can adjust it so it’s very important and I think, but it was not in our mandatory plan at the beginning of the project but it’s true as all the KPIs and dashboard are very easy to understand. We can explain to non-expert departments how we allocate and why we took this decision and this develops trust and team spirit in the company and we are not fighting anymore with retail or digital and it changed all the things.
Tristan Oualid: Thank you. So just before going back to the to the audience for potential question so I’d like to thank Carole, Ines and Joannes for the shared experience. So we’ll take now questions. So don’t hesitate to give me a sign and potentially if the question is dedicated to a specific person in the panel to indicate it.
Audience Member: My question is for SMCP women here. I was wondering how you switched from supply planner to demand planner more and what is the percentage of time you spend on demand? And let’s say supply and inventory planning. I don’t know if it’s clear enough. Ines Muguruza: We in this particular project since we’re looking at store allocation, we don’t do demand planning as we did before we don’t do forecasts. Lokad is taking care of that and doing it pretty well. So no we didn’t change the roles Carole’s team is still dedicated to allocation but they are helped by this tool and as I said they are taking a higher view of things and developing a more strategic way of doing allocation. We still have demand planning but not at this level. We still have demand planning for replenishing our warehouses, taking decisions on rebuys in the season and things like that. We are talking with Lokad about these subjects. They are not covered in our project yet. For the time being, Lokad is really dedicated to store allocation and store rebalancing. But maybe nice things to come in the future. We’ll see.
Margarita (Amazon Supply Chain Services): Hello, Margarita from Amazon Supply Chain Services. Thanks a lot. Very interesting, very clear about the allocation piece and the improvement that you made. I was curious apart from the allocation piece both for store replenishment but also for the e-commerce piece if this has also an impact on the way you are able to have stock to deliver to customers both through stores and also through the D2C channels and if in this area there are any other let’s say challenges that you are currently facing that you would like to face in the near future. Thanks a lot.
Ines Muguruza: Okay. Our stock is mutualized. We only have one stock for all channels. If we talk about Europe, we have one stock for all channels for all the countries. So this was already the way we worked. Before Lokad we used to reserve stock for the e-commerce in a quite, let’s say, cautious way. A big portion of stock was reserved for e-commerce from the beginning of the season to make sure they were going to do their sales and their targets. And this was maybe one of the biggest changes we made because now the e-commerce is treated as any other store. So we only reserve stock for them for the weeks to come and not for the whole season. So the calculation of stock needs for e-commerce is done every night in the same way as we do for stores and well I don’t know about your companies but in ours e-commerce sales are slower than stores. We tend to sell a full new collection at full price earlier in the stores and then when the end-of-season promotions come the curve for e-commerce ramps up very fast and this is taken into account by Lokad. It was one of the things that was harder to adjust to detect fast enough this growth in e-commerce, but it has completely changed the way we allocate stock to this channel. I hope I answered your question.
Tristan Oualid: Okay. So, if there’s no other question, I think we almost reached the speaking time. So that’s perfect. So again I think we can all applaud Carole, Ines and Joannes for their intervention. Thank you everyone.
