Value Chain Disruption and Resilience

[00:00:00] H.J. James Choo, PhD: Now we actually come to something that we all live and are affected by it, which is the supply chain or your value chain, as it’s actually titled here. Phil, I’ve known Phil for about 27 years and he’s just reminded me we were both actually five years old then maybe six years old. Now I met Phil when I was actually doing my doctorate at Berkeley.

[00:00:29] H.J. James Choo, PhD: He actually ended up serving on my dissertation committee, and he actually taught the first industrial engineering and operation science course that I ever took. Somewhat of what I do, it’s his fault. Okay Phil is the senior principal scientist at Amazon, where he leads transportation marketplace science team.

[00:00:51] H.J. James Choo, PhD: And the Earl Isaac Professor Emeritus in the Science and Analysis of Decision Making in the Department of Industrial Engineering and Operations Research at UC Berkeley. Please welcome Phil and his interesting topic.

[00:01:04] Phil Kaminsky, PhD: Thank you very much.

[00:01:10] Phil Kaminsky, PhD: Let me start with the disclaimer. I’ve been working for a long time, 30, 40, some large number of years on, supply chain management issues primarily for, in a research capacity at Berkeley with lots of companies. Most recently a lot of companies in the biotech, biopharma space, other manufacturing companies.

[00:01:37] Phil Kaminsky, PhD: And and then of course for, the past four or five years also, at Amazon. But I don’t know, I don’t know that much about construction or, capital project management. I think every few years these guys pull me out. I To talk about what I do know about and hopefully people in the audience are smart enough to, find some connections or, some new perspectives that, that you can apply to the worlds that are that you live in.

[00:02:12] Phil Kaminsky, PhD: I, I think so that’s, how I think is I work a lot in, in, in product and supply chain management now and in retail supply chain management. Hopefully there’s some some lessons that you can, draw on. So the topic that I’ve been most interested in, both for the last five to ten years in my research, and also it’s been a huge focus in what I do in Amazon, is broadly thinking about risk management in supply chains.

[00:02:50] Phil Kaminsky, PhD: How you think about risk, What you can do about risk how you measure risk and, like I said not so, much in, in, the industries that, that many of you are part of but, again hopefully, there’s something you can take away from it. In general, just to set the stage because maybe my background is a little bit different than yours.

[00:03:16] Phil Kaminsky, PhD: When I think about supply chain management and the way it’s evolved over the past, you 20 to 30 years, there’s two buckets that I divide things into. There’s some stuff we’ve been doing for a while that we continually refine and try to, get better at. I’ll talk a little bit about that. And then there’s the stuff that we’re really focusing on now, the stuff that I worry about day to day at Amazon, the stuff that the other companies that I talked to and work and work with worry about.

[00:03:48] Phil Kaminsky, PhD: And these things. Are things that tend to be around data, visibility, AI, lots of AI, and more than anything else lately, risk management. We’ll talk a little bit about why that is, but before we do that, I just want to set the stage. For 35 years, for 30 35 years, since we really started talking about supply chains, we’ve been focusing on a few different things.

[00:04:24] Phil Kaminsky, PhD: We’ve been focusing from a strategic level on cost minimization, on matching supply and demand, on thinking about how do you most effectively design your supply network. We build giant optimization models. Still, I spend a chunk of my time every day at Amazon thinking about building big optimization models to figure out where in the world we build, we put different kinds of facilities to make the supply chain most effective.

[00:04:55] Phil Kaminsky, PhD: We’ve been thinking about inventory management. Exactly how much stuff do we keep where? I’ll talk a little bit more about that later. But even though we’ve been talking about this for a long time, we continue to get better at it. We continue to find new tools to make it more effective. We focusing to a large extent on, generative AI tools lately and how we can think a little bit differently about inventory system design, using those kinds of tools.

[00:05:24] Phil Kaminsky, PhD: We’ve been thinking about supplier selection. We’ve been preaching, I think I’ve done it here before, preaching a little bit about strategic partnerships and how the contracts that you structure in relation, in relationship with your partner drive decision making and enable us to, globally optimize supply chains.

[00:05:41] Phil Kaminsky, PhD: And we’ve been talking for a long time about pricing and the effects of pricing and how pricing makes your supply chain more effective. And we’ve been talking about from an implementation and execution point of view How do you use tools to make, to optimize decisions, right? How do you, it’s easy to wave your hands and say whatever.

