Subscribe
Share
Share
Embed
Increasing costs. Evolving consumer expectations. And accelerating disruptions. Today’s supply chain leaders face more challenges than ever before. Hosted by Karl Siebrecht, the Logistics Leadership Podcast features industry interviews, warehousing and transportation deep dives and the unique insights supply chain leaders require to compete and win in today’s environment.
[00:00:00] Narrator: It's the Logistics Leadership Podcast with Karl Siebrecht and Ben Dean.
[00:00:11] Karl Siebrecht: Welcome back to the Logistics Leadership Podcast. I am Karl Siebrecht. We are happy to be with you again. Ben Dean, it's good to see you again.
[00:00:21] Ben Dean: Yeah, happy to be on the podcast any day of the week with you, Karl, and I'm really excited for today's episode.
[00:00:27] Looks like we're talking about navigating uncertainty, but there's a little uncertainty in that title. What exactly are we talking about?
[00:00:33] Karl Siebrecht: That's a good one. I like that. So yeah, navigating uncertainty. It is, frankly, it's one of my favorite topics because it is so real. The world is an uncertain place.
[00:00:44] And particularly as supply chain professionals, it's really, really important to try and get as good of a grasp on predicting the future as we can. You know, I'm also, you know, full confession here, I'm kind of an economics geek and when I think about navigating uncertainty and supply chains, it really comes down to econ 101, trying to marry
[00:01:08] demand with supply to get to this equilibrium point. So what do I mean by that specific to supply chains? Demand is really about your sales forecast, your demand forecast, of course. On the supply side, it is about the capacity of your assets to meet that demand. So what are the capacity assets in supply chain?
[00:01:28] Of course, this is raw materials. This is inventory. This is transportation capacity. It's warehousing capacity. It's labor capacity. So how do we navigate the uncertainty of the future to build the exact right amount of capacity to make our supply chains work well and be extremely efficient? So that's what we're going to drill into today.
[00:01:50] Ben Dean: Great. It's one that I'm especially passionate about and I can't get my full forecasting episode in, so this will have to do. I think I've got a perfect guest for us today because I'm talking to Rick Watson, Founder and Owner of RMW Consulting, and he does this for a living day in and day out, helps supply chain specifically for brands and retailers to navigate uncertainty in their supply and demand equation.
[00:02:17] But I feel like we can't just dive into him. We've got to talk about decision making criteria and uncertainty, kind of at that academic level.
[00:02:26] Karl Siebrecht: That's exactly right. So that's what we're going to do. I'm going to queue up here a conversation with Kris Ferreira. She is an Associate Professor at the Harvard Business School, and her area of expertise is exactly this topic, sort of decision making and navigating uncertainty.
[00:02:45] She teaches a supply chain course to MBA students and also teaches analytics courses in their executive education program. So she's going to set a foundation for us to talk about approaches, to talk about technologies that can help do this better and just really how business executives broadly can best think about this dynamic.
[00:03:09] So let's dig into that. Kris, welcome to the podcast.
[00:03:13] Kris Ferreira: Thanks so much for having me, Karl. It's great to be here.
[00:03:16] Karl Siebrecht: It's great to have you. It's good to see you again. So let's just start, if you wouldn't mind, tell us a little bit about yourself and what you do.
[00:03:25] Kris Ferreira: Sure. Well, you gave a very kind introduction and kind of touched on at least half of my job here at Harvard Business School, which is on teaching, right, and supply chain management topics, as well as analytics.
[00:03:38] But maybe I can do a quick introduction on the other half of my job, which is around research. So much of what I do for research is, I call it at least like human AI collaboration space. And what I mean by that is, you know, a lot of AI tools over the last five, 10 years have been deployed as what I consider like decision support
[00:04:01] tools. So they're made so that they're giving recommendations to employees who are then tasked with deciding, you know, if and how to use these recommendations to make decisions. And it turns out that oftentimes, as one might imagine, these employees, their intuition might be in conflict with an
[00:04:24] AI's recommendation. And they have to decide, like, hey, do I go with my gut and trust my intuition or go with the AI tool's recommendation? And that's a really hard question to answer, right? The answer is probably sometimes I should go with my gut, sometimes I should trust the tool. But how do you know kind of when to follow each?
[00:04:46] And so a lot of my research is on helping these employees answer that question. So how can I help them kind of know when and how to combine, you know, their intuition, their knowledge, their expertise, with an AI tool’s recommendation. So that's what I mean by human AI collaboration.
[00:05:04] Karl Siebrecht: Fantastic. So what we'd like to do in this conversation is to explore the intersection of kind of two of the worlds you operate in.
[00:05:16] One is supply chain, supply chain management, and the other is algorithmic decision making. So, starting at a very kind of general level, in your research and also maybe in your teaching, is there sort of a general level framework that you have to apply to the broader realm or domain of algorithmic decision making?
