MBA Masterclass Managerial Decision Modeling and Operations Management

Lisa Koengeter: Hi, everyone. Welcome. Good evening. Good morning. Wherever you are in the world. We have a large number of attendees. So we are going to just wait a minute or so before we get started, so everyone can enter from our waiting room. We're thrilled to have you joining us this evening or this morning, depending on where you are in the world. Looks like we're up to about almost 300 participants. The numbers keep climbing. Well, great. I'm going to get started. Our numbers, our entry, is starting to slow down. So, welcome everyone to the Chicago Booth MBA Masterclass series, and today's class on Managerial Decision Modeling and Operations Management with Professor Bariş Ata. My name is Lisa Koengeter and I'm the Director of Admissions for the Executive MBA program. It's my pleasure to be welcoming you on behalf of our full-time, our evening, our weekend, and our Executive MBA programs. In this Masterclass series, we are featuring our world-renowned faculty, which gives you a glimpse inside of our classroom experience and our very unique and distinct approach to business education. It's what we call the Chicago Approach. And you are going to experience that live today, and you'll see a couple different features of that distinct approach. The first is our data-driven, our evidence-based learning and teaching. It really gives you tools that allow you to ask better questions, become a more strategic thinker, and help you solve problems. You'll have an example of that data-driven, quantitative approach as you build a model in real time with our professor this evening. The second element of the Chicago Booth experience is our very collaborative community. So here at Booth, you join a supportive network between the students, the faculty, staff, and the alumni community. This group of individuals will champion you each and every step of your MBA journey. And then, beyond that, as you move into an alumni. In true Booth style of this data-driven approach in this collaborative environment, we encourage you to be involved this evening in our class session. Ask questions as we go. You can use the Q&A box in your Zoom menu and type in your question throughout the session. We'll queue some up to take live at the end, and I'll try to sprinkle some in for our professor as we go along, as well. So, let's meet your professor who will be teaching this very interactive session this evening, Bariş Ata. Professor Ata is the Chookaszian Family Professor of Operations Management. He takes a problem-driven approach to bridge theory and practice of operations management. His research interests include the delivery of healthcare services, sustainable operations, management of manufacturing, and revenue management. Professor Ata received his PhD in Operations, Information and Technology from Stanford. He serves as the editor for the Stochastic Models and Simulation Department of Management Science. And he's also one of our most innovative professors in teaching in this Zoom classroom, this remote environment. He's truly been a pioneer in online teaching. And you'll see those elements today in this class. Prior to his academic career he worked for McKinsey, and outside of learning in the classroom, he enjoys learning from his twins, Kevin and Emily. So with that, Professor Ata, I'm going to turn the mic over to you.