[00:06:02] Phil Kaminsky, PhD: Inventory’s bad and decrease lead times. And it’s a whole other thing to have tools and models that let you think about how to do it. We’ve been thinking about how do you design a product so that you make your supply chain more effective. How do you, for example, postpone differentiation of a product so that you can take advantage of Aggregate forecasts and the increased accuracy in aggregate forecasts.

[00:06:28] Phil Kaminsky, PhD: We were just talking about it in the in the in the previous talk, right? How do you have semi finished goods so that you can deal with the variability in the details of the final ordering? We’ve been talking about this for a long time. It’s still a very important thing. How do we minimize bullwhip or variation in supply chain?

[00:06:47] Phil Kaminsky, PhD: How do we execute against supplied contracts? How do we do sales and operations planning? How do we And these are what I would consider the baseline tools of supply chain management. Everybody certainly who’s taken a course from me, hopefully James and, many people since then and, or who’ve worked with my book are familiar with these kinds of tools.

[00:07:15] Phil Kaminsky, PhD: This is, I think, nothing new, but what we have been thinking about really is, like I said, data visibility and risk management, but even beyond that. I would say that people over these years have started to use these tools, supply chains became leaner and more effective, and maybe not surprisingly but as a consequence of that, supply chains actually became riskier, more brittle.

[00:07:47] Phil Kaminsky, PhD: We lowered inventory to a large extent, we decreased lead times, we got rid of all the waste in the supply chain And these kinds of decisions, they had downsides. And so what we, like I said, what we’ve been really focusing on more recently is managing risk in the supply chain using data using visibility.

[00:08:11] Phil Kaminsky, PhD: And to expand on this picture that I told a second ago if, we were focusing strategically 25, 20, 30 years ago, I realize I don’t have the same times on all the slides, you get the point, a long time ago, before most of you were born basic elements of optimization.

[00:08:34] Phil Kaminsky, PhD: More recently, we’ve been thinking about how do you do this in a dynamic, responsive, adaptive way? How do you design systems so that your supply chain can change as the inputs change? How do you think about segmenting your supply chain so that instead of having a single supply chain to meet all the needs of your of a broad set of customers.

[00:08:57] Phil Kaminsky, PhD: You can think of an overlapping set of supply chains to meet different needs. We’ve been thinking about like I said, risk management. And from a sort of implementation and execution perspective, we’ve been thinking about how do you do all the things we used to do in this world where you have huge amounts of data.

[00:09:20] Phil Kaminsky, PhD: How do you take advantage of testing? So a lot of what we do at Amazon, a lot of what I do at Amazon day to day is, I come up with some theory about how something’s gonna help some system work. Nobody really knows if it’s gonna happen. Maybe it’s nice to build a simulation. Nobody really believes the simulation. So a lot of what we do is thinking about how you roll out and test different things in a real world production environment.

[00:09:50] Phil Kaminsky, PhD: How do I try different delivery strategies? How do I try different ways of engaging? So a lot of what I do is engaging with all the transportation providers that move the stuff in the Amazon network. We move as an aside, we move something like 150, 000 to 200, 000 truckloads of stuff every week in, the U.

[00:10:12] Phil Kaminsky, PhD: S. We don’t own any trucks, so we have to, roughly speaking, we own a few trucks, not so many trucks. So we, have to spend a lot of time engaging with tens of thousands of small carriers every day to move all the loads that we need to move. The big, the tractor trailers that you see on the highway. And so, we have different ways of engaging with these carriers, which if you want to put me aside later I, could speak for hours about.

[00:10:42] Phil Kaminsky, PhD: All Probably not so relevant to the folks here, but the point is that we have to test these ways of engaging with carriers, right? We don’t quite know. You can’t just on the basis of a model and a, simulating environment tell what’s going to work and what’s, not going to work. So we spent a lot of time thinking about how do you effectively test different things in the real world.

[00:11:04] Phil Kaminsky, PhD: And then we think about robustness, designing systems for robustness, right? So if our focus 25, 30 years ago was how do you minimize cost, now we’re starting to ask this question. In fact, we put a lot of effort into answering this question. Maybe what if you pay a little bit more? What does that get you in terms of decreasing the risk in the system?

