 Hello, everyone. My name is Sina Valera, course lead for SC1x. Welcome to our live event. I'm here joined by Dr. Chris Caplis. Hi, everybody. Executive Director of the Center for Transportation and Logistics. And we are really excited to have you guys and cover some updates about the course as well as a problem for practice that Dr. Caplis has for us today. So let me go through the slides. I hope you all see this very well. So if you're, if you've joined, you should be already in Slack, sorry. And I'm just going to share with you some tips. Make sure you include your names when you ask a question. You will be, uh, so you don't show up as anonymous. Also take, take some time to look at the Questions tab and also the Polls tab. Polls tab is going to come handy right now because I have the first poll for you. So in one word describe your experience with SC1x. So we actually want to know how you feel when you think about the course as we're almost towards the end of it, right? When you think about it, what was the first experience, first word that comes to your mind as far as your experience? Was it too hard, difficult, easy, whatever? So I'll give you a few minutes, a few seconds to do that. While you're doing that, can you go back to the slide to explain Slido? So the thing to do now, what we're going to do, we have some canned parts of what we're going to talk about. I'll talk about practice, problem from practice. Sina is going to talk about a wrap up and we'll do some of those things. But this is also an opportunity for you to ask us anything. So ask me anything but use the question format in Slido. Now's the time to put those in and also as you put those in, you'll be able to see the other people and upvote them. The ones that you want answered as well, upvote those and those are the ones that we'll ask. So don't wait until 55 minutes in to ask a question because I guarantee we won't get to it. But if you get in early, we'll start answering these as they come in. So use the questions and then upvote the ones that you want answered as well and we'll try to filter or feather those in as we go along. All right. Thank you. Okay. So I'm also going to share some statistics from the MicroMasters program with you. We have over a quarter of a million learners enrolled in any other SCX courses so far. From those, a little more than 17,000 are unique verified learners who have earned over 25,000 certificates so far. From those, about 1200 have been awarded the credential. We just finished our CFX exam and over 200 more people received their credentials. We are really excited for those. Hopefully you will also be one of these in near future. We've done three CFXs so far so we expect this to keep going. The next one will be in September, I mean, excuse me, in February. But you guys probably won't be ready for that unless you've taken some courses out of sequence. But now as you're wrapping up SC1X, which is kind of the core course, in my opinion, you're really well positioned to continue on. We'll talk more about that a little later. Okay. And there are only a few countries from which we don't have students. We almost cover the globe, which is very exciting. And let's actually go and see how the results look like on the poll. So you should be able to see the poll results after you answer. Yeah. And it looks like someone put in supercalifragilisticexpialidocious from Mary Poppins, I was born. But it's fantastic, intense, interesting, great, but a little intense of supercalifragilisticexpial. Yeah. So that's interesting. So, okay. All right. Okay. So you are one of the 14,000 learners we have in the course. And we also have over 1,100 verified learners. That's great. Wonderful. Yeah. We have 172 countries represented in our course. And then we should also mention the C-case, who are doing a great job answering the students' questions on the forum and keeping the discussion lively. So this has been a great community. So, Sina, where are the flags on the bottom for? These are the top represented countries. How come the US isn't there? Yeah. Okay. Maybe it's assumed. Oh, come on. Okay. Okay. Yeah. Yeah. So, yeah, we should have included that. But I have a second poll for you guys. So thinking about the topics that we've covered in the course, which one have been most interesting to you, whether it's personal interest or it's been most useful for your career. So is it forecasting, product segmentation, or any of the inventory modules, or the mode selection and freight transportation? So I'll give you again a few seconds to think about this before we move to the slides. Yeah. So as we talk about, this is kind of the main topics in the sequence with which we learned them and talked them to you guys. So hopefully you see now how they all fit together and we'll reinforce that. But as the votes are coming in, it looks like forecasting is more interesting. That's leading pretty substantially as being the most interesting. Which is interesting because I think in years past it wasn't. I know it's the first thing we did, and we've actually cut down some of the forecasting because we went into a lot more detail in that before. So maybe that is making more sense. Usually the latter things that we covered tend to be the most interesting, but it's funny. You guys are approaching onto the first things we learned, which are inventory, or excuse me, forecasting. Also interestingly, managing inventory with multiple skills is coming up. It's a runner up basically, which I also find super, super interesting because in the school we had learned a lot of these basic methods, but just to be able to do quick and dirty modifications to basic models to be able to make them applicable in real situations. That seemed very, very fascinating to me. So yeah, let's move on. All right. So let's see where you are in the course. Hopefully you will pass the course successfully and that puts you in the second step on the MicroMasters path. You will have three more courses to go, I see two X, three X, and four X, which will make you qualified for taking the CFX exam. Once you pass that, you will receive the credential. So what are you going to see in the upcoming courses? So hopefully if you took the courses in sequence, you've passed the analytics, which gives you the basic optimization on regression and other very useful tools. In fundamentals, you learn about inventory and transportation and forecasting. In ESIC 2x, you'll learn about supply chain design, how to design physical information and financial flows. All these so far cover more of a conceptual model, but in ESIC 3x and 4x, we'll actually go and teach you how real life complications and complexities could be handled with what we've learned so far. So in 3x, we will introduce complexity and dynamics and exogenous sources that impact supply chain. And in 4x, we will cover a lot of technology related factors in today's world. If you know the math very well, it's awesome. But if you don't know the technology, it doesn't really get you that far. And then the reason why we had the CFX at the very end is because we found if you just have these courses and you don't have something at the end, you don't remember two courses back. And so with the final comprehensive exam, it consists of two hour exams and they're designed very specifically to help you synthesize everything that comes together. So