[00:05:44] So you described one of the challenges is when to go with the algorithm, when to go with your gut? Sort of what are the parameters along which a manager, a leader, would think about that trade off?
[00:05:59] Kris Ferreira: Yeah, that's a good question. And if it's okay, Karl, I might even take it a step back and answer like maybe the first question that a manager should ask, which might be like, when should we even have an algorithm to advise on recommendations?
[00:06:14] Karl Siebrecht: That's great.
[00:06:15] Kris Ferreira: Because I think the second question is like the one you asked on, you know, the subset of problems, that you could make improvements in decision making with AI support. Then how do you kind of combine, you know, the human's expertise with the AI? But I think maybe one question, you know, when we take the supply chain lens that any manager, certainly supply chain manager, would have would be like, where might I use AI to help improve my decision making?
[00:06:44] And, you know, for that question, I tend to really think about, you know, AI or machine learning as it's really just predictive analytics, right? So it's all about making predictions. If you think about this in like a business context, like making predictions is essentially equivalent to reducing some uncertainty.
[00:07:06] And so I tend to advise managers or companies, like think through like, what are those biggest uncertainties that you face as a business that if you were able to reduce some of that uncertainty, you could make much better decisions? And you know, for supply chains, a lot of time, it's demand uncertainty, forecasting.
[00:07:29] So if we can better forecast and predict demand, we can reduce some of that uncertainty around demand and probably make a lot better supply chain decisions anywhere from, you know, how much capacity should we invest in, into how much product should we produce, you know, how we should, you know, get it from point A to point B, what we should price our end products at, what we should promise as our lead time, et cetera.
[00:07:54] So all of those decisions revolve around kind of good demand forecasting. And so if you can improve your forecast, reduce some of that uncertainty, you can make better decisions. That's a common one in supply chain. Now you can't always use AI or machine learning to reduce uncertainty and make better predictions, right?
[00:08:15] There's going to be, you know, some predictions or some uncertainties that it's going to be hard to use data and analytics to predict. And so, I mean, gosh, think about as an example of like COVID, right? We couldn't, you know, have planned for or predicted COVID. No one saw that coming. So instead of, in that case, using AI to reduce uncertainty and make better predictions, then you're in the space of trying to plan for the fact that you're going to face some uncertainties that you can't predict.
[00:08:47] And so how can you kind of plan to be a more responsive supply chain?
[00:08:51] Karl Siebrecht: Right. And so what conditions need to be present or should be present to sort of be a better fit for an algorithmic decision making? Maybe one easy one, I'm no expert, but like, do we have a decent amount of data that we can plug into the machine to give it a basis to sort of make a forecast or inform a decision?
[00:09:14] So one of the conditions is you gotta have a lot of data.
[00:09:16] Kris Ferreira: Yeah, yeah, a lot of relevant data. Yeah, it might be helpful maybe if I answer your question with like contrasting a couple of examples, and we can even stick in the world of like demand prediction, right? So if demand prediction is so tricky, like when could you use AI or machine learning for demand prediction?
[00:09:35] You know, you can think about like two different types of, maybe on the extremes, like different products and, you know, companies that are trying to predict demand. Like on one extreme, you could have, you know, say an established supermarket that's selling like a gallon of milk. And on the other extreme, you have, like a new fashion designer who's selling a very fashion forward dress that, you know, my
[00:10:04] MBA students might buy, someone who's more fashionable than I am. And you know, here you have, on the one hand for the gallon of milk, you do have a lot of historical data, just like you said, Karl, right? So you have a lot of historical data that's probably representative of, you know, what demand is going to be like for milk in the future.
[00:10:26] Whereas on the, you know, fashion designer case for that dress, you don't have that much historical data of that dress. And even if you did have some historical data, who knows if it's going to be representative of the future or not. Whenever you have, I guess, a famous athlete or, you know, person, I don't know my famous people.
[00:10:48] I'm not cool. Kim Kardashian. If that's still relevant, you know, wears this dress and posts on Tik Tok. Okay. Now demand is going to skyrocket. So now your historical data is not as meaningful. So it's kind of volume of data and how representative that historical data is of the future. And I think the other thing to compare on those two different types of products would be how easy is it to, or how feasible is it, to kind of characterize those products in terms of their attributes?
[00:11:21] Right? So if I, you know, had you close your eyes and I started describing like a gallon of milk, right? I could say it's a whole gallon of milk instead of a half gallon. It's whole milk instead of skim milk. You know, it's an opaque carton instead of a transparent carton, et cetera. And you would have a pretty good picture of that item in your head, right?
[00:11:47] I can pretty much fully characterize that product with concrete, quantitative attributes or categories in your head. Whereas if I had you close your eyes and I tried to describe this dress, you know, it might have many colors. And so it's hard for me to describe maybe what the color pattern is, what it might look like.