Baris: Thank you. It's very nice meeting you, sort of; I don't get to see you. I think that's kind of unfair, but I will look forward to meeting you all in person hopefully in the near future. I also wanna thank Lisa and Kim and others for giving me this opportunity. So, let's get started here. Here's my slides. First of all, you know, I don't know how you felt about it, but I sent you kind of some soft homework already, right? So you got the case, you got the Excel spreadsheet and so forth to build our model. But you know, first of all, if you wanna do it, great, I'm thrilled. Please do it. All you would need as background information is what the function F4 key does. If you know it, great. If not, I'll go over it and then you'll be ready. But if you feel like, you know, it's maybe too early in the morning where you are and/or too late at night, et cetera -- if that happens to be the case and you just wanna watch, that's OK for now. Once you are here, it'll be a different story. So I wanted to do something here because as you probably know, we are looking forward to getting back to the normal -- or rather the new normal -- and be in-person, and because I think that's really, really nice to do. That being said, the last year has taught us a lot of things. And one of the things I personally learned is how well we can harness the technology to reach wider audiences. So I wanted to also give you a glimpse of some of the things. In a webinar setting, this is a little bit harder to do because it's harder to get you engaged. In the Zoom environment, I wanna say some of the things I've been doing. You know, normally, I ask you to turn the cameras on. Unfortunately, I can't. And I would welcome your participation through the chat, and Lisa is so kind, she'll read them to me, and I can answer. There are lots of things that we use. You'll do maybe a poll or two today. And I also use these reactions, and ask questions, get yes/no, and then call on people, and there's a lot of talking, two-way interaction, in my online classes. Now, I did ask you, I sent you a link for the Google document, and I asked you where you are. Thank you for filling that in. It's just amazing. I see people all over the world, and Brandon is in Chicago. So am I. I see people in Japan, Brazil. You know, Portland, more Chicago, Boston, India, Toronto, and so on and so forth. And, again, I mean, you don't really have to fill it in real time if you don't want to. But see, this is one of the things that I would have these open-ended questions and ask for answers and, as I see the names, I'll start calling on people. They'll start jumping in and we would have this back-and-forth. It would be, for example, I would've asked you, "How is the weather there?" And I can say that it has been raining here. Right? So, do you see the rain in Chicago? All right. So let me get rid of that, anyway. So, with that, and there's more -- like there are other things that we can do in this environment that we couldn't do. Like, I have my TA with me and what I would do is create a separate TA room as a breakout room. TA is there, if any student wants to see a detail of something, they can go meet the TA and so forth. So, what I think we will be doing -- I was just talking to some colleagues today, not in-person, unfortunately, but we are going towards that. We are opening up more and more. And we, I think everybody is in agreement that we will be harnessing the power of technology more, even though we are looking forward to being in person. And here, I wanted to do two things. So when I was told about this opportunity, I figured, how do I pick some topic that sort of speaks to maybe multiple courses? There are two courses I've had the privilege of teaching at Chicago. One is Managerial Decision Modeling -- that is, solving decision problems quantitatively, typically using Excel, and somewhat additional add-ins and so forth. And the other one is the Operations Management course. I teach that in our Executive Program and I look forward to teaching it again. And the topic I'm gonna do today is really at the intersection of the two. So, let me jump right in. It is about resource allocation and decision making. When resources are limited, how do you make decisions? And, often, resources are limited, right? Now, the question is what is the best way to achieve certain objectives when you have -- I think I'm in your way a little bit here, so let me, maybe do this. OK. So you can see me. You can see the slides as well. All right. So that's called optimization modeling. All right? So, and that's what we will be looking at. And it has a ton of applications, right? So, for example, you can think about portfolio optimization and things of that sort, or other financial or investment-related applications. My own field operations management has a ton of applications, of course; marketing, like promotional planning, things of that sort. Or distribution network design, something of real importance these days -- vaccine distribution, for example. How do you design that network and in a decentralized way? You send the right quantity in at the right time to the right place type of thing, right? So it's like the detailed supply chain planning type of question. Anyway: So with that motivation, let me jump into the case. So this is a fairly simplified version of reality. Nonetheless, it's gonna help us drive some points home. So, the -- I'm hoping that you had a chance to take a look at the case. But just in case, let me refresh