00:00:00 Introduction
00:03:16 Why Black Friday promotions are risky
00:08:11 Should retailers follow competitors’ discounts?
00:12:50 Measuring lost full-price sales
00:19:29 Introducing the Dublin retail persona
00:21:32 Rate of return and probabilistic forecasting
00:29:33 Choosing which products to discount
00:34:37 Basket economics and retail robotization
00:39:13 Omnichannel allocation and stock rebalancing
00:44:04 Black Friday versus Christmas
00:45:06 Black Friday as business as usual
00:48:53 The aperiodic rate of return
00:57:03 Should Dublin participate in Black Friday?
01:02:16 Goodhart’s law and final takeaway
01:03:43 Conclusion
Summary
Conor Doherty interviews Joannes Vermorel (CEO and Founder of Lokad) about optimizing Black Friday without sacrificing long-term profitability. Using the fictional billion-dollar retailer Dublin, they examine discounts, probabilistic forecasting, product selection, omnichannel allocation, stock rebalancing, and rate of return. Joannes argues that Black Friday should be treated as part of everyday supply chain optimization rather than as an isolated event.
Extended Summary
Conor Doherty: Well, Joannes and I just wrapped up a discussion on Black Friday and it was a special discussion because it was in—oh, that’s my watch. Got to take my brain medicine. Leave this in. Special discussion that was requested actually it combines two audience requests.
One was to discuss Black Friday and how to optimize ahead of what is for many of our North American audience one of the biggest retail days of the calendar if not in fact the biggest. But also to get back to the basics with Lokad’s content particularly the supply chain persona lectures. So today’s episode is a bit of a pastiche and mélange of those two requests and Joannes and I take a supply chain persona and we analyze the decision-making that a big retailer would go through or would have to make ahead of committing considerable financial resources to satisfy demand on Black Friday. So we’ll try to get this out as soon as possible.
Hopefully no definitely ahead of Black Friday. So if you’re watching this, hopefully you’ll learn something useful. If yes, reach out to us and let us know. And if not, do not reach out to us and let us know.
But anyway, enjoy the episode. All right, so Joannes, today I’m trying to satisfy two audience requests. The first one is from our predominantly North American audience, many of whom work in retail and many of whom are currently gearing up for Black Friday. So what today is October 22nd, I believe we’re just over a month out from Black Friday.
They will soon see if the decisions that they took a few months ago and the decisions that they’re taking now will be prosperous. Will play out profitably. But they were curious about how we might approach Black Friday and the idea of promotions. We’ll get into that.
And for that reason, I’ve also tried to satisfy a second repeated audience request, which is a return to the supply chain persona series that you did a few years ago. Now, those things take huge amount of time, but what I’ve tried to do here is merge these two requests. And what I’ve done is I’ve created my own fictional billion-dollar omnichannel retailer called Dublin. They’re approaching—
Joannes Vermorel: Exactly.
Conor Doherty: They’re approaching Black Friday. We’ll talk we’ll talk through the decision-making chain that goes into approaching these promotions. I will give you more context later. But before we get into that and I think a good framing device for Lokad’s overall perspective on this kind of a topic, you’ve previously spoken quite a bit about your aversion to promotions and the idea that they are they’re a short-term cure if that.
So before we get into how you should approach optimizing Black Friday, basically the biggest day of at least in-person retail in the world. Before we get into that, what is your overall position on promotions and discounts and yeah elaborate on that please?
Joannes Vermorel: So if you operate brick-and-mortar retail networks you are operating with a lot of fixed assets you know that cost you a lot of money all year long. You can I mean all year, you’re renting you know all year long you can’t you can’t just liberate your stores for the month of August even if there is or July even if there is very few people there and even if December is for example a very good month for you can’t just overallocate you know a square meters just for this month so you have you have this flat you have this flat capacity and you pay for it all the time the same is kind of true for all your infrastructure the supporting fulfillment center you know your warehouses all of that is sitting in the back same thing you know it’s largely static you can you can accommodate a few things but it is it is difficult and then when it comes to manpower you I mean various countries have various degrees of flexibility but the reality is that it’s difficult to be super flexible when it comes to manpower because then you have people who are basically incompetent. So yes, you can bring more staff but if you want to have like high quality operations high reliability and whatnot flooding your operations with people who are just there you know for a few weeks is typically not like a super good recipe. So bottom line is if we could just consider that and that is true for most retailers and this Dublin, you know, persona I would say what you want is just to kind of maximize utilization all year long.
Now the interesting thing is that historically one of the ideas why Black Friday was popularized was to try to have people coming earlier in the low period before the Christmas rush. So you create some buzz some activity to bring people to the store ahead of the rush of the end of the year. So that is good that is well obviously nowadays nobody really decides Black Friday you know it just happens and it’s a habit. Now when it comes to the promotions I think you see we should really keep in mind that in here promotion is not necessarily I would say only discounts and in fact ideally you want to promote as in you know put forward a lot more than discounting.
