00:00:07 Large-scale disruptions and long tail risks in supply chains.
00:02:24 The difficulty of assessing and preparing for tail risks in large companies.
00:05:17 Importance of strong leadership in identifying and mitigating tail risks.
00:06:30 Historical context and predicting potential large-scale disruptions.
00:07:08 The fallacy of forecasting rare events and the need for probabilistic forecasting.
00:09:49 Stockpiling as a short-term and expensive solution for supply chain risks.
00:11:55 The importance of agile and adaptable production in mitigating risks.
00:14:14 Lack of agility in decision-making processes and its impact on supply chains.
00:15:10 Adjusting pricing mechanisms as a response to supply chain disruptions.
00:17:08 Producing enough essential products during crisis.
00:18:36 Smart pricing policy to avoid stockpiling and shortages.
00:21:10 Less obvious solutions for supply chain management.
00:24:31 Recomputing and optimizing supply chain strategies quickly.
In the interview, Kieran Chandler and Joannes Vermorel discuss tail risks and their impact on supply chains. They emphasize the importance of recognizing and preparing for these infrequent but potentially disastrous events. Vermorel highlights the need for strong leadership, probabilistic forecasting, and agile production processes to manage supply chain risks. He also stresses the significance of software agility and the ability to quickly recompute, reprocess, and reoptimize strategies during disruptions. The conversation underlines the crucial role of adaptability, programmatic expressiveness, and fast response in supply chain management to mitigate the risks of unforeseen events.
In this interview, host Kieran Chandler and Joannes Vermorel, founder of Lokad, discuss the concept of tail risks, long tail events, and their impact on supply chains. Vermorel explains that tail risks are infrequent events that most companies do not have routines to handle, and they are often what ultimately causes a company to fail. He describes tail risks as extreme, improbable events that are more likely to occur if a company is playing an iterated game, such as the repetitive nature of supply chain cycles.
Vermorel emphasizes that forecasts are not as important for tail events, as their probability is typically non-zero. This means that, given enough time, a company will inevitably face a tail event. He uses the term “Black Swan” to describe these scenarios that may occur once in a decade or even more infrequently. When asked about how people typically approach these scenarios, Vermorel notes that denial, wishful thinking, and general unpreparedness are common. He finds it intriguing that larger organizations, which are typically the ones operating supply chains, have more difficulty assessing risks.
The challenge of addressing tail risks in a company often stems from the difficulty in reaching consensus on the importance of these rare events. Many people within a company may not be aware of the risks or may view them as too improbable to be worth considering. This can lead to a dominant perspective within the organization that does not prioritize preparation for tail events.
The interview focuses on the concept of tail risks and their impact on supply chains, highlighting the importance of recognizing and preparing for these infrequent but potentially disastrous events. Companies often struggle to properly address tail risks due to their improbable nature and the difficulty of reaching a consensus on their importance within the organization.
They address the challenges that companies face in planning for rare, high-impact events and how to approach risk management effectively.
Vermorel emphasizes the difficulty of projecting the survival of a company in the long term, particularly when most employees have never stayed with a company for more than five years. To mitigate risk, he suggests companies need strong leadership to acknowledge and identify potential problems. He cites a TED talk by Bill Gates from five years ago, where Gates discussed the likelihood of a flu-like virus causing significant disruption in the world, showing that it’s possible to be aware of potential risks ahead of time.
However, Vermorel argues that the focus should not be on forecasting specific events, but rather on assessing the probabilities of various risks and prioritizing them accordingly. This is particularly important when considering that resources are finite, and companies often face multiple risks simultaneously.
He explains that if a risk has a one in a thousand chance of happening and the game is played every day, it will likely happen within the next two decades with near-perfect certainty. Therefore, probabilistic forecasting is essential to help balance competing investments in risk preparation.
When asked about other methodologies for preparing for risk, Vermorel mentions stockpiling as a common but naive approach. While it can help deal with disruptions to some extent, stockpiling is an expensive and often inefficient solution. Stockpiles need constant replenishment, as products degrade over time and have a shelf life.