your memory. So here, we are really looking at the simple situation where you have two products, right? They could have picked more creative names, but what we have is Model 101 and Model 102. And because you have limited capacity, you can only do so much of these. And the question is, how do you pick the best possible mix? Right? So how many 101s and how many 102s we should be producing? Now, that would be the one question we will look at and solve together. And then if time permits, what I'd like to do is sort of talk about a follow-up question. Hey, there's a new product introduction possibility, the 103. Should we do it? Should we not do it? How do we think about these types of questions? Now, so let me give you some of the data in the case. All right? So here, what we see is the manufacturing data. Again this is really simplified, right? So no manufacturing process will be this simple. That being said, I assure you that everything we are gonna do and the tool that I will show you is applicable to anything that's of real skill. So here, just to keep things contained, this is typically what we do. Cases are sometimes a bit simplified and guesstimate-- but the insights are generalizable. So here, let me actually go to my board and draw the process flow diagram here. So here, what I have is ... the ... First, we have two products, right? Engine assembly, that's step one. And you can kind of see, maybe not so much -- but if I make this a little bigger, you can kind of see the data here, right? So, while the engine assembly Model 101 takes one hour there; Model 102 takes two hours; and you have like, here on my slide, you have $4,000 available, right? So, this is sort of the data. Let me now draw the next step, and the steps thereafter. So we have a clearer idea of the process. This is sort of the, typically, the first step to studying systems is to draw a process flow diagram that kind of gives us visibility as to what's going on in the system. Here, it's gonna be fairly straightforward. But when we discuss, for example, the operations management cluster, there are some cases where this would be fairly complex. Right? So, but it's the step number one so that we understand what's going on. Visibility, right? So that's sort of the important objective here. So next, they go through the stamping operation. And thereafter they split and they have their dedicated capacities of 101 assembly... and... 102 assembly. All right. The simple, clean and simple, situation here. So let me go back to, again, my slide. Now, please have that picture in mind along with these numbers we got. OK? So we have more, of course, right? So it's not just the operational data we care about. We also care about the financial data, right? So here, what we have is, let me look at here. We have the direct materials, cost, direct labor cost, and, and sort of the overhead costs as well. You see your overhead costs right here, OK? So what we want to understand, obviously, is you can push 101, you can push 102, and we would want to get a sense of which one is more profitable, right? At the end of the day, we care about that. So that's sort of, we want to glean that information from the slide. Now, something that I'll sort of cut to the chase -- normally, this would be a nice point of discussion as to what to do about the overhead fixed versus variable and so forth -- you gotta think about the following, right? So when you change your decision, you are thinking about how many 101s versus 102s to produce. If I increase, say, their numbers by one or two and so forth, what are the costs that are changing with your decision? That's sort of what we need to keep in mind. So obviously the direct material and direct labor costs, they'll contribute per unit, on a per-unit basis. Right? If you think about the overhead, the same is true for the variable overhead but not for the fixed overhead cost. Fixed, as as you imagine, fixed is not gonna change with your individual decisions. So what we want to do here is to kind of focus on the direct material costs for 101, as you see it's $24,000; and then direct labor, that's $4,000; and I'm gonna summarize these in the next slide. And as far as the overhead costs go, we gotta be careful. What we need to do is really just focus attention on the variable overhead cost. Because this is the component as you produce more, this cost increases, right? Whereas the fixed part of the overhead isn't going to change with your decision. So this is kind of a tricky point as we think about this case. So just to summarize: What are the contributions or the marginal net benefit of making these production decisions? I'm gonna look at the revenue of each item and then subtract the variable costs. So if you do this for 101, basically, reading off the numbers from the previous exhibit you would sort of get, I think this is $3,000 per truck. And the marginal net benefit of the 102 would be $5,000 per truck, right? So in a way, you see that the 102 seems more profitable based on the financial numbers. Now a question we have is: The marginal net benefit of 102's is higher, should we push them? Like we can do both of them -- but if I, whenever I can, which one should I push for? Right? So I have a -- I just wanna get your feedback on this one. So let me end this. Should we push... Model 102? What do you think? So, let me give you some time to weigh on this one. You've seen the numbers. So, here is our... poll question, also. Oh, now what did I do? I think I, I didn't.