The problem that I have with extensively relying on discounts as many retail networks are doing is that it kind of destroys the credibility that you might have pricewise with your customers. You know if there is something that one day is sold 100 and then it is sold at minus 40% so it’s 60 and then back to 100 I mean the clients either they think that you have like extravagant margins and usually it’s not the case you know or and they will they will remember what you’re doing and then they will more and more get used to chasing the opportunities of your own discounts. And you’re creating perverse incentives where people will more and more remember and anticipate your future discounts and then will just come only when you discount. So that’s that creates and this sort of bad behavior can typically be made worse if you give bad incentives to your own sales team where they have to achieve targets especially in volume in which case people are desperate to be able to have access to more and more discounts.
You see, people, I mean your own employees who are in charge of essentially pricing and store allocation will just keep internally advocating for more and more discounts and just so that they can achieve their volume targets. But again those volume targets are not the goal. The goal is never the volume. The goal is never to achieve some arbitrary target.
The goal is only one thing. It’s a long-term rate of return. That is what you want to achieve. And here we can see that my experience is that the quasi-totality of the retail chains who are playing actively the games of substantial discounts are not very profitable.
You know at best they have like small single-digit percentage you know of gross margin and it’s not even you know brilliant. It’s for me it’s more like a sign of desperation than a sign of having true mastery of their own destiny and knowing exactly how they want to steer their own their own brand through the market.
Conor Doherty: True. But the I would say the immediate push back to that would be again just we’ll get into more details about Dublin but just any retailer they don’t set their competitors prices and they also don’t set the market norms and again we can say in a vacuum and then in reality in reality there are certain periods where customers not as a result of your actions just as a result of the market norms itself are programmed to expect discounts. So, for example, you have two retail two retailers. One who respects or witnesses or participates in Black Friday or Christmas discounts sales yeah discounts, excuse me, and then one that does not.
How can I parse or understand whether or not I’m making the right decision there when at face value if I do not participate, my competitor just vacuums up all of the all of the traffic?
Joannes Vermorel: Yes. But I mean first, do they? Because you see the problem is that your competitors operate with the same constraints as you do. They have fixed assets for them.
They just cannot absorb 10 times the demand you know because nothing follows you know there the square meters are not sufficient the warehouse capacity is not sufficient the staff is not sufficient. So this very quickly leads to very poor customer service where people wait for a long time and they are not really satisfied. And so my take is that you know this is this those mechanism are typically have been typically for most companies been abused in the sense they are again that’s corporate incentives where everybody internally is pushing for volumes because that’s the way they can achieve their targets. But the but because targets are easy to compute, easy to measure, targets in volume and never targets in rate of return.
You see that’s the thing is that we have a very strong disconnect especially for in retail between what is really the long-term economic interest which are the rate of return and the sort of quarterly targets that are given to employees. And here my suggestion is that again we have to think that it is about promoting the retail the retail brand. It doesn’t mean that everything has to be discounted. It doesn’t mean that you need to follow all your competitors at the exact same price.
You know Vuitton is doing a fantastic business and they don’t try to match the price of their competitors. You know it’s ultimately if you want if you look at if you want to be the cheapest of the cheapest the reality is that the people who achieve that tend to do very little discounts because it’s like the very low price every single day and that’s the way they build their trust with their clients. So you see that’s the thing is that usually the brands who are doing those discounts are usually brands that are not super competitive price-wise. And so my suggestion is typically obviously when all your customers are only coming from the discounts.
The cure is a decade-long journey where you need to gradually reduce the amount of discounts that you’re offering. We have done that for a few of our a few of our of our clients that had like massive discounts. I mean it was like in absolute volume it was like over 30% of total discount. So a and to shrink that by half and even that is still is still quite high and it took it took pretty much for a client I’m can think of it took literally five years to do that.
I mean you cannot you cannot do that overnight because again you have to gradually diminish that so that you don’t lose your customer base overnight. They still you still play on their anticipation but you want to convert to a place where you have much fewer discounts. You have some but the promotions can be just it can be on other things. It doesn’t necessarily have to be hard discount on price.
It can be also having limited series for specific products that will only be there for a few days because you have limited quantity and you advertise this scarcity rather than advertising the fact that you have a discount.
Conor Doherty: Well, before we actually transition to the to the actual persona Dublin, just you’re talking about targets, one of the one of the questions that’s was posed to me and I’m posing it to you is the idea of being able to differentiate between let’s say X volume of sales as a result of a 20% discount today. How can I know whether or not I would have sold that same volume next month at full price? Like teasing apart what is like basically the distorted demand signal that you get as a result of yes discounts plus well have I have I basically just sacrificed more or less guaranteed full-price sales a month from today so that I can have less money but right now and again evaluating whether or not that is even a valuable tradeoff.
Joannes Vermorel: Yes. I mean and here we are back to the classic supply chain theory is just so dysfunctional. So the classic supply chain theory do not even acknowledge that pricing exist. No, I have I have many books, fake books, university textbooks on my on my desk and they the thing is just absent thus essentially that’s why people are playing those game of volumes because that’s the only thing that they can see pricing is not even part of the theory.
So yes, you need a predictive model that tell you what would have been an estimated an estimation of the demand if you have had a different price considering the price of your competitors. So it is it is difficult and it is the difficulty is compounded by the fact that you have many elements for pricing that are very slow moving. So for example, if you are over if you are more expensive than your competitors, you don’t lose all your clients overnight. It just it is just a slow attrition of your customer base that is gradually moving to your competitors.