[00:12:08] And so I could maybe use words like, oh, fashion forward. You know, you're not going to have a clear picture of what this dress is in your mind. And it's hard for an algorithm then to have a clear understanding of how to relate a dress or a product to other maybe similar products if it's hard to quantify the attributes.
[00:12:28] Karl Siebrecht: That's a great example. Yeah, I get it. So you want lots of data. You want that data more likely to be historically representative of what might happen in the future. And that data you would also like to be specifically, even quantitatively, descriptive.
[00:12:45] Kris Ferreira: Yeah, exactly. Think like either a quantitative feature or variable or something you can group into a relatively small number of categories.
[00:12:55] Karl Siebrecht: Okay. Got it. So to apply this, you want to sort of first start in our business where there's a relatively high degree of uncertainty. And then within that pool, we would want to look for places where we've got data that's more representative, that's more specific, et cetera. And that would be a ripe area to potentially apply machine learning or an algorithmic decision making framework.
[00:13:20] Kris Ferreira: Yeah, exactly. You got it.
[00:13:22] Karl Siebrecht: So as we think about that, how does that apply to supply chain? We talked about demand forecasting, which is a key one. Are there other areas or other dimensions within the supply chain that either yield better or worse candidates for applying this type of management tool, algorithmic decision making?
[00:13:39] Kris Ferreira: Yeah, good question. There might be uncertainty in your company's supply, right? So how likely or unlikely is it to get certain supply and in what lead time, right? So uncertainty on lead time. There might be uncertainties around different aspects of production, right? So perhaps some of your resources or equipment might be kind of finicky and likely to fail.
[00:14:04] Can you make predictions around, like, when equipment might fail so that you could intervene and do predictive maintenance or have a better understanding of what your, you know, kind of output rate is at any given time? You might have uncertainties around labor, right?
[00:14:22] So one that might be hard to predict is our strikes, but nonetheless, like labor shortages, labor strikes, or even like labor turnover can be challenging for some supply chains.
[00:14:34] So lots of different examples of that. And then, you know, I would say within each of those, you know, we talked about looking for opportunities from an uncertainty perspective, but even given a particular uncertainty, so demand uncertainty, there's still many different decisions that you could make that are linked to that uncertainty to consider. For instance, you know, demand uncertainty plays a role, like I said before, right in terms of your investment and capacity decisions all the way downstream to your final pricing decisions.
[00:15:13] And so what you might do with better demand predictions or better X predictions, whatever it might be, there's often multiple layers of that, as well.
[00:15:25] Karl Siebrecht: So how should we think about time horizon? If we're trying to predict stuff, I'm simplifying a little bit, but we want to apply these decision algorithmic tools, AI as being one type, specific type.
[00:15:38] We want to apply that to kind of making predictions. How does time horizon of these predictions factor in?
[00:15:48] Kris Ferreira: In a huge way, in a nutshell.
[00:15:51] Karl: A bit of a leading question.
[00:15:52] Kris Ferreira: Yeah. I mean, even just to stick with that same last example, right? The, you know, investing in capacity, you often need to do years in advance.
[00:16:03] At least months in advance, depends on what it is, versus setting a pricing decision, which gosh, if you're doing like an online direct to consumer approach, like you can change the price immediately.
[00:16:19] Karl Siebrecht: Right. And then change it back.
[00:16:21] Kris Ferreira: Right. And so, you know, it might sound so obvious, but I see a lot of companies get this wrong, you know, whenever you're coming up with a demand prediction model or using, you know, AI to predict demand, you can only use information available to you at the time that you would need that prediction to make a decision.
[00:16:42] For instance, if you have to make an investment decision years in advance, you can only use predictive features or information that would be available years in advance. And so that cuts out a lot of possible information that you might have, like, what is the competitor pricing this product at?
[00:17:01] Well, you have no idea two years in advance, right? You don't even know what your competitors are, necessarily. Whereas for a pricing decision, like you would know in real time, kind of what a competitor's price of a similar product might be. And you could use that as an input to your prediction. And so because of this, naturally, you're going to have a lot more variation or uncertainty around your demand predictions.
[00:17:22] The longer that time horizon is from the point where uncertainty is realized, right? Your sale is made. Your demand is realized.
[00:17:32] Karl Siebrecht: Got it. And so with that, have you had experience with businesses or in your research, or even do you just have ideas around, there are areas where there's higher and lower certainty, there are time dimensions as a driver of the better or worse ability to predict?
[00:17:50] What are the implications on the actual planning process that you engage in along those differences?