Lisa: Professor, a few folks were getting, yeah, that are having trouble seeing the poll.

Baris: All right. Now, now I relaunch. I was wondering what the...

Lisa: There we go there. There come, there we see the votes.

Baris: Yep. Yep. Oh, I really, really appreciate this level of engagement. I miss teaching. I, you know, I finished my teaching in March and feels like it was ages ago, really. In this COVID time, time sort of , last year feels like a decade. Anyhow. So people are weighing in. It's kind of split. That makes it great for discussion. I wish we could -- we had more time to actually discuss this back and forth. Let's see. See, in the normal Zoom setting I would've also asked you to give your yes/no reactions. As you know, you can do that in the reactions button below in a Zoom meeting. I'm sure you've done that before. And I would quickly get the names and start asking for, why did you say yes? Why did you say no? Typically, there are good arguments for each decision and that gets the discussion going. All right. So here I have about 80 percent participation. Just in interest of time, I'm gonna end the poll and share the results. So 60 percent, 61 percent said, yes, we should push Model 102. And you what? 39 percent said, no. I'd be curious as to, I'd really curious to learn the reasoning for people who said no. But let's see. So let me offer you my thinking process here. So let's look at the following. Let's continue the thought experiment, right? So here, what is the maximum monthly capacity, monthly output for 101, see, if I focused on, focused on producing them, right? So you can look at each station, right? So here, we kind of have the different steps: engine assembly, stamping and 101 assembly. We know how much capacity they have. You know how many hours 101 takes on them. So you can look at the ratio of each, you kind of get like how many 101s you can do. And that's my calculation here: 2,500. Yep? So then I say, "Hey, if, what's the maximum 102s if I just focus on the 102s?" You can do that calculation and that gives you 1,500. OK? And again, for 102s what you would want to do is, you go through these three steps, right? So they differ in this last step. And, at each step, you say, "Okay, if I look at this one here, I have 4,000 hours. Hey, it takes two hours to produce a 102 here so I can do 2,000 based on this one. Based on this step, I can do 3,000, 'cause I have $6,000 divided by two. In the last step, however, I have 4,500 hours available. Each 102 takes three hours so the maximum I can produce is 1,500. That would be limiting what I can do, OK? So now the question is: If I really push 102s and produce... 1,500, what is left for the, here, remaining capacity of each resource for 101? So I'm gonna subtract what I use and the leftover will be used for 101s. OK? So if you do that calculation of using the leftovers, how many units can you produce? It's 1,000 trucks per month for 101s. So this seems like a viable solution. So the people who voted for "Hey, let's really push 102s," this would be your answer. Right? So now, what is the underlying logic? Right? So they have higher marginal benefit. So we are gonna push 102s as much as possible. That gets us up to 1,500 102s. We look at the leftover capacity and use that for producing 101s and produce as many as those. Now, if that was it we wouldn't be having this class. Right? So the question is: OK, what's the flaw in that reasoning? What is that 39 percent? So that, so far, we didn't discuss. The thing is that you can't just go with the financial data. That's what I've done so far. We also need to look at the trade-off between 101 and 102. In a way, if you look at, like, things of dollars per unit of capacity, that needs to be qualified further ... for the resource that matters. Right? So these are terms that we make really specific in, in the class. So the trade-off between 101 and 102 -- if you really can think of it like an exchange rate in some sense -- is not one-to-one. Because