If you don’t I mean all things considered equal obviously if you provide superior service a better availability better perception because the store is like is cleaner more orderly whatever then that’s a that’s a different that’s a different product that you’re effectively selling because what comes around is also part of the whole experience But fundamentally your that’s what at Lokad we refer to as functional forecasts. So it’s not just probabilistic where you have probabilities. It is fundamentally a function of price that you can decide and it is not just elasticity because the reality is that pro that customer responses tend to be highly nonlinear and also what people what is typically misunderstood about pricing is that usually people will not look first at your competitor for pricing. They will first look at yourself and they will make price comparisons within the same retailer itself.
Conor Doherty: I think you gave the example before I think the example you gave not long ago was milk. So for example, if you increase there’s a point at which you can increase let’s say milk. It’s a simple example but by one cent or one penny you’re not going to lose all your clients but if you just add one penny one penny one penny at a certain point it will just completely collapse.
Joannes Vermorel: Yes. Because you’ve gone too far. And also you have cannibalization and substitution that happens inside you know your own assortment because it’s not I mean we are not talking about one product in isolation is that usually you are let’s say for example you are a do-it-yourself store. Yeah.
You have maybe like three different ranges of screwdrivers. Yes. Like the super cheap, the good value for price and the sort of good brand last forever. Yeah.
And so people when they do when they start thinking about the pricing before checking the price of what your competitor is doing they will first assess the relative prices of what you are offering and I’m mentioning screwdrivers because for that’s a that’s an interesting case because you can buy them you know individually you can buy a pack that gives you like many heads do you want a box no box? Do you want something that is compact? Do you want something that is big? I mean, do you want something electrical or not?
You want Yeah, exactly. And you have like many options and most of the of the analysis of your client in terms of pricing will be just thinking about what is in front of this person. And that’s why and that’s where the strongest nonlinearity happens because if you have this thing that gets slightly more expensive than this one and it’s literally on the same shelf and the client is in front of it then they will have like massive switches you know much stronger than walking a few kilometers or miles or taking their car to go to another store that they don’t know. That is much more immediate and that’s so does so bottom line is you need a predictive model for what will happen short-term and long-term with your prices and how it will steer the behavior of your of your customers and we can say that there is a lot of fancy data science involved in involved in that and that’s true although I would mitigate you know the statement it’s only fancy data science is that the data science is only about pretty much the short-term analysis when it comes to long-term impact usually you have to have like high level intelligence and make I would say judgment calls because you cannot simulate or even try to extrapolate what will happen over the course of the next decade you need to put in your model judgment callss such as for example how people get used to a discount policy.
It’s something that will play its course through a decade. And you’re not going to learn the patterns of that through the data. I mean maybe the historical data can tell you a few things about that. But ultimately it will be a judgment calls from the supply chain scientist to model that correctly.
Conor Doherty: Okay, let’s push on and put this into practice. And before I do, in honor of this episode and because I know how much you love technology, I know you do. You love computers. And because of that, what I’ve done for you is I’ve written you a piece of paper.
And on that piece of paper in green ink, you’ll have the context of Dublin, which I will read. I wrote in capitals. I hope you can. That’s just for your ongoing reference in case you want to know what the turnover is.
Yes. Basically, so we’ll flash up the information on screen, but we want to test these ideas in the context of Dublin. You can read, I’m just going to say this for the audience. So, Dublin has a billion-dollars in annual sales, 100 stores, and a website that contributes approximately a quarter of its sales.
For the North American audience, essentially, think Target, but without the groceries. So, it sells clothing, homeware, toys, household electrical appliances. Now, last Black Friday, Dublin had record sales, but you know, decisions weren’t brilliant. So, they actually had a lot of stockouts, a lot of late deliveries because orders were placed too late, and they also had a lot of dead stock then in January, so they weren’t able to sell it in over the Christmas period either.
This year, Dublin wants to do better. So Joannes, given the context as you now understand it and everything we’ve said so far as a starting point, how should the executives of this billion-dollar general retailer consider the problem of doing better? Because you hear that a lot. I want to do we know we’re not doing brilliant.
I hear that a lot. We’re not we know we can be doing better, but in this context, what is doing better at ground zero? And then we’ll build from there.
Joannes Vermorel: And first we need to embrace the economic perspective. Okay? You know doing better is improving the rate of return. Okay, that’s as simple as that.
Conor Doherty: And express all of this now in the context I’ve given you. So that’s the framing device going forward.
Joannes Vermorel: So again the rate of return is I invest $1,000 I immobilize $1,000 in this business and within one year through inventory rotations how many dollars I have now. And you want ideally to have maybe after a year $1,200. That would be 20%. That’s nice.
That’s that would be nice. That’d be pretty good. And so this is this is the rate of return. So that’s you put a certain amount of money and it compounds through over the course of the year and you can look at how much money you have.
That’s literally the key idea of the of the rate of return. How people are in retail are playing this game. They just essentially buy and manage at a at a certain price and sell at a higher price point and then let the money rotate ideally many times through the year and that’s how you can have you know compounds interest rates and generate your profits. Okay.