[00:11:30] Phil Kaminsky, PhD: If you to call on the transportation example. If you change your mix of long term contracts in the spot market, for example, it’s going to cost you a little bit more, maybe, but maybe you reduce your risk. How do you talk to leaders about what the right measurements, the right ways to characterize risk are?

[00:11:54] Phil Kaminsky, PhD: And then again, how do you not focus on a single strategy but, design a system where the strategies adapt, for example, when costs go up or down, demand shifts in the marketplace? And I think I have a few more details on this. And so really this is the key world in which we’re operating in now.

[00:12:16] Phil Kaminsky, PhD: We’re operating in a world where, for different sets of customers, we actually think about designing different supply chain networks. For cost sensitive customers in certain parts of the world, we think very differently about how we design, for example, at Amazon, how we design the, distribution system.

[00:12:35] Phil Kaminsky, PhD: then for business customers. For certain kinds of expensive goods, we think very differently than how we design the distribution system than for groceries. And we’ve learned back in the day we used to think there was one bright solution. We’ve learned that we need multiple solutions and yet we need to figure out a way to make to leverage overlaps in these multiple solutions to lower costs.

[00:13:04] Phil Kaminsky, PhD: We need to think about the same thing for differentiation. You can think about it as product design strategies, right? Maybe you want to, again, calling on this, last presenter, for certain customers, you want to maybe have more finished goods inventory. For certain customers, it makes sense to have more semi finished inventory.

[00:13:27] Phil Kaminsky, PhD: For others, it you, you maybe want to design a different kind of supply network. You maybe want to have different kinds of, contracts, right? Todd talked about getting the, was it VPs or CEOs together in in the start of your discussion. So these are the sort of the informal way of, managing those kinds of relationships.

[00:13:52] Phil Kaminsky, PhD: But I’ve spent a lot of time in my career thinking about the formal ways of structuring contracts. So that you can achieve this, kind of alignment in supply chains. And, we continue today to think about there’s another big focus of what I do today, day to day, is thinking about how we design contracts with multiple companies so that we can get the supply, decision making in the supply chain to be as if it were a single company making these contracts.

[00:14:21] Phil Kaminsky, PhD: These kinds of decisions again, always with this kind of overriding question of how you do this in a way that still continues minimizing cost. And the way you do this, very roughly speaking, is it’s easy to go to pair me or somebody talk and you look at, some four by four grid and you sometimes you do this and you sometimes you do that.

[00:14:49] Phil Kaminsky, PhD: But ultimately, we use huge amounts of data, we run big simulations we do all kinds of analysis, we have big optimization programs that help us to set precisely set the, parameters in, these kinds of contracts. And then to implement them, we have huge amounts of of data. Same thing in designing networks that are dynamic or adaptive or responsive We update our prices every every few seconds. We mix our supply, change our supplier mix dynamically every day. We work to reduce lead times. We do all of these things using vast amounts of data and using visibility and using models like I think conceptually. All these things are easy to understand.

[00:15:53] Phil Kaminsky, PhD: In practice, what we have is lots of data, and lots of analysis, and lots of models to make these things work. I say us, but really I think this is how we think about modern supply chain management. There’s lots of data, lots of analysis, a couple simple concepts, but understanding that really to drive the value out of these concepts, we need we need tools.

[00:16:17] Phil Kaminsky, PhD: With that context. Let’s think about, a little bit about, specifically about risk management. First thing I want to say when I talk about risk management, we talk about, at Amazon and in the bio, with the biotech firms I work with and other firms I work with, we talk about risk a lot.

[00:16:38] Phil Kaminsky, PhD: And the first thing that becomes apparent is everybody needs something a little bit different when they talk about risk. And it becomes really important to understand when you’re having a conversation what you’re talking about. I think the kind of most useful initial framework is this unknown versus known unknown framework.

[00:17:00] Phil Kaminsky, PhD: I think most people have seen this before. Seems obvious, right? There’s some very no very well known understood risk, which is variability that you can quantify very, clearly. And then there’s, on the other end of the spectrum, a bunch of stuff that only happens, which is So infrequently that it’s as if you have no information about the likelihood that will happen.

[00:17:30] Phil Kaminsky, PhD: What’s the chance that a tsunami will hit central California? Particularly for the people who travel travel here once every few years, the odds are, that you are, will be here when there’s a tsunami warning, almost zero. And yet, somehow at some point the wave is apparently gonna come crash, crashing over the hill.