Instead, Vermorel suggests focusing on strategies that offer more options for facing various risks at once. For instance, having agile production processes that can rapidly switch from one type of product to another can be more cost-effective. This could involve utilizing similar components across a range of products, allowing companies to quickly adapt to changing circumstances and better manage risk in their supply chains.
The conversation highlights the importance of strong leadership, probabilistic forecasting, and agile production processes in managing supply chain risks and optimizing company operations in the face of rare, high-impact events.
The conversation begins with an exploration of how agility in supply chains has suffered due to IT systems and software, making processes more rigid. Vermorel notes that in some cases, physically moving goods between warehouses is faster than reconfiguring IT systems, such as ERPs. The lack of agility is not necessarily a problem with physical processes, but rather with decision-making and implementation, which can be slow even when solutions are obvious.
The conversation then shifts to the importance of having a lean and agile supply chain, which is not necessarily about predicting specific scenarios but rather about being able to react quickly to changing circumstances. Chandler and Vermorel discuss the role of pricing mechanisms in supply chain management, focusing on how governments often impose price controls during shortages. Vermorel argues that this approach is misguided and can make situations worse, as prices serve as a signal for companies to redirect their efforts towards areas that need more attention.
Vermorel emphasizes the importance of producing more to address shortages, rather than just dealing with their consequences. He uses the example of producing hydroalcoholic gel, a simple chemical product, and its plastic container, arguing that there is no reason why production capacity cannot be increased to address shortages. However, companies may be hesitant to invest in more machinery if they are unsure about the longevity of increased demand. Steeper prices can incentivize companies to invest in resources to address temporary shortages.
Vermorel then argues that retailers have failed to implement smart pricing policies to prevent disorganization during crises such as the COVID-19 pandemic. He proposes a simple solution to prevent stockpiling, using toilet paper as an example. Retailers could offer the first pack at a regular price and then significantly increase the price for subsequent packs, deterring customers from excessive buying. Loyalty cards could be used to manage these pricing schemes.
They discussed supply chain optimization and the importance of agility and flexibility in responding to disruptions. Vermorel emphasizes that companies need to be prepared for adversarial behaviors and should prioritize solutions that are responsive and adaptable. He believes that software agility is crucial for addressing organizational challenges during disruptions. Vermorel also suggests that companies should be able to recompute, reprocess, and reoptimize their strategies within hours to effectively handle supply chain disruptions. The interview highlights the need for programmatic expressiveness and fast response in supply chain management to mitigate the risks of unexpected events.
Kieran Chandler: Today on Lokad TV, we’re going to discuss whether these events can truly be forecast for, and what supply chain practitioners can do to limit their impact. Joannes, perhaps you could give us a bit of an overview about what long tail events we’re talking about today.
Joannes Vermorel: Tail risks are all the sorts of risks that happen infrequently, and thus, by design, most companies don’t have a routine to deal with these sorts of risks. Yet, when you think about what is actually killing companies, not humans but actual economical entities made of many people, it’s almost invariably tail risks. It’s something that was unexpected and frequently took very significant proportions to the point where the company dies. Obviously, there can be cases where some companies go through a very gradual death over multiple decades, vanishing to nothingness, but it’s actually quite rare. Usually, those end up with a phenomenon that just keeps happening faster and faster until the company bursts. This is what happened, for example, to Blockbuster, the video rental company. Netflix was a non-issue until it wasn’t, and the disaster just unfolded in a couple of years, like a slow-motion disaster. That’s what tail risks are about: extreme events which are very improbable at any point in time but, if you play an iterated game like supply chains, where it’s about repeating the same cycle, the same production or distribution cycle over and over, even something that appears to be exceedingly improbable, such as something that happens only once per century, will come to fruition. So, back to your initial question about forecasting, what’s interesting about tail events is that you kind of don’t care about the forecast. You just know that if the probability is nonzero, playing the game long enough means that at some point, your company will face a tail event.
Kieran Chandler: We kind of mentioned it in the introduction; we’re talking about those Black Swan scenarios that occur maybe once in a decade or something like that. What have you noticed about how people approach these scenarios?