So that’s that is the thing is that this is I mean for me this is now relatively self-evident because that has been you know the way Lokad has been doing things for more than a decade. Yes. But the reality is that most retailers are still not operating from this perspective. They think volume targets and they think as okay we project that and thus this will happen we will be good and that is just not there.
So because once we say okay the rate of return the next step is to think about the future is uncertain and thus especially when we have like a highly volatile event like Black Friday you cannot say for all your products I’m going to sell this and this and this and this because it’s you have to understand that the amount of uncertainty grows with the spike. You know, if you’re selling something super smooth and steady, then you can be relatively certain about how much you’re going to sell next week for example. But if you want to have a look something that is several months ahead and on top of that it is something that comes with a big push in terms of advertising and a big twist in terms of pricing then the uncertainty becomes enormous. You know that’s and also the fact that your competitors themselves are going to do a lot of things.
So you are you are setting yourself in a position where you have like tons of uncertainty and the reality when people have you know those overstocks and whatnot is that they just say oh yes we are super confident we will sell 2,000 units we just know it. We just know it and the reality is that they don’t know it and the volat the uncertainty is very high and turn out that maybe they sell they it went well they sold 1,500 so that’s three quarter but it’s the 500 unit might represent something that is more than what they are selling in normal conditions for the entire rest of the year and thus even if it’s most of the merchandise was indeed pushed and liquidated through this event. You may end up with month and month of inventory and that qualifies that as dead stock after a while.
Conor Doherty: Well, is that exactly what has happened in this scenario?
Joannes Vermorel: And the problem is that if you because you liquidated your stuff with usually a massive discount you had razor-thin margins and thus it doesn’t take much excess inventory to for the operation to become unprofitable. And that’s why again the way Lokad thinks of it is we look at all the possible futures and we want to average out what will be the average rate of return. So you take a trajectory say okay we are going to purchase this quantity okay let’s look at all the possible futures how this decision this option is playing out and what is its net rate of return once you’ve integrated all the events as they unfold but that means that in order to be able to do this risk analysis you need to have risk in the first place you need to model the risk and modeling the risk is Exactly what Lokad does when we say we do and we advocate for probabilistic forecasting. Probabilistic forecasting is a way to acknowledge that the future is very uncertain.
You assign likelihoods and that’s you have factors that just increase uncertainty. So your distributions probability distributions would be even more spread out especially if you push the pricing your prices in areas that you rarely visit. And if you do pro I would say advertising that you rarely do etc. I mean all of that is pushing the envelope of uncertainty into territories where your data doesn’t tell you much and the reality is that the consequence is really okay uncertainty is super large and the interesting thing is that when you are facing a lot of uncertainty what the typical consequence is in order to have a better rate of return is to be more conservative.
You see that just as simple as that it will the most of the time when we work with clients and we embrace and acknowledge all the uncertainties there is the conclusion is low is more conservative than their plan and that’s that goes you see the problem is that again usually we get a lot of push backs because employees have all their all their bonuses all their targets on them hitting volumes them hitting this and that and that and but the targets never for example you can people have like a bonus for a target for black Friday in volume but at no point there is anything that would you know mitigate that taking into account the amount of dead stock that is generated
Conor Doherty: The example earlier like you sell x volume at 20% discount would you have sold it at full volume earlier but if the t if the bonuses are tied to this time horizon Then I just did it now and that’s a problem. You’ve lost money in the long run.
Joannes Vermorel: Yes. And giving out bonuses and we are we are back to you’re creating perverse incentives that are going to be gamed because those incentives are not aligned with the long-term interest of your company and that’s as simple as that. So Lokad the way we address that is embrace the economic perspective then do what is necessary to not ignore the risk and in practice the best tool that we have for that is probabilistic forecasting and then you do your assessment with those functional forecasts which is a different things from the probabilities. It’s just about it’s just to reflect the fact that some decisions will be made in the future not yet but by yourself.
That’s why such as the prices. So the prices your prices are not uncertain. They are just undecided. You know there is a nuance between what is irreducible uncertainty and what is just undecided but you will be calling the shots, ultimately.
Conor Doherty: Well speaking of calling the shots one of like this the starting point beyond like before you even get to the prices is in fact choosing what it is in your assortment that you are in fact going to discount in this context. So again, we said about Dublin, you have your cue card. I believe I mentioned for again people who don’t have a cue card, a well-written cue card, clothing, homeware, toys, and household electrical appliances. For the sake of discussion, let’s say as big as fridges, I don’t know, maybe not air conditioners, but again, as a general framework, the perspective there is like how do you choose what it is that you’re going to going to discount?
Because obviously you don’t want to discount everything that would as a flat rule you want some things that bring people in but not everything.
Joannes Vermorel: Yes. And again the worst part about discounting is opportunity buys. So people who just would have bought anyway and they just discover on the day. Oh discount good.
I mean it changed nothing but thank you. Thank you very much.
Conor Doherty: Absolutely. Exactly. I bought a projector under those exact circumstances last week. Yes.
I would have bought it anyway. Full price.
Joannes Vermorel: And typically appliances are exactly in this category. An appliance is a is a decision that people will think about for months.
Conor Doherty: Yes. Exactly. And I use ChatGPT to limit down all my options.
Joannes Vermorel: And so people they don’t decide to buy, you know, a very nice double door American fridge just because it happens.