[00:17:58] Kris Ferreira: Yeah, it's a good question. And it might depend a little bit what you mean by planning process, right? Like if you incorporate kind of AI predictions in terms of a plan, or if it's more planning to be responsive. I personally like to think about both. Going back to kind of an earlier point on, you know, there's uncertainties that exist that you can kind of reduce and better predict.
[00:18:24] And then there's some that you can't. You know, when it comes along, you know when it comes, in terms of this like time horizon and planning process, typically you'll have more uncertainty the longer the time horizon is. So typically those decisions, like big investment decisions, are going to be harder to predict demand for and thus lend more towards, like, how can you plan to be responsive in these settings?
[00:18:56] And so, you know, there's ways that you can set up your supply chain or say, we'll stick with investment decisions such that you can be more responsive, right? So can you predict kind of a range of, you know, future demand that you might have and maybe build out your capacity to the lower end of that range and kind of use more flexible capacity for the upper end, right?
[00:19:20] That'll allow you to be responsive without having to risk having low utilization on a huge, you know, kind of investment early on.
[00:19:30] Karl Siebrecht: Perfect. Yeah, there was an executive I was talking to just a week or so ago that talked about, in the supply chain planning process, this was a big, it was a large retailer.
[00:19:39] They were working to sort of build more of an approach that explicitly identified areas of known unknowns. You know, it's kind of like the Donald Rumsfeld thing from years ago. But really to get clear eyed and clear about where do we know there are unknowns?
[00:20:01] Kris Ferreira: Right.
[00:20:02] Karl Siebrecht: We're trying to apply tools and analytics to predicting lots of stuff. In some cases, you know, not all cases will be equally applied, for the reasons that you articulated earlier.
[00:20:15] So if we can be clear about the known unknowns and then figure out what to do about that, how to operate, how to execute in that environment in a different way from where we have a higher degree of confidence that we can get the right answer.
[00:20:30] Kris Ferreira: Yeah, absolutely.
[00:20:32] Karl Siebrecht: Got it. So the other thing to sort of explore a little bit here.
[00:20:36] So algorithmic decision making that requires lots of relevant, clear data to apply the tools, I'm paraphrasing, right? So, supply chains by their very nature are chains. And along the chain, often there are different actual companies and there are handoffs. So if we, you know, simplify a supply chain, it’s all the steps that a product has to travel or the raw materials through to the product all the way through to the end consumer.
[00:21:06] You know, it's going from raw material provider to a manufacturer to a transportation provider, maybe to, you know, who knows, wholesaler to retailer, and these are different businesses, each of whom are collecting or have some data. Some of that data may be relevant to feed, you know, to feed the machine, but it lives in different places. Can you help us solve that problem?
[00:21:33] Kris Ferreira: And I would say even sometimes, even for companies that are vertically integrated, you still have all of these pieces often operating fairly separately. I would say your description probably holds for the vast majority of companies or supply chains out there. Yeah, you know, it's tricky.
[00:21:58] There's some interesting research. And also I see this over and over again in practice of like, obviously each piece of the supply chain, each node, if you will, so from a raw material supplier being a node, you know, your manufacturer being a node, your distribution center being a node, your retailer being a node.
[00:22:19] Each one of those is making, at a high level, they have to make a similar type of prediction and decision, right? What's my demand going to be from the next level? And how much should I produce or, you know, store, house to meet that demand? You would think that information from each of these supply chain parties could be very helpful to each other.
[00:22:47] And certainly, certainly that's true. However, I have yet to see so many supply chains kind of collaborate, if you will, in terms of sharing this information in a way to where all can make better decisions. And I think it's probably clear, but I think a big reason for this is that oftentimes you have different members of the supply chain incur
[00:23:14] risks depending on whether the forecast is too high or too low.
[00:23:20] Karl Siebrecht: I love that example. Yeah, even though everybody is trying to digitize their supply chains and make it easier to have clean data to share, that's a problem unto itself. We had a conversation earlier this season around the jungle of systems, technology systems, and platforms that most companies actually live with.
[00:23:42] And they're trying to sort of upgrade those, you know, evolve them over time. But there's a jungle that exists even internally, to your point, like even within the four walls of a corporation, sometimes hard to share that data. But this is a, this is an incentive
[00:23:55] dimension to the problem too, which can exist inside the company, as well, right?
[00:24:00] Each department has their KPI. And it's kind of local optimization maybe versus a global optimization of total cost or something like that.
[00:24:10] Kris Ferreira: Yeah, totally. And it also happens in cases where kind of multiple, say, departments are sharing a similar resource, right? So if you're trying to vie for a similar resource, you might over inflate to make sure that you have the right share.
[00:24:25] Karl Siebrecht: Yeah. Great. So back to AI a little bit. So you've been doing research in this space well before the sort of relatively recent AI bubble with the sort of advent of Chat GPT and the sort of consumer awareness of this thing. And to be fair, also I'm sure it seems like just in terms of capability, there's been a pretty steep curve of advancement.