[00:17:58] Phil Kaminsky, PhD: But all it’s not even really jokes aside, I was just, I, my, my neighborhood in Berkeley was evacuated, we had an evacuation warning this morning. So this is, this was a, serious thing, I don’t mean to, make light of it, But these are the kinds of things that, that clearly you need a different set of tools to manage risk in the, supply chain than you, you do for the, fact that demand is plus or minus 20 percent on any given day.

[00:18:30] Phil Kaminsky, PhD: But there’s also a spectrum in between, right? There’s a, maybe demand is plus or minus 20%, but maybe there’s a significant shift in demand. Maybe there’s a significant shift in, in marketplace costs. And so when we think now about managing risk in a supply chain, it’s useful to keep in mind that you need a different set of tools to manage different kinds of risk.

[00:18:55] Phil Kaminsky, PhD: But you look for overlaps in these tools because this, sort of risk continuum is, continuous. It’s not just that they’re it’s not just that they’re two kinds of risk but, various kinds of risk where the kind of degree of uncertainty And what we call in academia the, model uncertainty changes, right?

[00:19:17] Phil Kaminsky, PhD: In some cases, you know the distribution of underlying uncertainty. In some cases, you don’t know the distribution, but you have some sense that there’s a stationary distribution there. In some place, in some cases, you just don’t know what the hell is going to happen. You just don’t know that there’s no useful model of this of this, tsunami hitting us.

[00:19:42] Phil Kaminsky, PhD: Which was particularly interesting because I got to Amazon right when a bunch of stuff that dramatically impacted our business happened. COVID, and then Ukrainian conflict these things impacted lots of firms. Significantly impacted us at Amazon, right? In the in the, stores business.

[00:20:07] Phil Kaminsky, PhD: I know there’s some folks here from the from AWS. I probably should have clarified this at the beginning. Maybe it wasn’t clear. Where I work is we get the packages here. We deliver smiles, we like to say. Particularly this time of year. Particularly. This week, in fact.

[00:20:28] Phil Kaminsky, PhD: And there were all kinds of things we weren’t expecting. Workers weren’t available. Demand forecasts changed dramatically, right? I don’t know if you can remember, there was a time when suddenly everybody was desperate to get groceries delivered and you couldn’t get groceries delivered. And there was no toilet paper in the world.

[00:20:49] Phil Kaminsky, PhD: It seems like a long time ago now. It actually wasn’t it wasn’t that long ago. But these are the kinds of things that we had. The market rate for transportation dramatically changed overnight. Transportation got really expensive, then really cheap, then really expensive. We, have talked about this a little bit before.

[00:21:12] Phil Kaminsky, PhD: You couldn’t put two people in the same truck, driving the same truck, unless they were related at some point. And so it became very hard to drive stuff across the country. So all these kinds of things that we hadn’t really modeled. And we we reacted to in a way that involves spending a lot of resources to try and as quickly as possible, build up systems to deal with it.

[00:21:40] Phil Kaminsky, PhD: Which led to huge expenses after the pandemic was over. And, frankly led to a lot of this careful thinking about risk that we’re, doing now as we think about about addressing these kinds of things. So I talked about these are some of the changes, talked about how we reacted, how we thought of reacting.

[00:22:01] Phil Kaminsky, PhD: So people have been talking about risk for a long time, but often it’s in this kind of old fashioned kind of way. If you’re, if you got an MBA maybe a few years ago, you probably saw a grid something like this. Somebody would draw a, two by two matrix. They would say there’s some stuff you can control.

[00:22:26] Phil Kaminsky, PhD: There’s some stuff that has impact. You put everything somewhere on the grid. And the stuff you can control that has high impact, you invest in. And the stuff that you can control but has low impact, maybe if it’s expensive, you don’t bother investing in. And if there’s stuff you can’t control and has high impact.

[00:22:44] Phil Kaminsky, PhD: I don’t know, and the professor would change the subject and move on to the next topic. Obviously this makes intuitive sense, but really for 7, 8, 9 years now, we’ve tried to take a much more quantitative view of risk management in supply chains. Now what we’ve often done in my profession, and I’m, by training operations researcher, you like to take the impact of an event times the probability that event will happen, you figure out some expected impact of the event, and if that number seems high, then you do something about it, and if that number seems low, then you don’t do something about it.