Conor Doherty: It’s not an impulse purchase.
Joannes Vermorel: Yes. Because there is like a 15% discount today. So typically I would say appliances are like super bad for promotions. Typically what you what you want is something where people will be driven by impulse buy.
Okay. And toys are not bad. Toys in the example would be toys Ideally, if you can maybe clothing, but you can you can even have for example for toys, it would be even better if it’s something that would be most likely be you know exhausted before Christmas. So you see so that it’s something where okay it’s an opportunity and it’ll be gone before Christmas and even before and because I can’t just postpone the for Christmas because in Christmas there will be just none left and again it can be cheap doesn’t have to be rare it can be for example I don’t know you know something that is with a license it’s like a limited edition of something Pokemon something Naruto something Harry Potter whatever but pick your license but you see it’s the idea is that you want to find the products typically when you want to have a discount that would be my intuition something that is relatively cheap so that the volume of discount you can have like a big percentage and it will be more likely to trigger an impulse buy if it’s something that is quite cheap overall as opposed to give discounts on that are not going to be very impressive because you see if you give you have like a 3,000 euro you know double door expensive American fridge and you give a 10% discount it’s 10% is not that impressive and that’s already €300 worth of discount while if you’re selling something for €10 and you give a 50% discount, it’s only €5 that you’re giving away.
So you really want to think of how can you achieve you know something that is again you have to think about the promotion aspect so that it’s visible it is attractive but also it is fundamentally cheap and ideally it doesn’t you know it doesn’t interfere too much with the sort of price positioning that you try to have through the entire year where people will trust you to have reasonable price, you know, that are like stable over time and not people will not feel like defrauded because you see the thing with when you do a discount, imagine you buy an expensive fridge and then two weeks after you have like a 10% discount you can have resentment say ah I just the return period in fact you can bring it back. Yeah, exactly. And you can have people who do these sort of games to do. I do when you said I do know someone who’s done that recently.
Conor Doherty: It was not me. I still have my projector, but I do know people who’ve done that.
Joannes Vermorel: Yes. And again, that’s that is hell. And that’s really not what you want to do.
Conor Doherty: It’s interesting because when you also Okay. When you think about the purchasing habits, so again, people who might go to Dublin, most people will go in to buy multiple things. So again, we talked about the basket perspective before. Now, there are no groceries in this store, but typically people don’t go in and buy one toy for their child.
They’ll buy that plus other things. However, the buying patterns of someone who will come in to buy a fridge might in fact be that they literally just want a fridge. They are not there to browse. So, selecting discounts on things where it is kind of just a one-and-done purchase like it’s a one-shot visit.
They’re not they’re not carrying a fridge and a whole bunch of clothes and a whole bunch of plates and bowls, etc. So, how does how does like the basket perspective, which we’ve talked about before, that people buy things together, not having one thing might spoil the whole purchasing experience, how does that fit into the Black Friday discount perspective for Dublin?
Joannes Vermorel: I mean again the thing is that you have a lot of here we are talking it’s we have a phenomenon that is quite complex and it is an event that is at a point in time and again just like I was saying you have fixed assets in your companies for most everything including your pricing analyst and merchandising analyst. So my take is that here we’re entering a territory where you need to robotize all of that you see because again people need to spend time to just manage the day-to-day business and now they would need to have like a lot more time to do this to manage the to manage this once-a-year event and my suggestion that’s the way Lokad does it is that Black Friday and all the quasi-cyclical events need to be just business as usual and you have a model that just models this what we call quasi-cyclicity so it’s not exactly cyclical I mean it comes every year but just like Chinese New Year it can shift a little bit yes yeah and all the decisions that you generate day after day reflect that you know and again I think people overestimate the sort of super macro decision that are just for Black Friday and all the mundane stuff that happens such as for example if you want to just have your store ready for Black Friday, it means that probably up to probably three weeks before Black Friday, you will start sending more stuff to your stores to fill them up. And why several weeks before? Because you don’t want to end up with a massive delivery where your fulfillment centers have to work completely overtime so that they can ship twice as much merchandise as they usually do just in the last 5 days before Black Friday.
So you see that’s typically you want this retro-planning to be done smoothly and taking advantage of because for example your stores might be unequal. You might have a store that has some a little bit of excess square meters. Thus you can start sending stuff three weeks in advance. It’s okay.
It will fit. But you might have some other stores that are quite tight in square in shelf space and thus you can’t really preload them many weeks in advance because there is just not enough shelf space for that. So again that’s a lot of micromanagement and if you want to be able to do it right to do it smooth and to do it at scale all of that would need to be robotized. So we are you know this is like the third pillar you know economic perspective probabilistic modeling and then robotization so that all of that is done and not in silos but ultimately it’s just a continuum of decisions that are taken every day and this event you know black Friday yes it’s a point in time but all your decisions anticipate that and start to do little nudge as you’re going into Black Friday and afterward same thing you do little nudge in reverse to address to address the leftovers for example rebalancing inventory over your network
Conor Doherty: That was literally the next question was about network allocation so because again as we said right at the in the context Dublin’s omnichannel so it has 100 stores and it also has the website that accounts for 25% of its business or at least last year did 25% of its turnover. So the question would then be how do you allocate stock because you have the same the same items on sale both online and then you want them in the physical store because people are going to arrive and if they don’t have the item they want they might leave because again basket perspective people buy in baskets but how do you at a high level how does Dublin optimize the allocation there where there’s enough you know in warehouses for online but also enough in stores it’s 100 stores like not all the stores will have the is that
Joannes Vermorel: Again you go for the you know what we call the risk-adjusted you know optimal decision. So it’s you take into account all the possible futures and you call the shots and you’re essentially gambling and there is no other way small bets small bets on every store now because erraticity is very high. Some stores will liquidate certain products and be out of stock for those products. It’s fine.