they take different amounts of time at each resource. Right? Or at least some of the resources -- so especially, the very first step, right? So at the engine assembly, 101s take one hour whereas 102s take two hours. So if I backed off on 102s, presumably, I may be able to produce two of the 101s. Right? So that kind of changes the dynamics. Anyway, so let's go ahead to do this calculation. So the total contribution margin, the total contribution now is, here are the decisions, right: 3,000; so 1,000, this was 101s, and 1,500 102s. This is sort of the base case we started off with. Then I said, "Hey, what if I backed off on 102s, and did one fewer? How well will I do?" Now, you can go back and check, "Hey, you can actually increase the 102s not by one, but by two." Right? And if you did that calculation, your total contribution actually comes out to 1,000 more. Right? So then the natural question is, "Hey, can we do this? Can we do better?" And hint: Yes. Right? So, that's why we are here. And as you can imagine, if I give you enough time, I assure you all of you can kind of figure this out and get the right answer and so forth. Now, imagine you have tens of different products. Again, a large number of resources and there are all sorts of trade-offs. How do you make this decision? So for that, we have a general tool. And sometimes -- well, let me first go through the, how to think about problems of this sort more generally. So we wanna go through three steps. I call it the ABCs of Optimization Modeling When I was creating this slide, I guess my kids were into coloring a lot so you see the color coding right there. I also tend to color code the model so that they're more visible. So there are three components. One is the decision variables. These are the choices we are gonna make. It's fairly obvious here. We wanna choose the number of 101s versus 102s. Right? Sometimes, it's not that obvious. But still, we wanna identify what are the levers we can pull, what's in our control, and what's not in our control. And second one would be, what's our objective, right? So that, in this case again, it's fairly straightforward. That's a total contribution. And the last one is, what is not in our control and is restricting us? These are the constraints. Here we are gonna take the capacity constraints as given, right? You can also imagine savings where you may be looking into a capacity investment decision and so forth. But, here, they are the constraints that we are gonna have to live with. So here is how we approach it. Like, it's a more systematic approach to tackle this problem that's sometimes called linear programming. So what we will do is, I'm gonna give names or assign these symbols, mathematical notation, for the number of 101s and number of 102s. X1 will denote the 101s to produce. X2 will be the number of 102s to produce. Now I wanna express the various concerns I have, right? So, step A was to determine my decision variables issue we just had done: X1 and X2. Step two is to calculate our objective, which is basically this total contribution: 3,000 times the number of 101s. That will give me the contribution from 101s; 5,000 times the 102s, that's gonna give me the contribution from 102s. Right? So that's my objective, where as I play with the number of X1s and X2s, I wanna make that number as large as possible. But I can't just freely choose whatever I want because I have a whole bunch of capacity constraints. Now if you plug in X1 equals 10, X2 equals 10, the question is how many hours of engine assembly time do you need? Well, each 101 takes one hour so one times 10 plus two times 10. And you add those up, that gives you 30. And that's precisely this calculation over here. OK? So what I want to make sure is that as I choose these numbers, these two numbers, I wanna make sure that the hours I would need at engine assembly is less than or equal to total hours available, which is given to me in this table. So I cannot use more than 4,000 hours. OK?