[00:23:36] Phil Kaminsky, PhD: But it’s never quite clear exactly what that number means. If there’s something that has a 10 percent chance of happening, and it has a million dollar impact for you. What does that, hundred thousand dollars mean? It’s never going to cost you a hundred thousand dollars. So what we started thinking about was supply chain disruptions.

[00:24:00] Phil Kaminsky, PhD: Specifically the kind of mental model that we, think, when we think about managing risk, managing disruptions in a supply chain is what happens if you take out part of the supply chain? For some period of time. What’s the impact of that on the operation of your system? What happens if you I don’t know, can you see the mouse here?

[00:24:23] Phil Kaminsky, PhD: No. No. Maybe I’ll point then. What happens if you pull out some supplier’s plant for six weeks from your supply chain? What’s the impact? And you can, for every node in your supply network, you can Depending on the sort of the nature of your business, there are different ways of measuring this, but you can get some notion of the impact on your supply chain if that node is gone for some period of time.

[00:24:51] Phil Kaminsky, PhD: It doesn’t matter so much to you if it’s a fire, or a strike, or a tsunami. You can start to assess the impact of, removing nodes in your supply chain on your financial performance or maybe some other performance metric. You can do this a node at a time. We spend a lot of time building tools to understand so this is an example of a a, biotech supply chain, one of the biotech companies we work with, where we did this kind of an excuse me, this analysis, we build models, we try to understand, you take some node out of the supply chain for some period of time, given the standard inventory that’s kept throughout the supply chain, How long before your customers see the impact?

[00:25:41] Phil Kaminsky, PhD: What’s the impact of, doing this? We ran all kinds of simulations. You get a whole bunch of graphs that look like this. You try to understand what kind of service level, where service level depends on the business. Maybe it’s lead time to, to fill out some gap in the inventory. Maybe it’s percentage of in stock.

[00:26:06] Phil Kaminsky, PhD: If a particular node is down for some period of time And you’re operating a supply chain in general with some cost, some standard cost. What happens to the service level while that, node is down? So in other words Stock of, one thing we actually saw, One of the, one of the, one of the biotech companies we were working with, Particular kind of container The, there was a issue in a plant of the manufacture of that kind of container.

[00:26:33] Phil Kaminsky, PhD: It really wasn’t a very expensive component. But depending on the inventory that the firm maintained of that component for some period of time they, maybe be two weeks or three weeks or five weeks until the end customers started seeing the impact of this component being gone.

[00:26:53] Phil Kaminsky, PhD: And then depending on how long until the the, container man, manufacturer got up and running, you can get some assessment using these kinds of tools of the impact of this node being missing on the supply chain. And then what we could do is we could build a graph. Something like this for every for every core supplier of this company.

[00:27:14] Phil Kaminsky, PhD: And for a bunch of different potential operating costs. So different levels of, effectively finished goods and in process inventory that they, kept. Another project that I’ve had, did some work with is a similar analysis at Ford. I was a little bit involved in this project where we looked at an even simpler kind of analysis.

[00:27:35] Phil Kaminsky, PhD: Just ask the question you can see it here for a whole bunch of suppliers, what’s the total spend, spend of the supplier? And how many vehicles would be affected if that node was gone for some period of time? That supplier was down for a month. How many vehicles would be affected? And, graph that with, how much you spend at the supplier.

[00:28:02] Phil Kaminsky, PhD: And you see this. This, interesting set of suppliers out at the end there circled in red, where you’re not spending a whole lot of money, they really will dramatically impact your business. And these are important quantitative insights that help you think about managing risk in the supply chain, but to do this, you really have to put in the effort to understand the impact on each of your suppliers.

[00:28:27] Phil Kaminsky, PhD: Here you can see it even more now the, vehicle’s gone or been translated into financial impact. You can see there’s a whole bunch of suppliers that Ford is spending very little money, that will cost them a whole lot of money, a whole lot more than they’re spending at this supplier, if something were to, happen at this supplier.

[00:28:47] Phil Kaminsky, PhD: So we spent a lot of time working with different companies, and we spent a lot of time at Amazon thinking about how do you go from risk being this vague concept that leaders are concerned about, to things that you can put a quantitative measure on, And then once you do that the question becomes how do you manage this risk?

[00:29:09] Phil Kaminsky, PhD: So I think this sort of, again, two ends of the spectrum, there’s this sort of well understood risk. This is, I think a, kind of risk that, that folks in here are pretty familiar with, right? This is risk that aligns with variability or maybe well understood uncertainty of various kinds.