that's why we do that because it's a learning experience by itself, even though it's a peer assessment. So the first exam really covers the first three courses and the second exam covers the last two courses, but they all kind of tie them together because one of the skills hopefully that you'll learn is when to use which tool or which method at which time. And that's the real challenge that we try to get you to learn. Awesome. Okay. So as you had the chance to think about the concepts we covered, also we want to know which one was the hardest because we've heard different things over the weeks. We just want to ask for your direct response of which one was the hardest. Oh, we immediately have some responses. But let's wait a few seconds and see what you guys think. Yeah. So the first one was what's most interesting and now it's what what gave you the most difficulty. What do you find most challenging and I can guess what it's going to be, but we'll see how we have my guess too. But yeah, although we've done a lot of revision to this for certain weeks to try to as we get feedback each time we run it, we make revisions and changes. We don't necessarily change the content, but the way we deliver it. So we add recitations, modify problems, add more problems, add more sandbox tools. And it's all trying to figure out how to make it better, how you can learn the concepts better. One more thing that I want to mention is we really carefully receive and record any feedback that we get and think about it in weekly meetings and all all the time we want to think about ways to improve the course. So many of the changes that happen in the course actually come from student feedback or CPA feedback. So make sure if you feel like you have an idea of how the course could be differently designed to fit your needs better, just let us know and then we'll make sure to consider that. Okay, we have about you know 20 or so responses in it and it's not quite what I thought it would be. It's a multi-period stochastic inventory is the tops. Typically I see it as single period. Single period models, a news vendor, news boy problems, typically give students the hardest time. But you guys went with a multi-period. That's the SQ and the RS systems. Usually by that time you're getting to understand the total cost equation and how the distributions work for really looking at the probability of stock outs. But that's interesting. Also aligns with the grades that nobody had yet. Okay, interesting. Okay, so let's move on. I also want to briefly mention once you get your credential, well a lot of our learners proudly put that on their resume that helps them with career development or finding a better job or even become more efficient at their own job, at their current job. Plus this certificate also qualifies you for applying to MIT's blended masters in SCM, which is a five months residential program that culminates in a master of engineering degree from MIT. So that's a wonderful opportunity, but it's not just MIT that accepts, that uses this certificate as a pathway to graduate studies. You know, the Purdue University also recognizes the credentials of our closed logistic centers and the University of Queensland and many more universities. So let me just say a thing about that, because the reason we designed the MicroMasters when I set it up initially, it was not to be a pure feeder program to MIT or any of the universities, it's predominantly designed to be a standalone credential. So if you have the MicroMasters credential that means something. You put it in LinkedIn, you put it on your resume, as more people get this and we get the word out, it means something. Also it can lead you if you want to go to a graduate degree. And so not just the program at MIT, but also two other ones that are scaled centers across the world, the Zaragoza Center and the Malaysia Center and all other external universities. And if you want to find out more about this, just go to the website for SCM MicroMasters and we can link you to all of these. These are across the world and so it's pretty exciting to see all these other universities also recognize and accept the credits that you earn under the MicroMasters. All right, so now we're going to cover the problem from practice. Okay, I'm going to hand it over to Chris. Great. All right. Thanks, Seana. So let me just go in here. So what the challenge that we had, you know, we always try to get problems from practice. And this time, I took this from a thesis that we did about two years ago, Shang and Bernie, Wang and Cerno 2017. And they worked with a large retailer. And the problem was, it was a large US retailer, they had more than 1600 traffic lanes. And the traffic lanes in this case were, you know, point to point lanes, DCs, distribution centers to stores or DCs to DCs, some vendor to DC, but not as many. And they used for higher trucking or intermodal to move the product on the lane, but only one mode per lane. They did that for simplicity to make the management a little easier. You could think that sometimes you'd flip to truckload or intermodal on a lane, but that gets pretty complicated in operations. It's a lot easier if you have a load that comes, you know that if a load's the certain lane and it's always goes to truckload, you can go to that routing guide, or if it's always going to go intermodal. If you have to make the decision on the fly in practice, that's harder to implement. So they had the rule, one mode per lane, and the lengths of haul were from anywhere from 200 to 2000 plus miles. So the problem was that the company thought there may be opportunities to convert more lanes from truckload for hire to intermodal. That was the way going in because, and we'll talk about that in a second, but they thought they're missing an opportunity and they wanted to do this, but they didn't know which lanes made sense and how much would they actually save because they needed to present the case to their upper management. So as you look at this, you can say it's pretty straightforward how you can think about it, but it always, it's kind of every problem's the same here. And so what you're looking at is what's the problem that they think they have, and what's the real problem, and then how would I approach this? And so it's not just how you solve it, but then how do you, how big is the problem? How do you size it? So just to step back, and I'm sure if you guys think about it, you could think of what your initial approach would be, and it's obviously a mode choice, right? And so if we look at the two modes, truckload versus intermodal, they really are classic differences. So truckload is a point-to-point movement. It's fast, but it tends to be expensive. Intermodal, in this case, it's all land-based intermodal. So it's a truck or a drage, picking it up from the origin, going to a railhead, train will take it to the next, to the final destination railhead, and then a truck takes it the last mile. So you have first mile, last mile, drage moves, DRAY, that connect to the road network. And intermodal is always less expensive than truckload, but it's slower. It takes a lot longer. So this is the basic trade-off. So we talked about this. I think in the videos we used it in ocean shipping with intermodal, but it's the same concept because at the end of the day, all you care about is time and money, and so you just