Joannes Vermorel: The first thing that’s very common is usually denial, wishful thinking, and general unpreparedness. It’s interesting because when it comes to large organizations, companies who operate supply chains are almost inevitably large. You don’t operate a supply chain from your garage. These companies are inevitably large, and it’s intriguing that the larger the company, the more difficult it usually is to assess a risk. Just because it’s kind of the opposite of what you get from design by committee or when you want to have consensus, consensus is always going to be, “Oh, this thing is super improbable; we don’t need to deal with that.” Obviously, if you take an average survey, most people would not even be aware of the risk. So, if you do an average of the opinions of people in a company, the dominant answer will always be, “No, this is way too improbable to happen in the first place, so we don’t care.” That’s where, in general, companies have a hard time starting to approach those risks, just because it’s very difficult. By the way, you have to think that in a company, there are very few people who actually truly care about the long-term.
Kieran Chandler: Survival of the company is very hard. I mean, most people, if you ask, cannot say, “Could you project yourself in this company ten years from now?” Most employees would say, “Well, no, I’ve never stayed in a company more than five years in my entire career.” So, projecting myself a decade ahead is very difficult. So, how do you approach rare events that happen once a decade when your workforce is dominantly made of people who are not even staying three years in your company?
Joannes Vermorel: First, companies need to acknowledge these sorts of problems. Usually, it’s not going to be a committee, so it’s not a vote. You need to have strong leadership and then to identify the problems. Most of them are not that difficult to guess; it’s usually more obvious than people would expect. It’s interesting when you think about people who were really doing smart risk assessment. There are some TED Talks from Bill Gates from five years ago where, for his foundation, Bill and Melinda Gates, he was saying, “What do we consider as the most life-threatening risks that we are seeing for the world today?” Bill Gates was on stage five years ago, saying that a flu-like virus is probably one of the biggest concerns that the world is facing right now. So, it’s not an accurate forecast. Bill Gates was not able to say 2020 will be the year of the problem, but anybody who looked at history knew that these sorts of flu-like problems have happened in human history basically a couple of times per century and have been going on for centuries.
Kieran Chandler: Okay, but that’s one side of things: acknowledging that these rare scenarios might actually occur. But then, I guess the other side is forecasting for these scenarios and forecasting the impact. I mean, they’re so sporadic. Is this something that we can even begin to do?
Joannes Vermorel: First, I think the problem is wrong. You don’t want to forecast that these events will happen. It’s an iterated game; it doesn’t matter. That’s a fallacy when people want to forecast. In a sense, I think that’s also why, at Lokad, when we advocate things like probabilistic forecasting, it’s so hard to understand. If you have a non-trivial probability, it doesn’t matter if your estimate is very incorrect. If you say there is only a 0.1% chance that this thing will kill me, but you are playing the game every single day, in three years, you’re dead. It doesn’t matter if your forecast is correct or incorrect. You need to think of ways to assess the probabilities of those events so that at least you can have a rough prioritization. Is it something that has a chance in a million or a chance in a thousand? You need to be able to have relative scales of the problem. Then it’s all about what can you do to prepare your company for this event that will happen. Forecasting, especially probabilistic forecasting, is just about having the right level of prioritization between competing risks, just because usually there is not one thing that threatens your continued existence. You typically have a series of risks, so you need to prioritize because your resources are finite.
Kieran Chandler: So, what we’re sort of saying there is, even though there’s a very slim chance that something will happen, because it’s iterative and we’re playing this game every day, the chance that it will happen at some point kind of increases?
Joannes Vermorel: It’s not that it will happen; it’s that there is a probability of one. That’s something that that people don’t understand is that if you have something that has only a chance of one thousand of happening, if you play this game every day, it will happen with near perfect certainty within the next two decades. It will happen with priority one. The trick is that it’s iterated. So, forecasting is not to assess whether the risk exists or not; the forecasting is just to balance your competing investment in preparing for the risk.
Kieran Chandler: Yes, it’s a very kind of scary prospect. Probabilistic forecasts are one way of approaching this. Are there any other methodologies out there that also take into account these kinds of scenarios?