[00:29:31] Phil Kaminsky, PhD: And as we talked about before today, you hedge that with buffers. Todd and I were just debating. I believe there are four buffers, not three. I don’t know if this is some sort of PPI sacrilege to add to the buffer list. I don’t know if I’m, I don’t know if I’m allowed to do that. But in, I’m going to go ahead and do that anyway.

[00:29:55] Phil Kaminsky, PhD: The buffers are, everybody said before, time, inventory, and capacity. I think importantly, flexibility is a buffer too, right? You can maybe argue flexibility is a subset of capacity, but in my mind it’s a little bit different, right? There’s un, if there’s uncertainty in the kind of demand for a product, if you have a machine that can make all the kinds of products that there might be demand for, you have a buffer that you don’t have if you only have, if you have separate machines that make the individual kinds of products.

[00:30:27] Phil Kaminsky, PhD: I think everybody here knows about all these kinds of things. But By the way, when do I have to? We started late and I don’t know what the current schedule is. Are we, am I close to wrapping up? Am I not? Is there anybody official to tell me? Depends on the flexibility level. Yes, exactly. Oh,

[00:30:55] Phil Kaminsky, PhD: we have plenty of time. But it says it’s 3. 30. 3. 30. So I’m already at minus, I’m already at minus half hour. That didn’t feel like it.

[00:31:09] Phil Kaminsky, PhD: Say that again? Twelve more minutes thank you, that’s useful. Okay, got it. Again, yeah, we’ve, it’s amazing time went backwards, sideways. So I think the key is you have these four kinds of buffers, but still, and I think folks are, pretty comfortable with that.

[00:31:39] Phil Kaminsky, PhD: The key is using these buffers correctly. The key is understanding the trade offs associated with these kinds of buffers and the key is understanding if you’re on the efficient frontier, right? So, what do I mean by this? And this is a specific example from from this, biotech risk management project If you’re using inventory as a buffer you gotta be pretty sure that you’re getting the most buffer you can get for that expenditure of inventory.

[00:32:12] Phil Kaminsky, PhD: And for this you need tools. Yeah let me leave it there, happy to, connect more about that. But the point is, you can’t just arbitrarily add buffers. You need, the kinds of tools to let you know the impact of increasing time, the value of increasing time, where in the supply chain it makes sense to focus on decreasing elite time, where you want to add inventory because you want to be pretty sure that if supplier Y goes down for a month, which they, do every few years, it’s You can still meet your customer service requirements, those kinds of things.

[00:32:52] Phil Kaminsky, PhD: And then in terms of flexibility, the reason I like to talk about flexibility is because the interesting thing about flexibility is that a little bit of flexibility goes a long way, is generally the case, right? This is a classic example that I love. You can imagine that you have five plants, that make five products each plan is dedicated to a product.

[00:33:20] Phil Kaminsky, PhD: If the demand for product B disappears and the demand for product a doubles nothing you can do about it. Now, if you have five plants, each of which can make all your five products, then the worst thing happens. The demand for everything that goes, down to zero, except one goes up five times, you’re still covered.

[00:33:42] Phil Kaminsky, PhD: But it turns out that if you have just a little bit of flexibility, In most cases so here you can see this is, a, a famous flexibility example called the closed chain, which says each plant makes two of the products but it’s connected in a cycle like this. So you can imagine sliding all of your demand down a a manufacturing network like this.

[00:34:07] Phil Kaminsky, PhD: You can get most of the benefit of flexibility. So this is a sort of very specific manufacturing in plant example, But like this notion that a little bit of flexibility actually buys you a lot is true in many cases. And a useful thing to think about when you’re thinking about buffering against risk.

[00:34:29] Phil Kaminsky, PhD: Here is going back to the, Ford example. This is how we, talk with Ford about thinking about these kinds of things. For the low cost, but high impact suppliers have two of them. It just doesn’t make sense to to place so much risk in a, place that you’re not spending a lot of, money at anyway.

[00:34:59] Phil Kaminsky, PhD: For the, high cost, high impact suppliers, that’s when you want to think very carefully about relationships and partnerships. Thank you. For the low impact, high cost suppliers, then you focus on on controlling costs. So again, this all starts from thinking in a qua in a quantitative way about the kinds of risks that concern you, rather than a qualitative way.