have to convert everything. So when faced with this, Chang and Bernie, they developed the total cost model. And this can look somewhat familiar, but not totally. So let me talk you through it because you're used to the total cost model. And I think Cassina will talk about that at the very end, but I presented in class. This was a little different because working with the company, they had slightly different priorities and approaches that they wanted to incorporate. So if you look at the total cost of moving a shipment, and all the shipments were on the same size, they're 53-foot containers, because whether it goes intermodal or whether it goes truck, it can be in a 53-foot size. So it's not like the modes have different volumes that they're shipping. So you can think of it as having two main components. The first is the direct cost, and the second is the inventory implications. So the first side, the CPL or the cost per load, what the carrier is actually charging, and it's cost per load times the number of loads per year. That's the D, the annual demand in that lane for the number of loads. Then in the brackets on the right, we have a very complicated term, but you'll notice that it's just one term. You could think about inventory and you would want to have cycle stock, safety stock, and pipeline inventory. They decided that the cycle stock is not affected. The way that they set the cycle stock they said is not really impacted by our mode choice. They did not want to consider it in here. And the same thing with pipeline inventory, which is kind of interesting. We debated, and you could put the pipeline inventory in. We did not for this because they didn't want to consider it. So instead, what we have in that bracket is all safety stock. So H is the annual holding rate, and 8,716, excuse me, are the hours per year. And so this is down to an hourly basis, and it'll make sense in a second. C is the value of the load, the stuff going in the container. How valuable is that? Is it a hundred thousand dollars worth of material? Is it a million dollars worth of material? The big D is again the annual demand on that lane. The little D is the distance in miles on that lane. And then the last term is distribution. And we actually used the log normal distribution here. Not a normal one. It's a little more complicated because it's a multiplicative incentive and additive property, but the concept is the same as everything you learn for normal distribution. And so what this function returns is the time required for a given service level. So 95 percent if you want to make sure that you arrive within certain time, 95 percent of the time, based on the historical mean and standard deviation, u and sigma, and you set a service level, it tells you the number of hours it would be. And so actually hours per distance. So that's why the units for H is in hours. So if you modify this all out, multiply it out, it's dollars. So at the end of the day, this is what they came up with. So my question for you, it's a polling question, as you look at these data elements, which do you think is the most difficult to obtain? And so I'll give you a second to look at these. And because it's not quite what you guys might think. So we have, what, one, two, three, four, five, six, seven, eight elements in there, right? Where the cost of per load for truck load, cost per load for intermodal, then also what we had, the annual demand, the holding rate, the value of the load, the distance of that lane under question, the parameters of the distribution of time it takes to move over that lane. So what do you think was the hardest piece of data to collect? And so very few of you are participating right now. So you need to make your vote and see what you think. Because the challenge here, and I think you'll see this in any problem in practice, is that the math generally is not the hardest part. The hardest part, in my opinion, is getting senior management everyone to agree on what the real problem is. The second hardest part, I believe, is getting them right data. How do you collect the data? And so I'll give you guys a few more seconds. But the leading vote right now is for the annual holding rate, H. And while you're voting, I'll just let you guys know that was not the hardest thing to do. That was actually quite easy. That's given to us by finance. And so that's a number that's usually in every company, someone knows that number, the CFO, the Chief Financial Officer, or someone in his or her organization will tell you what the H is. So that is not the hardest thing. That's actually the easiest thing that we've put in. Somebody took back their votes. Yeah, people are moving their votes. All right. So it looks like besides H, which you guys said was the hardest, the annual demand. No, that was pretty easy too, because you had historic history, and you could guess what that's going to be, that rough thing. So you had a pretty good estimate for that. So you guys are striking out right now. The standard deviation of time, that one was a little difficult. So I'll go with that as a harder one. But what's interesting, let's see, do I have them all on here? Let's see. Yeah. So what's interesting is that the hardest part was finding the cost per load. And the reason for this, and it's kind of a trick question, they didn't give it to us. And even if they gave it to you, it doesn't mean that they're going to be accurate in the future years. So this project that we worked with, they were not allowed to release their cost information, what they paid the carriers for some contractual reason. So we had to figure out what the cost would be for truckload or intermodal. But even then, they had a problem, even if they gave us their rates, because remember I said there was only one mode typically picked per each lane. So they might know the truckload rate on a lane, but they wouldn't know the intermodal and vice versa. So a lot of times, and you'll see this in SC2X, if you take that for supply chain design, if you're designing where a new facility is going to go, you don't know what those rates are, because these are rates you've never seen before. So how do you estimate the cost per load that a truckload carrier or an intermodal carrier is going to provide to you? Let me tell you what we did. We worked with another large data set that had cost for transactional data, and this came from a company called Chainalytics, and that I'm affiliated with. And so we did analysis on that, and we developed our own models. And our models, essentially, used this equation. We used regression. And so we regressed on all these transactions. And the data set usually consisted of an origin, a destination, a cost per load, a distance, and some other information. And all we did was regress based on cost per load as a function of the distance, the origin state at this point, the destination state. And then some other things, in this case, we looked at the volume, because the volume on a lane can have an impact. And actually, more volume you have lowers the rate on a corridor. So to explain what we've got here, let's say here's the western part of the United States, and let's say I have a load that's going from a California DC to a Minnesota sorting center, right? So going from California to Minnesota. The question is, what will that cost be? We all know that it's not just the cost per mile