Joannes Vermorel: When people think about preparation, usually they think about the naive approach, which is stockpiling. It’s very interesting, and we are seeing that in a live display where people start to stockpile things like toilet paper. Stockpiling is a very expensive mechanism. Yes, to a degree it helps in dealing with disruption, but it goes completely against all the lean, Kanban approaches. Stockpiling is an answer to a certain class of problems, but it’s a very short answer. The problem is that for every dollar that you can invest in stockpiles, it’s typically not very lasting because you are dealing with a problem that is going to be very infrequent. Even if you stockpile, you need to replenish the stockpiles all the time, and that’s a very expensive process just because stocks don’t last forever. Things have a shelf life, products degrade, and plastic degrades. There are very few things that you could put in a closet and come back a decade later, and it will still work. If you put a car in a cave and come back two decades later, your car will not start, and you will need extensive maintenance operations to make the thing work. The same thing applies to batteries or whatever. So, stockpiling is part of the answer, but there are many other strategies.
Kieran Chandler: If stockpiling isn’t a good strategy for a supply chain practitioner, what is a better strategy that they could take?
Joannes Vermorel: When I say stockpiling, it’s not very cost-efficient because it doesn’t give you many options. The problem is that you don’t know in advance which one of your scenarios will strike you down. It’s better when you want to think about preparation to think of options, what you can do to give yourself more options to face many risks at once, which can be much more cost-effective. For example, you can have more agile production where you can rapidly switch from one type of production to another type of production. Let’s say hydroalcoholic gel. What do you need? You need a plastic bottle, a plastic pump, which is pretty much the same as what you need for a large variety of products that have nothing to do with hydroalcoholic gel. You also need some very basic compounds like alcohols and water, which are very basic things used in entire series of industries. So, a question is, if you’re in these industries, even if you’re not producing this type of product, can you start producing them almost overnight? And that’s also something very interesting in modern supply chains is that frequently, IT systems and software are slowing down processes. Over the last couple of decades, supply chains have gradually become more rigid, frequently because of software setups.
Kieran Chandler: With many e-commerce companies, it’s actually faster to move stuff from one warehouse to another physically as opposed to reconfiguring the IT system or the ERP. In which case, moving stuff from one warehouse to another can be done literally in two days, while switching from one warehouse to another with regard to the ERP can take six months. Agility is frequently lacking nowadays. It’s not at the physical level that agility is weak; it’s literally at the decision-making process, and especially when facing tail risks that start to unfold. What I see is that most companies, even when they gradually start to acknowledge that the problem is real, they are still very slow to actually implement anything, even when it’s very obvious. There are very obvious solutions. Can you guess what I’m thinking of?
Joannes Vermorel: Well, what we’re explaining here is that by having a leaner, more agile supply chain, it’s not about forecasting which of these scenarios might occur, but it’s by putting yourself in a position where you’re able to react quickly. That’s the thing we should focus on.
Kieran Chandler: Yes, moving on to the next topic. One way of reacting is by adjusting pricing and the pricing mechanism. Why is that so beneficial?
Joannes Vermorel: Pricing mechanisms are interesting because when you read the mainstream press and see what governments are doing, they are literally enforcing price controls. This has happened in France and California, with laws against price gouging. I believe this is a misguided approach that will only make things worse. Prices are a very important element for markets to work, and pricing is a signal that helps companies redirect their efforts where things need attention. When people see prices skyrocket, the instinctive response is to enforce price control. But the problem with this is that you remove the incentives for companies to produce more and remove the shortage altogether. Nowadays, there is no reason why we can’t produce enough of a product, like hydroalcoholic gel, for example. It’s a super simple chemical product, and the plastic container is also incredibly basic. If we had to produce one bottle per person per day, it wouldn’t even stretch our capacity worldwide. But the problem is that producers are facing a temporary situation. Are they willing to invest in more machines that might only be needed for a few months? Having a steeper price is a way to fund the resource investment for those companies. I believe that in the current crisis, where retailers have been failing, is to implement smart pricing policies to avoid disorganization. For example, with toilet papers, there are super simple ways to deal with the situation. You could say that the first pack is at a regular price, but you need to have your loyalty card. If you present your loyalty card, you get a discount on the first one, and then the second one is at a higher price.