[00:24:49] With all that you've sort of learned and seen over time and certainly you've been right in the middle of this last couple of years, what's your view on, is it real that the sort of core capabilities of these tools are improving such that we should expect to see the tools be better and be able to help us managers sort of make better and better decisions going forward?
[00:25:14] Is there a lot of reality here in addition to the hype just in terms of AI that you're seeing? Specific to supply chain.
[00:25:21] Kris Ferreira: Yeah, I think it's a mix of both reality and hype. And maybe going back to kind of my research a bit here. Part of what I'm trying to do is to educate the employees and managers.
[00:25:36] Kind of what are the limitations of AI that will always be limitations of AI, and in particular, what are the limitations of AI that the human experts, the employees, could improve upon, right? And so I agree with you that, you know, AI has been gaining obviously in height, but also in capabilities. And I think that's a good thing.
[00:26:03] Where I think the hype sometimes crosses the line is thinking that AI will eventually get better than human experts. And maybe that's true in some areas, but I think in general, there's some limitations core, just limitations of any AI tool that human experts would be able to make improvements on.
[00:26:23] And I think finding it, that's why I've dedicated my research to this question of like, you know, how can you kind of take the best of both worlds, take the best of what an AI prediction is gonna bring, but also bring what the valuable kind of intuition and expertise that the human has, that the AI can't have, to make a better decision than either one frankly could.
[00:26:48] I personally don't think it should be like a comparison between, you know, how good can an AI tool do versus how good can a human do at the same task or decision, but rather kind of adding, I always say, like a third horse to the race. How good can a human equipped with an AI recommendation do, and I think there's a lot of instances where that should be better, but it's a matter of how can you help that do better?
[00:27:15] Karl Siebrecht: Right. And the conclusions there are sort of similar to what we talked about before. It's like when these conditions are present, this is when it's more likely to be useful. Or you talked about, you said you dedicate a lot of your research to helping to teach employees and managers like when, kind of when and how to use these tools.
[00:27:35] Kris Ferreira: So most of what we talked about before on like, when would you use AI or not, was kind of like, when is AI an even valid or valuable recommendation? I would say, you know, if you're looking at the subset of cases where AI would probably give you a valuable recommendation, then that's where my research kicks in.
[00:27:59] So then how do you kind of deploy those? And so here now you're getting a little bit more nuanced of saying, what are situations where an AI's recommendation will give you a good prediction or where will it make mistakes? And usually if I can describe it kind of simply, usually think like an AI tool can process a lot more data than any human can.
[00:28:27] So think in terms of any quantifiable, you know, feature information, it can use that and process it a lot better than I can. And it's looking at historical data, right? So now think about when, you know, what are the limitations in that? And put differently, what are the strengths of the human? Well, when the human knows, when your human expert knows something that is not quantifiable,
[00:28:55] then they're going to be able to make an adjustment to an algorithm. So an algorithm predicting fashion demand, a human sees this, you know, very fashionable dress I was talking about earlier, and they have a sense of what other products are kind of maybe most similar to that one. It's really hard to quantify that and assign attributes to it, but a human merchandiser, a buyer, that brand, that designer, would certainly have an idea of this.
[00:29:23] So they have this kind of private information or knowledge that an algorithm can't really quantify and access. Another example is like when the history is no longer representative of the future, right? So think, for example, I teach a case to my students on Wayfair during COVID, right? So, all right, so Wayfair is selling furniture.
[00:29:46] Living room furniture is a big category. They have algorithms predicting demand of couches. COVID hits. Now everybody wants to invest in their living room furniture because they're at home all the time. And so their demand totally soars, right there. The previous historical data of what demand should be for couches is out the window.
[00:30:08] Now, for, you know, someone who works at Wayfair in this furniture, in this living room furniture category, sees this coming, right? They understand what COVID is. An algorithm has no idea what COVID means. And doesn't understand the economic implications and really the lifestyle implications that that would have on their consumers.
[00:30:26] But a living room furniture buyer at Wayfair certainly would. And so they understand that people, that this is going to be the trend. And so that they know that, hey, we should be, you know, changing our price or changing our advertising, changing our sourcing of couches because of COVID where no algorithm could predict that.
[00:30:45] So things like that. So, you know, understanding that you're limited in AI based on, you know, how representative your historical data is, what types of attributes can be quantified versus not, et cetera.
[00:30:56] Karl Siebrecht: Great example. It's a great example. Bet that's a great case to teach.
[00:31:00] Kris Ferreira: It's a fun one. Yeah.
[00:31:01] Karl Siebrecht: Yeah. So let me shift gears just a little bit here and tap into your perch as a business school professor.