Lisa: Quick question, professor. We have a question: "Can you please clarify how you got 3,000 as the coefficient for X1?"

Baris: Sure. I'll have to backtrack a little bit but let me do that. So if you go back to -- sorry, that's... that is the calculation here. And I may be...

Lisa: Thank you.

 ...here little bit. So let me go talk about this further. If you look at this data here, right? So I am looking at the... we, first of all, know the, how much they sell for, right? So, they sell for $39,000. And then I say, "I'm gonna figure out all the variable costs." So $24,000 direct materials; I believe $4,000 direct labor; and $8,000's variable overheads. So what I then do is, I go back to that calculation and that difference gives me the $3,000, and I do the same thing for 102. Hopefully, this addresses the, clarifies the question.

Lisa: That's perfect. Thank you. Marginal net benefit. Thanks Philip.

Baris: Yep. All right. So let me come back up here. So this is where we were. And so now, what we want to do is... all right, so metal stamping. We kind of wanna do the same thing on the list. I wanna calculate the, how many hours I would need if I go with X1 and X2, is the number of 101s and 102s. Now each of them take two hours, right? So two times X1 plus two times X2 is the number of hours you would need if you are going with these decisions. That should be less than or equal to the number of hours you have. So that gives you the second capacity constraint. Now at the final step here -- and on my slide, you see the process flow diagrams -- this is the 101 assembly, that only 101s are going through this step. And each of them are taking two hours and we have 5,000 hours available. So two times X1 would be the, hey, if I said X1 equals 100, I would need 200 hours there. Right? That better be less than or equal to the number of hours I have, which is 5,000. Now I need to do the same thing for 102s and that gives, this is that capacity constraint. OK? So the, this is the very last step. So now, we are ready to solve this. Again, you know, this is a fairly small and manageable problem. It, you know, you could even use like graphical methods and when I was an undergrad, which is now many years ago, we used to solve some of these problems by hand and so forth. But we don't have to do that, obviously. This is why we have Excel. And we -- let me now switch over to my Excel. And again, if you have this file -- even if you don't have it, it's not hard to type this -- and if you wanna do this with me, please do so. I'll try to go sort of step-by-step and hopefully we can do it together. I do welcome the clarification questions and so forth. If there are any, please send them through the chat box. And Lisa, I'm gonna rely on you again. Thanks so much. So let's go ahead and build our model here. So let's see. So we will start with our decision variables, number of 101s and number of 102s to make. To start with, I'm gonna just put some arbitrary numbers: let's say 100 and 200. These are completely arbitrary but we are gonna ask Excel to adjust these numbers for us and let me highlight them in blue. They are my decision variables. Step A is done. Now I'm gonna look at my total contribution. And if you remember, that was just 100 times 3,000 plus, is 200 times 5,000. So I'm gonna write here as, hey, it is 100 times 3,000 plus 200... times 5,000. OK? So that is my profit. Let me highlight that one in green, for dollars, I guess. OK. And let me highlight the formula once again. And I may be able to... magnify this a little bit. So let's see if this is gonna do. C4 times C5 plus D4 times D5. All right. They can improve this feature, but anyway -- all right. So: The next step is to go through our constraints. We had really four capacity constraints, right? So here, if you noticed we already have the amount of, number of hours available. So let me first highlight this one in red. OK? So these are given to us, right? So -- and now, what I want to do is I want to calculate the hours used. If I make the decisions that are highlighted in blue, how many hours I would be using at each step? Right? So if you look at the engine assembly, you are looking at, hey -- 100 here times one plus 200 units times two -- that should give me, what, like 100 plus 400 is 500. Right? So -- but ultimately, what I will do is I'm gonna ask Excel to start playing with these numbers in the blue to adjust them, to maximize the number in green, while respecting my constraints that are highlighted in red. So I wanna recalculate and put that formula in there. So let's say this is equal to this 100. Here, if you'd like I -- let's do this. Please press the F4 key, Function 4 key. Times this one... Plus 200... Times E7. OK? What the Function 4 key does is if I were to just copy this formula down, it's gonna keep those references the same. Right? And I know many of you know this but I don't assume that everybody does, so we can cover it quickly. Lisa, go ahead.

Lisa: I do have a quick question coming in here. In the model, are you assuming that the 102 assembly line cannot be converted to producing 101?

Baris: Yes. That's an assumption that we are, we are making.

Lisa: OK. Thank you.

Baris: That's a very interesting question. If we could, right? So how would, and that would be a great homework question, maybe for next year. That's something to think about. We can certainly think about a lot of questions that, that is gonna build on what we are about to do. And I would encourage everyone to sort of think about them, and I assure you, like, if you formulate a question, there's a very good chance that we can build a model to answer it. All right. So let's say, OK -- and here, let me just copy this down -- basically, this repeated the same formula but for these other resources. So let me highlight these in yellow. I do like color coding. OK. So now we are ready to ask Excel to go ahead and do this. Maybe I'm gonna pause here for a second to see if there are any further clarification questions. Sounds like we are OK.

Lisa: I think so. I do have a question that came in a bit earlier about popularity. So what about the popularity of 101 versus 102? Are you assuming that Merton will sell everything they produce?