It’s normal. While some others might have excess. Again, that’s exactly what will happen when the erraticity is very high. So you will end up with a lot of imbalance for your network where you have store A and store A has one product in excess and another out of stock and for store B it’s the exact opposite.
Okay. Now what we need to happen is gradually rebalance. The problem is that in stock rebalancing is conceptually relatively straightforward. You know it’s just about you take some stuff you put them in a truck and you move it to another store or you do a circuit.
The complexity is mostly that you have a combinatorial explosion of all the possibilities. Because you have to consider every single SKU in every single store and how you’re going to do a big you know potential shuffle and you have nonlinear transportation costs. You don’t want to move one spoon from one store and just iterate like that. You have to think in terms of shipments and maybe you have stores you have sorry trucks that are already doing circuits.
So you want to take advantage of that, you know, I put something in the truck today as the truck is delivering stuff to me. I put stuff in the truck and then the next stop for the truck, we just unload the stuff and it’s it contributes to rebalancing. All of that is again back to automatization. Those things are not very difficult and although it might look a little bit scary because we are talking of tens of thousands of SKUs per store and 100 stores.
So it’s like obviously millions of combinations the reality is that those are not big numbers for modern computers. So to a large extent although you can’t just naively brute force a problem you don’t necessarily need to have something that is hyper sophisticated in terms of numerical analysis to just be able to optimize that but the key idea is that again it must be business as usual so you it’s not like we do inventory rebalancing for black Friday no you have a logic that suggests inventory rebalancing continuously and This logic has an economic foundation. So it will not suggest inventory rebalancing it is if it’s not economically viable. So if there is not enough for example mass yes to rebalance then we say no I’m not going to suggest to move one spoon from one store to another that would be crazy.
So I need to reach a certain volume so the opportunities are qualified. Do they reach the critical mass of profitability? Again, we’re thinking rate of return and it will just turn out that after Black Friday the reality will be there will be more opportunities popping up just because well we have a situation that is just after a big a big event and thus naturally your system will suggest those rebalancing but it’s not for black Friday it is just business as usual and the day-to-day numerical recipes would just capture the fact that the situation where there is more opportunities lying around at this period.
Conor Doherty: Well, it is when you talk about the like balancing the distribution of inventory across again an omni channel network if you were to compare approaches between how you prepare for Black Friday and how you might prepare for Christmas. Though you though you advocate okay it’s a business as usual perspective they shouldn’t be radically different there are clear nuances here that are worth pointing out so for example Christmas discounts are more like a period like there’s the weeks leading up yes it’s more it’s not as concentrated so for example oh if I if I have a stockout in a flagship store on December 21st well I still have I can still get there the 22nd or the 23rd and there will still be people going there you is it fair to say it’s a sort of you have one shot at getting the allocation right for Black Friday when all of the foot traffic is
Joannes Vermorel: I think it’s a mistake of analysis to think like that you know again your business is a continuum you need to have to serve your clients before and after you don’t want you can
Conor Doherty: The volume won’t be there the day after is the point though that’s what I’m getting at the volume of people not the volume of inventory
Joannes Vermorel: Yes but you will be left you will have inventory leftovers and whatnot. And again, if Black Friday starts to become, you know, a two-digit percentage in your yearly turnover, that’s a massive problem. That’s okay. Yeah, that’s a massive problem.
Again, all your assets are fixed. You know, you pay the rent for your store every week of the year, not just the one week where you have Black Friday. So it is again there are some businesses maybe online businesses that can say you know what we do 80% of our of our yearly business on this day and that’s okay because everything is virtual and they can serve 1 million clients one day and zero the next and it’s just fine. Okay, maybe for online businesses, but when you have infrastructure, you want that infrastructure to be used all year long.
And if you don’t do that, a smarter competitor of yours will be doing that and they will just out compete you just because they make a better use of their assets. So for me again you want to think of your supply chain as a productive asset as something that generate decisions day after day and you want all the things that are highly I would say that are just business as usual to be managed business as usual. Black Friday is business as usual. That’s just an one event because you see what about Chinese New Year?
What about Ramadan? What about Easter? What about Mother’s Day? What about Father’s Day?
What about X? What about Singles Day? There is there is there is a lot of things. Some obviously they don’t all have the same magnitude.
Fine. But nevertheless, they are all like quasi-cyclical events. They should be modeled as such. And you want your supply chain to be pretty much on autopilot for those things.