[00:31:09] Over the past several years there's a lot, there is a lot of talk about a shortage of talent in supply chains. This is a topic, you know, if you look at like a Gartner survey, you know, to supply chain officers, what are your top concerns? You know, it's been top three probably for years.
[00:31:28] Kris Ferreira: Yeah.
[00:31:29] Karl Siebrecht: So you teach a lot of capable folks presumably in supply chain.
[00:31:35] Have you seen any changes in the level of interest that students have around supply chain jobs? You know, supply chain used to be this backwater thing you may hear and now it's really exciting because COVID put it on the map and now the CEO and the board really cares. And, you know, there's a lot of data.
[00:31:57] So maybe there are some interesting technology and data problems to solve there. So is this like a cool thing that the students are clamoring to get jobs in? Or has that not actually happened yet?
[00:32:10] Kris Ferreira: I think you're the first person that has said cool and supply chain in the same sentence. No, that's interesting.
[00:32:20] I certainly agree with you. And I've seen the same thing with companies as well as with student interest on the fact that COVID kind of put, most of it was worse, but for better, kind of put supply chain more front and center. And a lot of people in general, like general public's mind, let alone our students.
[00:32:43] So I think that's a positive thing. I also agree with you that people are now seeing more applications to maybe cooler topics like analytics, which, hey, one day, 10 years ago, analytics was not cool either. So maybe it'll have its chance, you know.
[00:33:03] Karl Siebrecht: Anything's possible, right?
[00:33:04] Kris Ferreira: Anything's possible. But I do think that there is more kind of
[00:33:09] new analytical type work that's done in supply chains too, that might make it more, of more interest to some of our students that are leaning that way. Whether it's MBA students or the lots of more folks compared to 10 years ago that are graduating with undergrads or masters in kind of like a data science or analytics
[00:33:29] area. I do see a trend a bit, although it's very recent so I'm hoping that it continues for our MBA students to be more interested in joining companies that make stuff and get stuff to people as opposed to kind of big tech companies, which have been, historically, that's been the trend for a while.
[00:33:53] Obviously, you see changes and kind of the big tech industry a bit with some layoffs of larger companies under government scrutiny, this kind of stuff. And so, you know, I think because of that, perhaps there's some more interest moving back towards companies that are making physical products. Yeah, I would say I'm hopeful that there's more of a trend in labor supply this way, but I haven't seen a huge shift on the MBA side yet.
[00:34:22] Karl Siebrecht: You know, there are a lot of people within supply chain, I may be one of them, who think it's just getting really, really cool these days.
Kris Ferreira: I think so.
[00:34:30] Karl Siebrecht: Do you have any thoughts for companies that are trying to attract more talent that is kind of analytically inclined or even develop talent internally, to develop their pool of talent?
[00:34:45] Again, the people who can really successfully apply these kind of AI based tools.
[00:34:51] Kris Ferreira: Yeah. Yeah. Good question. I think it's good news actually for companies on both fronts. I think in terms of like attracting, you know, maybe young folks that have a background or recent degree in data science or analytics, you know, there's a lot of these folks coming out, you know, now compared to five, 10 years ago that, you know, given the, you know, trends of and kind of decreasing demand in big tech fields, which were traditionally like soaking up all of the talent in data science, I think a lot of, you know, recent grads now and new, relatively new employees are starting to seeing like, hey, there's actually a lot of demand for these skills in other industries as well, including in supply chain management.
[00:35:43] You know, and I think a lot of supply chain companies have invested a lot in, you know, storing data over the last many years and have troves of data to work with now, yet don't have the employees to go for it. And so this is like a fun situation for data scientists, right?
[00:36:02] It's like we have big problems, big data, and we can really kind of dive right in and make a difference.
[00:36:10] Karl Siebrecht: Well, Kris, this has been delightful, actually. I love that we've been able to kind of get into some detail and kind of nerd out a little bit on this, in an area that is super important. And I think there are more questions than answers around this, but there's, I'd say, uniform agreement that there's a lot of value here.
[00:36:28] You know, learning how to better apply these tools. And then also right along as these tools get better, you know, there's more value to be had. So it's been really great to hear your perspective.
[00:36:41] Kris Ferreira: I really enjoyed the conversation too. It's always, always great to learn from you. Thanks so much for having me on.
[00:36:56] Ben Dean: Wow. Now I can see why she's at HBS. I would have loved to have a professor like that when I was learning the foundations of economics and supply chain.
[00:37:05] Karl Siebrecht: Yeah, that's right. She's great. She not only brings a super interesting perspective, she's also got great energy and is clearly very, very passionate about her field.
[00:37:15] So, really, really fun to talk with her. The one thing that broke my heart a little bit, though, if I'm being honest, is when she said that not all of her students were dying to get into the supply chain field. You know, when I go to supply chain conferences, that's what you hear is like all the best people want to be in supply chain. It turns out from her perspective, not so much, but that's okay.