direct for the distance. The origins and the destinations matter, right? Because we talked about, I believe in the videos, like if you went into Florida or out of Florida, it matters because there's sinks in sources. The challenge for this transportation is a balance problem. So in this case, leaving California and entering Minnesota has different effects. So how do we capture those? Well, essentially what you do is on this large dataset that you regress and you say, okay, there's an origin effect leaving California. Every load that leaves California has that binary variable in the regression model. And the same thing for the origin, anything going into Minnesota. So you're able to aggregate because one of the challenges you have with any truckload transportation or any kind of transportation network is that they're very sparse and they're very unique to a shipper. So if you have 100 shippers lined up, less than 5% of the lanes will overlap each other. But the origins and destinations are pretty common, especially if you go at the state level or the large three digit zip level. So we're able to regress and find these coefficients. So we're able to estimate what the cost per load for truckload and intermodal would be based on more recent data. So kind of interesting what it took and what we did for this, but this leads us to my next poll. And so what we did, let me just go back, you can see the poll. I'm asking you what you think the results were. What we did is we developed this total cost for every lane for truckload and intermodal. So every one of those 1600 lanes has two costs now, right? What the truckload cost per load would be on an annual basis and what the intermodal cost will be on an annual basis. And so we can decide which one's lower, right, is the mode we're probably going to pick. And so the question that I'm asking you is what do you think the results were? And I gave you four questions, four options, rather. The first is that the model mimics the current mode split. In other words, it just replicates what happens. The next is that the model moved more intermodal to truckload. And the third was the model moved more truckload to intermodal. And the fourth was the total mode split stayed the same, but specific lanes flipped. But on average, it's about the same. And so I'll give you guys a second. And you're pretty much split here, but we'll let you guys come in because this is the thing that when you do research, we always go in with a hypothesis. And if you took SC0X, you understand hypothesis testing. It's a really good way to clarify what it is you're trying to solve. So we went in with the null hypothesis that this company was using more truckload than intermodal and that the model would flip more to intermodal. That was our null hypothesis. So then you go and you do the analysis, and you see what the results are. By setting up your null hypothesis ahead of time and having what you think the result is going to be, it also helps you as you do more of these to improve your forecasting ability, your ability to understand what the results will be. Because as you do more of this, and you see this with experts in your own firm, I'm sure, the experienced people can usually have a rough idea what the results going to be going in because they've done this enough times, you get a good second sense to that. Okay. So we have a number in and say the number one response was it moved more truckload to intermodal. And the next was the split stayed the same, but specific lanes changed. That's great. That's exactly what I thought it was going to be, too. And so we're all wrong. Here's the results. This is a confusion matrix or just a two by two. And the columns are what the model recommended for truckload and intermodal. And the rows are what the choices, the current situation was. And so you see that for the current stores, it was 65% truckload, 35% intermodal. And the model recommended 81% truckload and just 19% intermodal. So on the diagonal, the top left and the bottom right are where they overlapped where the model predicted truckload and actually is currently truckload. And then on the other side, where intermodal was the current and they'd act model said, yes, that makes sense. But on the other diagonal, bottom left to top right, those are the ones where they're saying you should change. So 4% of the lanes switched from truckload to intermodal. That's what our recommendation was. And for intermodal, then there was 21% that where it's currently going intermodal, we said should go to truckload, which is really interesting and not what we expected and not what the company expected. They were actually pretty surprised. And so what you do when you come with this, it's always great to get a result that you didn't expect, because it makes you think. And so when you do something like this, you can do a couple of different things. You can think about it lane by lane, and try to understand which ones flipped, or you can try and do like a sensitivity analysis. And that's what we did to try to understand why things moved, because we have to make some assumptions here. And so what I have here on the top right is a sensitivity analysis that the student team did. And let me explain the chart. On the horizontal axis is that C value, the value of the load that 53 full container. And these are in dollars. So 250,000, 500,000, all the way up to 1.75 million. On the vertical axis is the ratio of truckload to intermodal costs. So I put this red line, and that red line is where they're equal, right? Because remember what we did here on every lane, we calculated here's the truckload cost, here's the intermodal cost. So this is just the ratio of truckload over intermodal. So if it's above that line, that means the cost per load for truckload is greater than the cost per load for intermodal, which means you go intermodal. The bottom line is where the cost per load for truckload is less than intermodal. So what is this telling me? It's telling me as the value gets higher, then I'm going to go truckload. Why? Because it's time. It's time. So you want to be faster. So that kind of makes sense. Now it was tough to get that C value. That's another one that was tough, because these are products that are constantly changing, right? You can think of a store and over the course of a year, it stocks different things. So that was a challenge to get. But this is a nice proxy, because this is a nice result that they can use, because chances are for the store, they didn't have too many million dollar shipments. So if I had to do this for them, I would really shorten the access and maybe only go to half a million, because then you'll see that everything below about, what is it, about 120,000 dollars makes sense to go intermodal. Okay. So then they also looked at it in terms of the service level. So remember, we use that log normal distribution, and my chart showed 95% on time. Well, this is just the opposite. So if I have a very low service level, then what am I going to do? I'm going to go intermodal, because I don't care if it shows up on time. So up until, what is across the line, about 80%. So if my service level, this is cycle service level essentially, probability of not stocking out, if I'm below 80%, intermodal makes sense. Above that, truckload is going to make sense, because again, it's the time factor. So we came up with