And that’s where I say, you see, because the problem is that if you approach that Black Friday as a silo, an opportunity disconnected from the rest, you’re going to have a lot of stuff that just collide. You know you buy stuff for Black Friday but you have like near-perfect substitutes that are already in the store and you have an excess because people you don’t realize that in fact your sales have cannibalization between each other etc I mean for me it is wrong to think that you can you can segregate the black Friday you should not and that’s typically a mistake that people do because they want they are lacking the robotization component that they are doing it with people and the way they want to manage people is divide and conquer slice and dice the problem and say oh we have a task force that is dedicated to black Friday boom and this team is separate from the one that does the day-to-day business and I think it is logically incorrect you want to have the same numerical recipes that just deal with the whole package and the Black Friday being just one event among many others. Maybe a bigger one, but that’s it.
Conor Doherty: Well, this so again I’m glad you said it one event amongst many because again it is as my last question established is typically Black Friday decisions are typically sorry they presage or they foreshadow Christmas and you mentioned earlier about well we should measure everything all the decisions the impact of the decisions through the rate of return. Okay, let’s assume we accept that people listening though will naturally have a follow-up question which is okay rate of return over what time horizon. So for example, I take decisions towards Black Friday. Those decisions are obviously going to impact the demand that I experience at Christmas.
So should it be after Christmas? But what about the sales that I made for Black Friday? Let’s say November I think it’s 27th. They have a 45-day return period.
So now we’re into 2027. So at what point do I start can I reasonably start to evaluate whether or not the decisions I made a were successful in terms of rate of return.
Joannes Vermorel: So that’s a little bit a tricky question and it is I’m aware that it is and by the way there is a technical answer and it is in your book in your book it’s called what I call the aperiodic rate of return but fundamentally you want to look up till the end of time
Conor Doherty: Yeah we can high level answer to this I realize it is technical
Joannes Vermorel: So first all the arbitrary cutoffs are wrong you know there is you should not say I’m only looking up to one day why one day or one week why one week one month one etc. So logically there is no reason to ever cut. So morally directionally what you say I look up till the end of time. Okay.
Now that’s okay. By the way that’s what people do in banking with interest and whatnot. So it’s not like it’s yes in practice you have you have FX numerical FX that just make the return of money that you can make a very long time from now decay into nothing. So that’s just a numerical effect.
But then the thing is that you want to compute your rate of return. And what I’m saying with the aperiodic perspective is say the rate of return vary day by day you know till the end of time. The problem is that it’s a curve that will hit a maximum and then decrease when you look at a series of operations you know because you take a decision, it costs money and then it brings money and then it costs money and brings money and for a given operation such as stock purchasing it will decay into ultimately into nothingness because your inventory is exhausted or you just throw it away anyway you know if I say I purchase something I’m spending now I will have money that comes in as I sell it and at some point I would just liquidate and get rid and it will convert to zero. The thing is that if I if I look at my operation let’s say for example I purchase 1,000 units and within year one I have I have I have sold 99% of it but the 1% that remains may take an entire year to be sold.
So if I look at just year one, maybe I’ve done like a rate of return of let’s say 25%. And then because the 1% it’s a product that has nice margin whatever the 1% is barely touching that I could write it off and that would be I write it off I have 25%. If I extend my horizon of analysis on two years, I would need to divide this 25% by two.
Conor Doherty: Exactly. And that would become 12.5.
Joannes Vermorel: But is it fair? No. Because by the end of year one, I already had the money back. So why should I wait for one more year to think about my rate of return?
This doesn’t make sense. And imagine that you have one unit that takes a thousand years to sell. You know, I understand you would you would end up with a rate of return that would be completely negligible and this is completely fictitious. So the perspective that I propose with the aperiodic rate of return is to say look at the evolution of day by day of the rate of return and what you will be counting will be the day in the future where it’s the highest because that’s your effective rate of return because at this point this is not theoretical.
This is a point where you have the money back. So it’s you it’s ready to be reinvested into something else. So that’s why I say you think of your rate of return in the aperiodic perspective. So until the end of time, but to compute the actual rate of what you got, you take the point where you have the highest effective rate of return.
And that’s the analysis. It’s a little bit technical but think of it as a fancy way to say I want to compare decisions that are just that do not have the same horizons. That’s why I say it’s aperiodic. It’s because it is not something where you decide annually I have a decision that will that will be you know over one month from now and another that will take a year to unfold.
We are not on the same horizons and thus we need to have a rate of return that is not that’s why I say it’s aperiodic that is not defined according to a preset period
Conor Doherty: Someone listening to that will and again you are just describing it’s perfectly fine I understand what is required to track and measure that what you the periodic rate of return what you what it’s not
Joannes Vermorel: It’s not very complicated again the formula I’m not going to describe it but the formula is
Conor Doherty: But there you’re talking about basically a dashboard that they look at.
Joannes Vermorel: I mean not even because the reality is that the rate of return it’s at the moment when you want to take you consider all the options and you need you see the thing is that computing the aperiodic rate of return after the fact is very trivial. It’s something that you could do in Excel but you have to pick a point. The problem is
Conor Doherty: Yeah even in Excel it could be done easy.
Joannes Vermorel: The problem is this is not for the past that you’re interested in. You want to do that for the decisions that have not yet been made. You see that’s the tricky part. Because again analyzing the profitability of the past decisions is relatively straightforward.