[00:35:24] Phil Kaminsky, PhD: Now for unknown unknowns, just very briefly, there’s less you can do. You can’t do so many run simulations and, try to refine value at risk kinds of calculations. The tsunami’s gonna hit your in fact, we work with a plant Bayer the German Bayer, the biotech company that, the pharmaceutical company that has a biotech plant in Berkeley right at the water level.

[00:35:50] Phil Kaminsky, PhD: This would have been gone, or, I shouldn’t say would have been, that’s gone if this tsunami hits, right? It would have been gone today. It makes a third of the factor eight in the world the, hemophilic drug right, by the water there. To hedge against that risk, you need to think very different, very differently.

[00:36:14] Phil Kaminsky, PhD: It’s not like you can keep, you increase your inventory by 5 percent in a few places and you’ve hedged against that risk. You need to think in a very sort of concrete way about, about how to deal with this risk. In my mind, there’s three core ways to deal with these kind of unknown, risks.

[00:36:30] Phil Kaminsky, PhD: risks. First thing is just to know the risk is happening before everybody else does. Call this rapid sensing and response. The second thing is flexibility. If Bayer has another plant somewhere else that can make factor eight maybe they paid a little bit more for this plant, but this lets them shift production if, necessary.

[00:36:50] Phil Kaminsky, PhD: And the third thing is this notion of adaptive supply chains, which is really rapid response and redundancy and flexibility on, on, on steroids. And in the interest of time, let me go super quick. My favorite sensing and responding example is an old example, but I love this example because and this is, this shows that I’m older than most of you in the room.

[00:37:15] Phil Kaminsky, PhD: I used to love my Nokia cell phone. It was the coolest thing. It was like this little thing with the, you could type the little letters in it. And then one day, Nokia stopped making cell phones. The question is what, happened? No, this is I, told the story, I screwed up the story. I screwed up the story.

[00:37:38] Phil Kaminsky, PhD: Yes re, rewind. Ericsson, I used to have my Ericsson cell phone. I used the wrong, I had this little Ericsson cell phone with a little flip up I don’t know what, do you even call it. It was like a candy, bar shaped phone, but then it had a little thing that flipped. You guys remember?

[00:37:55] Phil Kaminsky, PhD: Nobody, but it wasn’t a real, it wasn’t like the Motorola flip phone. It just had the little, anyway. I love that phone. And then one day Ericsson exited the cell phone market, and the whole reason was because there was a, Philips plant in New Mexico that made a chip that Ericsson and Nokia used in their phone at a fire in this plant before Ericsson realized about, knew about this fire, Nokia had built up, bought up six months of the world’s supply of this chip.

[00:38:27] Phil Kaminsky, PhD: Ericsson couldn’t get it. They sold no more phones the rest of the year, and then they went out of the, cell phone business. So, one way to deal with this kind of risk is just to know about it. Another way is, flexibility. Quick case study. The idea here with flexibility is maybe it doesn’t maybe it doesn’t cost so much.

[00:38:51] Phil Kaminsky, PhD: So this is, again, very quickly a consulting job we did with With a, CPG company, had 40 manufacturing facilities in the world. We did a study to to figure out what the cheapest way for them to manufacture these goods were. There was a new design left no plants in North America or Europe, but minimize the cost and fully utilized all the remaining plants, but that left no flexibility in the system.

[00:39:25] Phil Kaminsky, PhD: And what we found is there was another solution. That used seven more plants, including a plant in North America, and a plant in Europe, that cost just a tiny little bit more. But left a great deal more flexibility and redundancy. These are the kinds of things you can’t You can’t tell without models to to, assess them.

[00:39:50] Phil Kaminsky, PhD: And then, oh, I’m going the wrong direction here. Finally, so just as an aside, I’m just up against time. It’s more complicated, leveraging this kind of flexibility is more complicated than just saying, hey, you can make stuff in different places. There’s often costs and time associated with moving production around.

[00:40:10] Phil Kaminsky, PhD: This is something we, we spend a lot of time modeling in the biotech industry. Happy to talk more about that. Offline, but the final and most powerful tool for dealing with this kind of risk in a supply chain is what I would call adaptability. How do you adjust the design of your supply chain itself to meet structural shifts, to meet disruptions, to meet changing demand patterns?