[00:37:33] We'll get over that. And we know that the real coolest of the coolest, of course, do want to be in supply chain.
[00:37:39] Ben Dean: Well, speaking of cool people who are in supply chain, I can't wait to get into my conversation here with Rick Watson, who founded RMW Commerce and runs that business, because he adds that practical element to what we just heard from Kris in terms of the decision making, how to navigate uncertainty.
[00:37:54] How do you apply that to your own supply chain? He gets really into the details here and things that you can apply to your business. Let’s take a listen.
[00:38:03] Rick Watson: Obviously it's a super broad topic, so you need to think of a few different aspects to it. You can almost think about it from a market uncertainty, which is, you obviously have a certain set of customers.
[00:38:16] So what is the likelihood that your customers are going to change their preferences? Or how are your customers changing over time? Are they becoming more likely to buy your product or less likely to buy your product? So I would call that kind of market uncertainty and put a number of risks in that bucket.
[00:38:32] Second is, you know, I mean, you should call the sort of like global and structural uncertainty. And so you end up with things like is the world a stable place or is the world a politically uncertain place? It kind of feels like if you go back 25, 30 years, every era you think it's completely uncertain while you're in it, but then if you look back you're like actually it's more stable than I remember compared to today. The U.S.
[00:39:00] is not necessarily the only dominant superpower in the world, which creates a lot of uncertainty and trading relationships. The notion of trading partners and trading relationships certainly affects supply chain and logistics in all sorts of ways from tariffs and codes and regulations and postal.
[00:39:22] And so there's a whole like, I would say, trading and political certainty that gets into sort of the supply chain angle of it.
[00:39:31] Ben Dean: So Rick does a great job there of teeing this up by categorizing different types of uncertainty. But I want to take a step back here and show us why Rick's perspective on uncertainty is a bit unique because of where he sits between
[00:39:47] shippers, clients, and 3PLs in this space.
[00:39:51] Rick Watson: I’ve always just loved eCommerce and the fact that it's so dynamic and seeing it from all aspects, from software company, inside of a retailer, service provider, and now as a consulting firm, I get a little bit different angle from it because I get to sit on the client side of the table as they navigate these challenges.
[00:40:12] A lot of times a client hires a service provider and the service provider’s like wondering, I wonder what's happening in the boardroom while they're thinking about this decision. They're like on the outside of it. As a consultant, one of my goals is to be on the client side of the table, which is one of the things I find most fascinating.
[00:40:30] Karl Siebrecht: That's such a great point, and it really echoes one of the parts of the conversation I had with Kris, where she talked about how inside an enterprise corporation, even within departments or owners of the different pieces of the supply chain, there could be struggles around sharing data across those points, where one department might have data to optimize his or her area of the flow.
[00:40:54] If each owner does that independently without sharing that data effectively, you're sub optimizing the whole, you're sub optimizing the supply chain overall.
[00:41:04] Ben Dean: Yeah. Got so much personal experience with that, with talking to warehouse GMs who are doing things that hurt their transportation team, for example.
[00:41:13] And, you know, in Rick's case, I think the consultancy offers this opportunity to, you know, break down those silos, but I was really interested in the things that Kris was saying about the new technologies that are enabling that as well. At the same time, she said sitting on top of all that are people and
[00:41:33] I've got a good clip here from Rick talking about people as an uncertainty factor. So let's listen to that.
[00:41:40] Rick Watson: And then kind of more closer to home, you have a little bit business uncertainty, which is like, what's the likelihood that your people will stay with you? Do you have stability in your workforce?
[00:41:53] You know, for instance, if you're planning your employment and staffing for a distribution center for the holiday season, like as everyone has already kind of done at this point, what's the risk that you're going to wake up Monday morning and half your staff isn't going to show up? And do you need contingency from like a third party placement firm to get someone in that day if you need to get products out the door?
[00:42:14] So that's kind of like, I think people uncertainty on the ground level who are actually performing the tasks you need every day to service your customers is another big risk that people are trying to remove from their businesses. Some cases with AI and robotics, some cases with automation and some cases with, you know, backup plans and contingencies.
[00:42:39] Ben Dean: So people's a big uncertainty vector, but obviously it's mostly an internal one. Whereas in the conversation with Rick, we're talking a lot about both internal and external uncertainties. And I want to get to another clip from Rick, where he broadens that to global supply chain issues and how you plan for contingency at a macro level.
[00:43:03] Rick Watson: The reality is, if something changes five years, you might need to change your approach. It's not as hard as it used to be to change your approach. And so if you want to take advantage of a near term opportunity that you see in the next three to four years, let's say with nearshoring, it's not an irreversible decision.