a solution, and a lot of times students think, you know, we stop with that little chart on the left, but here's a solution. But that's really only going halfway. The real challenge that you want to do, and the thing that I think is most helpful to companies, is to look at the sensitivity analysis, and understand why it made decisions as it did. And it could be that, you know, we overestimated the value of the product going through, because that would definitely shift things more to truckload. And same thing with the service level, do you really need 95% because as you up the value and up the service level requirement, you're going to shift to the faster mode. So hopefully this gave you a sense of the type of project that we did. The other thing that just to make sure you guys understand, and then we can answer if there are any questions that come in, is that when we looked at this, it's really easy to do this on lane by lane. Let me just go back. So here's a little equation. Yeah, we can do this for one lane. The challenge was we want to do this for 1600. And we wanted to constantly do it. So we didn't want to just have to do this. It can be complicated. So the other challenge that they had is how do you scale this? How do you scale it and make it usable for the end user who won't be going into the math? We ended up delivering them a spreadsheet where they could enter in new data and actually come up with their own solutions. And this was a little complicated because like I said, we use the law of normal distribution, which is a little more complicated to implement within a spreadsheet. So that was the problem that we wanted to talk about is a mode choice problem. But I thought it was kind of interesting and it fits right into what you're learning right now, mode choice. And what's really interesting, what drives transportation is inventory cost at the end of the day. I'm also going to challenge you with, you know, exercise. So try to think about the total cost equation and try to see if you can predict how things would move. If you can do that, that means you really understand the dynamics of how different factors relate to each other. Look at how the formula works. For example, you have cost per load times demand. So that means there's a linear relationship between the two. So if demand goes up, that cost element also linearly increases. If you think about all other elements like that, then you get a very good sense of what happens to the system overall as different parameters change, as costs change and so on. Yeah. So let me explain. Someone just asked a question for me to explain the charts. I thought I did, but let me just do it again. The two charts you see on the slide, the x-axis of the top one is the value of the product that's in each container. It goes from zero to 1.75 million. And the horizontal axis on the bottom one is the cycle service level. What probability do you want to make sure you're not going to be late for the transit time? So the takeaway is, as you increase the value, as you increase the service level, you'll tend to favor the faster mode, which is truckload. As you move in the other directions, you'll favor the slower mode, which is less expensive than the intermodal. So I hopefully answer that question. All right. So another relevant question to what you just said is, how do companies estimate ordering costs and cost of carrying inventory? You just said some of these are directly received from the finance department. Yeah. So we talked about this in the videos a little bit for the total ordering cost. What you can do as a really rough thing for activity-based costing is you look at the total cost of the software you use for order management, the headcount, their annualized, fully loaded salaries and divided by the number of orders that are coming in. For transportation, it's a lot easier because you're paying a third party. And so for this, the cost per order here is actually the cost per load. Right? Does that make sense? So it's the cost per load here that I'm paying for the truckload carrier or the railroad or the intermodal corporation to move that load. So it's pretty easy in this case. The cost of the C sub e that I use in there, that's really two pieces. One piece is the value of the product, which you should know that's pretty easy to get from a master skew file. The other is the holding cost, which I guarantee every company has one. And it's usually depending on what their, there's many different approaches, but it typically falls into what either the cost of capital or the potential other use of the investment dollar. So more high-tech companies will have a higher value. They'll have a, you know, 0.25. Other companies might have a 0.1. It's really just a lever. As you increase that up, you're going to hold more inventory. And so that's all it's used for. In this case, I don't, I don't recall exactly what the H was that they used in here. I want to say 15%, but I'm not positive. But that's usually a number that you get pretty easily. And so that C sub e is just the product of those two things, the cost of the value, the cost of the product times the annualized holding cost. And then just make sure you're in the right units of time. Okay. Okay. Great. All right. All right. Okay. So let's wrap up the core, what we covered in the course, basically. So I'm just going to remind you about what we've done so far. We've, this has been an exciting journey for us and for you. We could be in the beginning, we covered demand forecasting, we looked at different approaches for forecasting, time series analysis, we covered naive cumulative and exponential smoothing, which I'm sure you had a lot of fun with. Then we, so this model looks at history and past events to predict future. But then we said, there's another method that uses causal analysis. So when you know other factors, factors other than previous events, how can you use that to predict future? So that gets us to the OLS regression. We also covered some special cases of exponential smoothing. Then we switched gears and started inventory management discussion. We started it with deterministic models, EOQ, for example. Then we introduced the stochasticity for single-period models and multi-period models. Then we looked at special cases. What would you do when you have multiple SKUs? Basically all of the models before that was about one SKU. How would you handle multiple SKUs, multiple locations and more realistic cases? We talked about warehousing a little bit. And then in the last week, we talked about transportation management fundamentals of freight transportation and also how does transportation links to inventory. Let me just say one thing about demand forecasting, the special cases. The two cases that I believe we did was one was for new products and the other was for sparse or intermittent demand. It's shocking to me how often these are major problems for companies. That's why I wanted to make sure we address them. If you work for a large manufacturer of an OEM, an original equipment manufacturer, like an auto or heavy equipment manufacturer, you'll be dealing with sparse demand, where it's very infrequent. Crosston's method is very valuable and it's a nice thing for you to pull out of your pocket if you're put in that situation. New product forecasting is just everywhere. It's no one does it well. It's a really