[00:40:38] Phil Kaminsky, PhD: And I think you do that by designing this ability to respond into your supply chain. Having multiple suppliers. Carefully about how you deploy given that you have a set of suppliers thinking very carefully about how you deploy those suppliers, how you use them, how you keep them all engaged at a certain time.

[00:41:02] Phil Kaminsky, PhD: How do you decide where to ship demand? And you do this to a large extent by not just treating them as arm’s length suppliers, but really work, working to share objectives to, to build engagement around a shared culture of risk and, trying to ensure that each of them are flexible and adaptable.

[00:41:26] Phil Kaminsky, PhD: So I wanted to end with this idea that Amazon has made a big deal about over the past year. This notion of how Amazon designs its distribution network we’ve made a big deal in the past year of moving to what we call a regionalized network in Amazon. So when Amazon first started when, about the time that James and I first, met each other a couple of big distribution centers that served the whole country The idea being you could pull demand and lower costs that way.

[00:42:04] Phil Kaminsky, PhD: And then over time as delivery lead times became more critical, Amazon moved to this sort of hodgepodge network of distribution centers all over the country. So when you ordered some stuff from Amazon, some of the stuff would come maybe from a near location, some would come from a far location.

[00:42:25] Phil Kaminsky, PhD: We spent a lot of time and effort Building algorithms that try to figure out the cheapest way to get everything to you. If something if the demand was very low in the country for a certain component, we’d maybe keep it in one place. As, the amount of demand increased, we would spread it out more, standard inventory theory stuff.

[00:42:46] Phil Kaminsky, PhD: And then at some point, the system like this became very hard to manage, and became very inflexible when demand shifted regionally. Or when transportation costs changed a lot which, they’ve been doing over the past few years. So, most recently we moved to what we call a regionalized network, where we divide the country up into a number of regions.

[00:43:07] Phil Kaminsky, PhD: I took this picture not from an official Amazon document, so I can neither confirm nor deny that this particular consulting firm got the regions wrong a little bit. But the bigger picture, that we divided the country into regions, and now almost everything you get distribution center within your region, which enables us to deliver things more quickly.

[00:43:33] Phil Kaminsky, PhD: Which is becoming increasingly important. Hopefully most of the people here get stuff in the day. Like you’re shocked at how much stuff you get, maybe not this week, but other weeks. You’re shocked at how much stuff you get quickly from Amazon. That’s because we’ve thought about, as the market shifted, as the demand changed, as our objectives changed.

[00:43:55] Phil Kaminsky, PhD: We had a supply chain that let us keep the same physical set of buildings, but redesign the underlying algorithms to utilize those assets in different way to, to react to challenges in the marketplace. I am way over, I think, so I’m going to skip the last few examples. I’m going to skip my final thoughts, I’m going to, I can’t do that so I’m just going to go to my conclusions, my takeaways which are as you think about risk management and supply chain solutions, the solutions have to be as adaptive and responsive as the supply chains themselves.

[00:44:37] Phil Kaminsky, PhD: The world we know is changing more rapidly, demand is changing more rapidly something is on, on some social media network, some influencer. Is pushing something and suddenly we, need to get an inventory of that all over the country. To have these kinds of adaptive solutions, you need data and visibility, but that’s not enough.

[00:44:59] Phil Kaminsky, PhD: You also need algorithmic tools, simulation, smart decision making. You need to build flexibility in the system from the start. As I said, it’s not good enough to have a big business school concepts. You need tools and models to actually decide on the actual numbers of things. that are kept places and lead times that I have in the system.

[00:45:22] Phil Kaminsky, PhD: And given all this and the complexities of the systems the way you save money is taking that next step and integrating these solutions. So if you have two separate supply chains for different sets of customers, to the extent that you can combine those supply chains in certain parts of the country that’s how you save money.

[00:45:40] Phil Kaminsky, PhD: So I’m gonna stop there. I actually have no idea if I’m over or under, but I feel like maybe over.

[00:45:50] Gary Fischer, PE: Holy cow, do I have any time left for me? Phil, you’re going to be around for a while?

[00:45:52] Phil Kaminsky, PhD: I’m going to be around, yes, so happy to take questions.

[00:46:02] Gary Fischer, PE: I encourage you to, you gave us a lot to think about there. I would encourage you to corner Phil after we, break today and cover him with your thoughts and questions. A lot of really good stuff.

[00:46:17] Gary Fischer, PE: Is it four or is it three? I’m looking for some advice here. Very good, thank you.