But what is what makes it difficult is that you want to do that for the upcoming decision. And here we are back to we have probabilities into the mix. Probabilities do not mix with Excel. There you need to have instruments that let you do this probabilistic modeling.
So that’s a guess is that and it’s not that it the problem would be the exact same if we were picking another indicator than the aperiodic rate of return. The real problem is not this indicator. The problem is that is the probabilistic modeling that is going to represent like the big thing that is difficult to handle and that do not fit into Excel.
Conor Doherty: Okay. Well, as an overall concluding thought in this persona framework, you would or would not advise Dublin to participate in Black Friday discounting.
Joannes Vermorel: Again, I think the point is that it depends on how you do it. You see that’s the thing that economic perspective doesn’t say you know it doesn’t think in terms of absolutes what it says is that you need to think of the rate of return what are you doing is it profitable in the long-term interest so promotions can be good but there are so many ways to promote and what is exactly the range of your assortment that you’re promoting all of it probably wrong probably wrong a small you know a small selection a small selection that only exists for the duration of the operation in the store. Probably good. Is can you have do you have a big payback?
You know, if you can you serve because you see the thing is that there are so many problems that are contingent on that is what if your stores are just not able to absorb the capacity? What if your customers resent you or is it a great experience and they love you for that? You see, so it’s really all the nuance. What I’m saying is that you need to robotize the decisions so that Black Friday just becomes one thing among the rest and then you really need we because you’ve liberated a lot of time for people to think it through and to really think of how do we approach Black Friday?
What do we promote in the sense put forward you know what is the image that we want to convey what is the message what is the story that we tell you know those are like the very good questions as opposed to just go like hyper aggressive and then give like a massive you know 30% plus discounts that are only going to create problems down the road because then customers don’t trust your day-to-day prices, then they will wait and they will start to game. And it’s and ultimately you will be out competed by a competitor that makes a better use of their of their fixed assets, the same sort of fixed assets that you that you have to.
Conor Doherty: Yeah. I think I mentioned this anecdote before, but we both know somebody who they have an Excel spreadsheet and they track at the local stores of Monoprix. We’re in Paris and they track they have a very specific coffee brand that they like and they are periodically on it appears as if there’s just randomly on discounts but he actually tracked that it’s approximately every two and a half months more or less. So he but what happens is he goes in and then he will just clear out the shelf.
So you have one person taking I don’t know 10 15 units of this very particular coffee and then goes to the next store the next week when they have theirs and just clears it all out. And again the Netherlands is very famous for I mean the population of the Netherlands is especially famous for that. So again it’s not a it’s not a fictitious trend that you’re describing there. There are no are they the majority?
No. But again think about the demand signal. You have a you have a stock out because one person walked in and bought 15 units, that doesn’t mean that well I got to put 15 more back or 15 more back, excuse me, immediately because they might not sell at all. So again, the more embracing the uncertainty perspective as you’ve been talking about at length now.
Joannes Vermorel: Yes. And again, a very holistic perspective where you really think of the long-term interest of the company. And that’s why I say, you know, the economic perspective. I’m not talking about the quarterly financials.
You know, it’s usually it’s very narrow-minded. It’s myopic. What you want is to have this economic perspective which is much more demanding intellectually. But you see, it’s not actually that much more complicated in terms of analytics.
It’s just more demanding in terms of you have to really think it through in terms of what is the strategy, what are we even trying to do, how are we trying to serve and out compete our competitors and as opposed to I would say relatively dumb and lazy shortcuts like just give the teams a big volume target because those sort of things are exactly sort of things that a dumb and lazy corporation will do and it will invariably backfire. It’s like a all this sort of all those things are just going to be gamed by your own employees and ultimately represent a disservice for your customers.
Conor Doherty: That’s Goodhart’s law, right? When a metric becomes the target, it seems
Joannes Vermorel: Exactly. Goodhart’s law for the for the audience interested. But yes, it’s a and that’s by the way this is the Goodhart’s law is also a problem for the numerical recipe because we set the targets and that’s why the way Lokad works with numerical recipes that have like targets is that constantly revising the target is a journey. See that’s we that’s the thing is that the modelization of the economic of the long-term interest of the company and what goes into this rate of return maximization.
It is not something that is that follows I would say a Descartes perspective where you just decide first that and then everything follows. No, no, it’s a it’s a it’s an empirical process and I called it in the series of lectures empirical optimization where you set your objective, you look at the decisions you that are that end up generated according to this target and then you identify you identify the insane decisions and then you refine the target and you keep iterating and that’s so the target the formula for the target is constantly being revised. That’s the job of the supply chain scientist. Not gratuitously, but just because it reflects the ever-moving change of strategy of the company.
Conor Doherty: All right. Well, Joannes, thank you very much. I don’t have any more questions, but I know that we will come back to a very similar topic when we discuss Christmas and how people should optimize for that. But, until then, I don’t have any more questions for you.
Thank you very much for joining me. And to you, thank you very much for watching. I hope our audience in North America appreciates that. If you want to learn more about optimizing inventory and allocation ahead of, well, actually just as business as usual, but certainly ahead of big events, feel free to connect with Joannes and me on LinkedIn or to reach out to us directly at contact@lokad.com.
And with that, we’ll be back soon with other topics. So, there’s nothing left to say except get back to work.