hard thing to do, but I think the real takeaway from that, hopefully you saw from that when I showed the funnel diagram, the stage gate diagram, is to make sure if you're in this, if you're in a company that's introducing new products, you want to be at the front of that funnel. You want to know as soon as possible so you can start forecasting what your supply chain needs to do to be able to support this new product. It's one of the biggest challenges that we have. A lot of the thesis projects that we have are dealing with new product and promotional introductions, which causes a lot of interesting for-demand forecasting. I also had a student who just got a job in a medical company. Their job was to innovate and come up with new products. He actually came back to me after the course and said, that's my main problem. How do we predict the amount of this kind of product? We have no idea. You have no history. Yeah, great. Let's quickly see how the topics I just talked about link together under one equation. Chris also talks a little bit about this in week 11, the wrap up. The idea is we have this amazing total cost equation that has a flexibility to include a lot of different scenarios. So we've talked about purchasing costs, element in there, we include ordering costs, then we have a larger component about inventory costs, which includes cycle cost, safety stock, and pipeline inventory costs. Then we can also include stock out costs. So far this is a lot about inventory, but how does the forecasting fit in here? For forecasting, basically if you look at the formula, we see a lot of these. That's the level of demand. So forecasting first helps us understand what the level of demand is. That obviously has that big impact on the model, but it also helps us understand what is the variability of demand. If you look at the carefully, you know that variability has a completely different impact compared to the level of the demand. So it's also very important to have a good estimate of the variability. And let me make sure I understand, reiterate that the sigma that you use here, the root mean square error is actually not the variability of the demand itself, but of the forecast error. So if you know exactly what the forecast is and you have no error, then you will have no need for safety stock. But as your forecast gets worse, so that's really the forecast error in there. That's what you've got to cover for your safety stock because you're covering that period from when you're being replenished. So that's where forecast definitely fits right in there. Okay. Then what about transportation? Okay, so transportation may have fixed cost for every time you ship and order things. It may also have variable cost for the quantity of items that you ship. So those will impact the first two components, the ordering cost and also the purchasing cost. So if you have a per item transportation cost, that will be added to your purchasing cost. If you have a per order transportation cost, like every time you order a truckload, that will be added to your ordering cost. So that's how you can actually capture transportation in the total cost model. It's also, again, going back to the idea of variability, it's also important to have a good idea of the lead time and the demand that happens over the lead time. That is obviously directly impacted by transportation as well. So basically the two other concepts besides inventory that we have talked about also link very nicely, fit very nicely under this model. Okay. So I have a final poll for you. So what, basically, we want to understand what features in this course really work for you. You really like it, whether it's related to the flexibility of having an online course, being able to schedule your sessions anytime, or the specific delivery methods that Chris uses and other people in recitations and other content, or maybe this specific, or maybe it's about the specific content that we covered, forums, help email account, inventory simulation game, and so on. So just let us know in a few words what really was the most helpful in taking this course. And while we're waiting for those results to come back, let me just answer a question from Chris that said, commenting on my clarification of what the sigma is, the variability of the forecast of demand, not the demand itself. But typically most companies won't have that. And so then you're going to assume that the variability of the demand that your forecast was actually the average. And therefore the variability of demand would be the variability of demand itself. Because you're assuming that your forecast was what the average was. So therefore the variability of demand is the forecast error. If you have forecast error, use it. If you don't, use variability of demand. Because you're assuming that your forecast is the average. But typically most companies, very few companies that I've seen actually use the demand error, because they just don't keep it. It's a harder thing to do. Okay. So let's see actually, we have a few more questions coming. Maybe we can cover one more. Yeah, sure. We can work down these and we can probably switch off the pole if you want to go straight to the camera. I think are we done with everything? No, we still have some more here. Yeah, we have a little bit. Yeah, so let's go to that. Let's stay on the PowerPoint. Okay, so let's finish this and then we'll come back to questions. All right. So naturally, if you want to take the courses in sequence, which is what we recommend, then your next step is a C2X. So Chris is going to talk a little bit about what you're going to see in a C2X and the content. Yep. So okay, so you've taken SC0X, I hope, and you have all these analytical tools in your tool belt. You've taken SC1X, you understand the fundamental trade-offs of forecasting, inventory, and transportation, how they all try to get together in that one equation that binds us all. They all fit into that. SC2X builds on this and it says, okay, we've got these three flows that hopefully you guys are very familiar with. You've got the information flow, physical flow coming back at the product, and then a financial flow, some payment. And then you have also information for status as things are moving in progress. Well, what we do in SC2X, we take each of these three flows and say, how do you design for them? And the most obvious one that occupies a lot of supply chain time is the physical flow. So that's where we start. But then we get into some of the information flow. We specifically look how information flows across the supply chain from procurement, all the way through manufacturing, distribution, to managing your customers. And then for the financial side, we try to introduce some financial concepts so you guys can speak the language of the CFO, because everything we've done so far is all good for if you're talking about the supply chain guys, right? They'll know what cycle service level is, and they'll talk about, you know, Sigma and all these different things. The CFO doesn't care. They might not even understand. So what he or she needs to know is, what is the impact? So that's why we focus in on the supply chain finances. So here's how it's set up. Again, the same structure, 10 total weeks, four weeks of content, a midterm, four weeks of content, a final. And you can think of it at the first half, weeks one