Kieran Chandler: Like instead of being two euros per pack, you go to 20 euros, and then suddenly people cannot rush anymore and just, you know, buy with 200 euros the equivalent of four months or six months of regular consumption for a regular individual. So, this is an example of stopping people from stockpiling by using those kinds of loyalty cards. What do you reckon the customer response would be to people actually limiting what they could actually buy? I mean, we are heading towards a Soviet state kind of example.
Joannes Vermorel: The problem, I mean, obviously my advice would be, it’s very reasonable in this sort of situation for people, especially people with families and kids, to stockpile a little bit. It’s reasonable. Obviously, I would not think, I don’t think that stockpiling for the next six months is an ethical response. You’re creating a large problem for the rest of us. Nonetheless, I think it’s very important to avoid relying on wishful thinking. You cannot have a plan that relies on the fact that everybody is going to be honest. That’s not how the world works. Whenever you want to do something, you need to expect that we are talking about humans, and there will be adversarial behaviors. People forget about that. It’s not that all humans are bad. The statistical reality of societies, and pretty much all societies actually, is that the overwhelming majority of people are honest. Otherwise, if you had more than a tiny fraction of a percent of people who are murderers, societies would collapse. The level of violence would be unbearable. So, the vast majority of people are very honest, disciplined, and actually acting for the good of their neighbors.
Kieran Chandler: You mentioned pricing being one of those potential solutions to this. I think it’s the most obvious short-term solution, and it’s very desirable. So what are some of the maybe other less obvious solutions that could be beneficial in these scenarios?
Joannes Vermorel: Less obvious solutions would be to directly decide, for example, to stop producing certain stuff. When you see that there will be a drop in demand, let’s say I am a company that is producing plastic for plastic bottles, plastic containers, and I can use that for shampoo and for hydroalcoholic gel, if we come back to the same example. Well, at some point, you need to take a very steep decision and say, “When should I actually stop producing shampoo and just say I’m going to use the same plastic bottle and just put hydroalcoholic gel in that?” Yes, it’s not going to be perfect. It’s not exactly the right container, but it can work. So that’s why, that’s the sort of things where you need to be ready, be imaginative, and you need to think about ways where you can potentially redirect production to things that matter most. And again, it boils down to supply chain agility, and frequently nowadays, I would say it’s software agility, because all those things, the problem is very frequently not at the physical level, but we are literally at the software level.
Kieran Chandler: So, to conclude our discussion today, the main lesson about tail risk is that it is a scenario that will inevitably occur. The best way to protect ourselves is by being as flexible and responsive as possible in our supply chains.
Joannes Vermorel: Yes, and again, a shameless plug, but I think it makes sense. These are the situations where disaster mitigation really matters. Having programmatic capabilities to rapidly implement approximately correct strategies is crucial. If your ERP is rigidly geared towards safety stock policies that take forever to update, or ABC analysis that becomes meaningless because it reflects the past, then those tools are completely unsuited to deliver a fast response. You need to have programmatic expressiveness.
At Lokad, we have been building systems that allow companies to the extent they can, do something and develop non-standard, non-mainstream plans that can potentially revisit entire strategies within hours. That’s why our four pillars are: forecast all possible futures, look at all possible decisions, prioritize everything in terms of dollars or euros of value, and robotize everything so that you can recompute everything in one hour.
The reason you want to be able to recompute, reprocess, and reoptimize everything in one hour is that, if there’s a big disruption taking place in your supply chain, you need tools that can roll out a completely revised strategy, even if it’s fairly approximate, literally within hours. Nowadays, most companies are years, or even light years, away from these sorts of things. Thus, whenever they face an existential disruption, it can lead to bankruptcy. They frequently lack basic response tools and resort to things like Excel, which are very crude when you want to execute with a certain degree of reliability, rapidly and at scale.
Kieran Chandler: Okay, we’re going to have to wrap it up there and probably go join those queues in the supermarkets that haven’t quite implemented these strategies yet. So, that’s everything for this week. Thanks very much for tuning in, and we’ll hopefully see you next time. Goodbye for now.