[00:43:23] So that's the other great thing about some of these things. Like there are service providers all over the world now that have experience doing these and you don't have to say, it's not like changing your company name, you know what I mean? Or changing what products you're offering, you're just changing like where you land your inventory or what warehouse you put it into.
[00:43:44] And so Amazon, like just for its own sake, is trying to make logistics like Amazon Web Services, like, you know, Amazon Web Services data centers all over the world. If you want to spin up a new compute cloud in Asia, do that. If you want to spin it up in the U.S., do that. And so these global logistics providers are trying to become the same way.
[00:44:05] And so if you want to manufacture in India and China and Vietnam and Taiwan or Mexico and distribute it into different countries, they're making it easier to not make that sort of a one way door that you have to walk through and can't expand from that base.
[00:44:25] Ben Dean: All right, Karl, I wanted to pause on that one because I know this echoes things you've said similarly in the past about one way versus two way doors, as well as the decision making and planning cycles speeding up in supply chain.
[00:44:37] What did that spark for you there?
[00:44:39] Karl Siebrecht: Yeah, that's right. You know, when we think about this broad topic of navigating uncertainty, you know, for me, it kind of boils down to this. Look, we can put investment cycles into getting better and better at forecasting, right? We can apply AI tools to it. We should, and those should be able to help us get better with our forecasting.
[00:45:02] And our forecasts are kind of always going to still be wrong, particularly if you start talking about forecasts a year out, two, three, four years out. The best AI in the world is not going to solve that problem for me. So the real foundational answer to how do we navigate uncertainty is to actually build flexibility and agility into our operations so that when we are wrong, we can adjust very, very quickly.
[00:45:29] We bake that agility into our infrastructure because we know, we admit to ourselves, humbly, that our forecasts aren't going to be right. And so we're really prepared to move quickly. And so effectively to use Rick's word, which is the Amazon phrase, you know, we're creating more two way doors instead of one way doors so that we can adjust on the fly because the world's an uncertain place.
[00:45:55] Ben Dean: Yeah, and that won't change. I think Kris's comment on that in terms of uncertainty over time horizon, uncertainty grows. So the shorter time horizons that you can operate in, the less uncertainty there is, just inherently and mathematically. The full interview here was gold, but we didn't have enough time for it on today's episode, so check out our YouTube channel and we'll post the entire thing there end to end for folks to listen to and watch.
[00:46:21] All right. I want to put a bow on what we had from Rick here. And I think he had this great summary around the age of efficiency when it comes to supply chain. So maybe this will summarize our thoughts. We'll see.
[00:46:35] Rick Watson: We've kind of moved, I would say, into this year and into what we're calling the age of efficiency here at RMW Commerce, which is means you have cash, you turn that cash into inventory, and then you need to turn that inventory into sales and how many times per year that you could turn that inventory in sales
[00:46:51] will determine how efficient your business is from a cashflow point of view. And so that age of efficiency, I think in 2025, is part of the age of efficiency is like optimizing your business for success. Meaning like, how do you ensure that you don't have excess waste? Are you prepared for multiple scenarios,
[00:47:13] like if interest rates stay high for longer than we expect, or if they keep decreasing, that's two scenarios right there. If we have a trade war with China, that's another scenario right there. If the consumer continues to feel stressed, that's another scenario right there. So I think ultimately, look, you can AI or whatever all you want in the face of that uncertainty.
[00:47:40] What's important is to have options for plans, which means plan A is like assuming that we're going to be in this situation for a while. Meaning like, I'm just going to go back like three years, what if we don't discover a cure for COVID for like five, 10 years? Like we could be just how we are. How would the business operate then?
[00:47:57] What if it gets better? What if it gets worse? And so how do you come up with this sort of ABC planning cycles? And the goal is not to be correct, but to think thoughtfully about the future and how the business might react in these different scenarios.
[00:48:15] Ben Dean: It's great to have a plan to, you know, work with multiple scenarios and think about all those possibilities.
[00:48:21] But you also need to actually build your supply chain to be that agile, to respond to those scenarios.
[00:48:27] Karl Siebrecht: Yeah, that's exactly right, Ben. You know, I loved my conversation with Kris and really enjoyed listening to yours with Rick. And you know what this says to me is we need to do an episode on supply chain resilience.
[00:48:38] So let's plan for that. And until then, let's keep this conversation going.
[00:48:46] Narrator: You've been listening to the Logistics Leadership Podcast presented by Flexe. The opinions of the guests aren't necessarily the views of their company. If you'd like to learn more about the podcast or join the Logistics Leadership community, check out this episode's show notes and visit flexe.com/logisticsleadershippodcast. Keep the conversation going. Email us at leadershippodcast@flexe.com. The Logistics Leadership Podcast features original music by Dyaphonic. The show is produced by Robert Haskitt with Jeff Sullivan, Ben Dean, and Karl Siebrecht. Thanks for joining us.