through four, it's all about physical flow. And so afterwards, week seven to 10, it's more on the information and the finance. So in the first two weeks, we really look at network design, facility locations, where should I put my distribution centers? How should product flow from manufacturing to the DC to the customers? What customer should be served from a from each DC, all those kinds of questions. And the big tool we use is mixed integer linear programming. Now, if you took SC0x, you know what I'm talking about when I talk about a mill. If you SC1x is your first course, you probably have no clue. Mixed integer linear program is a very powerful technique, but it isn't used as much when you're looking at just inventory policies or forecasting or transportation. It's heavily, heavily used for supply chain network design, facility location, flow optimization, it's sometimes called all of those things. And that's what the first two weeks are. And that's great. That's really fun stuff. Then we use the exact same tool that mixed integer linear programming to talk about production planning, which is something we haven't talked about in SC1x. It used to be part of SC1x, but I moved it here because it makes more sense. We're using that mixed integer linear program tool. And it's all about what products should I produce in what sequence and how does that influence when I want them to be received. So you'll see a lot of the techniques that you use in SC1x, some apply, but you have this demand where it's indentured. So if I'm ordering all the parts that go into an automobile, I'm not going to necessarily, well, the demand for, say, the carburetor is going to be heavily tied to how many engines I made, which is very tied to the number of cars that I make. So you look at the end item and the demand for that end item and the demand for all the subsequent components filters down from that. So we use a bill of material. We use something called an MRP. So we talk into all that and it really sits upon a mixed integer linear program. Then in week four, we flip and we start looking at the other end. We introduce something called SNOP, sales and operations planning. And this is all about how do I make sure my manufacturing and my capacity matches my sales and what marketing ones. So how do you make those things come together? Because then we look at distribution channel strategies. How do I hit that customer? So weeks one and two are all about location of my facilities. Weeks three and four are how do I get product in and then how do I serve my customers? So after the midterm in week seven, we look at procurement strategy and Professor Yossi Sheffi comes in and talks about just basic procurement things because while procurement might not sit in supply chain, you'll work with those people every day. And then we'll also do some, he'll talk about optimization based procurement, which is work that he and I have done over the years here at MIT. And it's really a pretty common technique now. So he will spend some time on that again using a mixed integer linear program. And then Dr. Jared Gensel and Mr. Jim Rice come in and they talk about supply chain finance. We give you a quick whirlwind about basics and accounting and just financial flows. And then we go into more detailed cash flows and discounted cash flows and specifically supply chain financing, which is where you're actually paying your vendor faster for a better term. And so we'll talk about this, but the real purpose of weeks eight, nine and a half of 10 is so you can speak the language of your CFO. And it's really our transition into the second half of the whole micro masters, which is more, it's more qualitative in nature than the quantitative models. And then the last lesson in week 10 is the last element is designed. How do you organize your function? Because there's a lot of debate over what's centralized, what's decentralized, what goes in the business unit, what's done under geography. I just finished meeting with a company, large pharma company right before this webinar. And this was their major challenge. They're going from a very centralized model to all the business units have to be self-sufficient. So what does that mean for supply chain? So that's what we talk about in that. So the whole overview is we talk about flows, the design of the physical flow, the information flow and the financial flow for supply chain management. And it's a fun course. It's one of my favorite courses. It's in my top five. How's that? And with that, let me turn it back to you, Sienna. And then we can see what questions we can answer. Okay. So let's actually stop the video with the slides. All right. So you get to see us now. All right. So Venetius is asking, does it make sense to remove outliers from the cost per load? And maybe this is a more general question. When you try to use the historical data to come up with specific parameters, how do we treat outliers? Well, because we used regression, it got rid of the outliers, right? So I came up with an average because regression is giving you an average. Now, we all know if you took SC0X, you realize that regression, ordinary least squares regression anyways, very susceptible to outliers, right? It's the heavy kid on a seesaw. As you have a bigger outlier, you can do that. So what we typically do is I like to put collars. So you look at the data that comes in and you know that if something's a penny a mile, it's wrong. If something's $10 a mile and it's going like a hundred or a thousand miles rather, you know it's wrong. So you can actually put some collars to get rid of the obvious exceptions because for the transportation, what you want to get rid of, they're not outliers in that they were errors necessarily, but there might be a spot load. Might be a Friday afternoon and the primary carrier turn it down. So they have to go to the spot market and they're paying 3x. Well, you're not going to plan on that. That's another problem that we can talk about another day. You want to find out what the typical contract rates are going to be for the majority of your loads and you don't want to have that one load influence the model. So yes, I think it's always makes sense to look at your outliers. To remove them, you have to be very careful. And when you take SC4x, we talk about this that you never ever actually delete data. It's just the way what you flag and what you consider. You always want to go back to audit to make sure you want to throw something back in because a lot of times as you're working on a problem, what you think is an outlier is actually the primary case. And so you've got to make sure the company and your sponsor understands that. And the other thing is sometimes the outliers, and this is something that I worked with Bruce Arnson. He mentioned this case where he did an analysis for companies for lead time of products coming in and they found the average, but then he started looking at the outliers on the fast time, like things and if the average time was like 24 days, they'd find some cases where it took 12 days. And so instead of trying to model to the average, they said, how do they do it in 12? Let's replicate that. Let's find the outlier on the positive side and say, what happened? What needs to happen to make sure we always have a delivery on that outlier side rather than defaulting to the average? And so I think looking at outliers, outlier analysis by itself is always a good thing to do. It's always a critical thing to decide whether you're going to keep something or not include any analysis. And I can add that you should always be very careful about ratios as well. Sometimes the culprit for outliers are the ratios. So for example, if you have a cost per employee, then in a specific branch or specific day you have only one person hired, then that's going to create a huge change in your value. So we have a very popular question asking about what's the best way to revise or review SC0s and SC1X to revise or review. The reason why we, so I don't think any of you, well if you're taking this course now, you didn't. The first time we offered this course was in the fall of 2014. And at that time, it was a very different course in that we didn't have a midterm final. And also there was no key concept document. And the number one complaint students had, besides there weren't there were no community TAs, there was one TA, it was just me and they complained about the response time, but the real complaint they had was there was no book. And that's why we created the key concept document. To me, that's the best way to do it, to understand, refresh the concepts and then do practice problems. That's the best way that I think it takes to review. There's also some good reference books that I put in there. I'm a big fan of Silverpike and Thomas. Used to be Silverpike and Peterson. Thomas from Penn State, it took over. Actually, he's not a Penn State anymore. He moved. Do you know where he went to? He's not a Penn State. I have to review. He's at oh, UVA. I think he's at UVA now, Dartmouth School. But it's a really good book. It's a great reference book. So the way I view any of these kind of things, there's an entrance ramp and then there's a reference guide. So we use the videos to learn the concept. And there are some textbooks that are great to learn from. But there aren't necessarily the same books you want to have as a reference guide on your shelf because you don't want to go through all that text that you need to get that on-ramp to understand the idea. You want something short and concise. That's what the key concept is for that document. And also, Silverpike and Thomas is a great reference book because it goes into a lot of detail. In fact, someone asked a question. I just saw it. I think Jamie or Jaime asked a question. His company uses order fill rate instead of item fill rate. And how do you convert? Go to Silverpike and Thomas. They'll talk about that. It's a little more complicated. Someone said that they think they're the same thing. They're only the same thing if your orders are of a single item. So sometimes it's like the perfect order stat. So there's a little complications. And the other thing that I find that's complicated for companies is that they are sloppy in what they say. They'll say service level. And that could mean anything. And so understanding how they measure order fill rate, it could be the number of orders shipped perfectly. And so to a customer, that might consist of multiple SKUs. Or it could be some other mix of statistics. So I think the best thing to do is understand exactly what it is capturing. And then you can look and see what that translates into an equivalent item fill rate. It might not go exactly, but Silverpike and Thomas is a good place to start. Okay. So Pram is asking other plans to introduce MILP in anywhere in SC1. I think since we covered in 2x. 0 and 2. I don't see it. I don't see it. We could do production planning. In fact, the course this is modeled after is SCM260. I teach here at MIT. We do production planning in there. But SC2x, it fits better in there. So that's why we kept it there. Okay. And another top question is, does an optimized inventory increase company profit or it basically improves asset turnover and hands ROI? And if I got it correctly, I think you're not separate things. But I think it just works both ways. So optimized inventory. So when you say optimized inventory, what we're doing is we're making a trade off, right? And so typically you're assuming a stock out. So let's say it's really lowering inventory. Let's say it's that way. It could increase the profit, but it depends how you optimize the inventory. Think of the difference between a single period model and a multi period model. In the single period model, we're optimizing for profitability. But typically what you saw in those problems, the optimal stocking level is usually less than what you would think. It's the 80%, 70%. It's really a function of the costs there. And typically when we think about service level, everyone says 90%, 95%. So a lot of times by going up to that level, the optimized level, setting a cycle service level, that might not be optimal for profit. But it's the optimal making the trade off between stocking out whatever that cost you put there and the cost of holding the inventory. So does it always increase profits? No. When you do a multi period, it's minimizing your total cost. So it's a little bit different because we're not looking at the revenue because the margin doesn't go in anywhere in that equation for the total cost. It does when I do single period, but it really isn't incorporated when I look at multi period. And it's also not solely about reducing an asset turnover. Sometimes you might actually find out that it's better to hold more inventory. So I think we're at an hour. So any last burning question you want me to answer? It's your pick. You're the course lead. Okay. So let's see. Well, I love the concept of sandbox within the edX platform. But some other games are outside the edX platform. Karam is asking. Karam, you're asking so many questions. So many questions, yes. But other plans to integrate them? No. Do I need to talk more? We found it doesn't make as much sense. So in SC3x, we play the fresh connection. It's a third party game. So it's just easier to do that. And we might integrate by making some of the UI stuff, but the user names already pass through. Some of the other games like the inventory game that you guys developed here, that's integrated or is that separate? Well, it's a separate platform, but being included in us. Most of our effort has not been on the integration to this because each of these tools are used in other platforms as well. We use them online, we use them for executive education. So we developed them externally. So I don't think it's a big, it's not a big burning issue for us to integrate them in. Where we put a lot of effort is the sandboxes. Those are integrated. And also all the randomization, which you guys don't see, it's behind the scenes. All that is integrated and changed the way the platform is being done. As well as the change where for those of you who will take the CFX, and that for those cases, when it's a proctor exam, we don't tell you if the answer you submit is correct or not. That's another thing that we added to the platform. It just tells you that we received it. So those are the things we did for the platform. Whenever we have a content tool, we usually develop it outside and we try to make it so it's accessible, but we don't work on the integration too much, maybe in the future. And with that, all right. Thanks a lot, Chris. All right. Thank you, Chris, for joining. Thanks. All right, guys, take care.