Baris: Terrific question. And yes, I should have said that. We assume here there's plenty of demand of each kind. That being said, I mean, you have to be crazy to assume that. At the same time, we can easily throw in the demand constraints here. Right? So they would just be additional constraints that, hey, my X1 cannot exceed the demand forecast. My X2 cannot exceed its demand forecast. But then, whoever asked that question will likely ask, "Hey, but how do you know the demand? Isn't that random?" Then I would say, "Terrific question." Right? Because yeah, it's random. And that's also something we address in the class. How do you make these decisions under uncertainty and so forth, which is the Monte Carlo simulation, right? So today, of course, we are only barely scratching the surface but nonetheless, I hope this gives a sense of the type of things we cover. All right. So let me now go to Solver. To find Solver, you wanna look under Data and then go all the way to the right to the Solver. If you have it, you click on it. If you don't have it, here's what I suggest. If you go under File, then under Options, you would see the add-ins. If you click on the add-ins, then you say, "Go," I know, a lot of steps here. You wanna make sure you check the Solver edit. OK? And again, at this point whether, you know -- if you did it, it's nice. But if you are having technical difficulties, I wouldn't worry too much about it. So let me go to my Solver here. It will say, "Set objective." That's the green cell. I wanna maximize it. Right? And let me clear these up completely. Let me delete this as well. So I wanna be changing the numbers that are highlighted in blue. Those are my decision variables. OK? Again, I think I'm in your way. Let me... And the course has the Star Wars theme. I'm like the Jedi, Obi-Wan Kenobi here. Anyway. Usually I get a laugh at this Zoom, but now I can't see it. Right? So... So here, let's add our constraints. We have four constraints. So what you could do is select these in yellow, and say they are less than or equal to these ones. Let me say, "OK." If you do that, Excel knows that you have four on each side and knows that it should compare them one by one component-wise. So this is how your Solver menu should look like. OK, let me move this way, maybe. And if you have it, great. Now we can say, "Go ahead and solve." So if you are with me on that one... maybe I can move over here... OK. So let's go ahead and solve. Solver found a solution. "All constraints and optimal conditions are satisfied." That's the message we want to see. And just for future reference, what I'd like to do is I'd like to get the sensitivity report. So here is my answer. The total contribution margin is $11 million, I suppose. If you look at the answer, though, what do you think we are doing? Right? So where we started from in the discussion earlier, was maximum 102s at 1,500, and 101s was only at 1,000. Now it's completely reversed. So those who voted "no" were right on because I think they were doing this calculation of dollars per unit of capacity type of thing. Right? So, now I wanna ask you a question. A good question is -- and again, I, we won't have time to get into this but ask yourself -- if I could give you an hour a day today, how much is it worth to you? Right? Time is valuable because if you had more time we can do more things, right? So, you know, I could work out and get rid of some of these COVID pounds. After this, I'm gonna go on my Peloton, hopefully, and work out with Ally Love, or maybe Olivia, and one of these wonderful instructors. So: The point is that if you have a resource, if you had more of it, that should be worth something. Right? So here, I look at the engine assembly. How many hours do I have for 1,000? That's in red. How many hours do I use? 4,000. Now, we use everything we got. We're at capacity. Do you think that if I could give you one more hour at engine assembly, that's worth something? You can put that to good use, right? Presumably, you can. What if I took away an hour from you? That should lower your profit because you were using everything you have. If you have fewer hours, you are not gonna be able to do the same things that you were doing before. So the question is: If you get an extra hour or if you lose an hour, what is it worth? That has a name. That's called the "shadow price." So let me go to my slides here. Shadow price indicates when you have, indicate the change in your objective: in this case, the total margin. How much does it increase when you get more of the resource? Or how much does it decrease when you lose some of your capacity? Right? So it's a sort of a sensitivity analysis; as things change a little bit, how do things change? And let me skip over this one, and show you when you solve... so let me go back to Excel again. Right here, I have my sensitivity report. So here is -- and I'm gonna have this in my slide shortly. For example, for the engine assembly, you see a shadow price of $2,000. Here's what it means. It means that every hour of engine assembly is gonna get you $2,000. This is true up to 500. This is the allowable increase. And if you were to lose an hour of engine assembly capacity, that's also gonna cost you $2,000 and up to 500; that's the allowable decrease. Also for the metal stamping, the shadow price is $500. That means if I gave you an extra hour of metal stamping, you can make $500 more. And if you lose an hour, you would lose $500, and here are the ranges for it -- allowable increase, allowable decrease. Now, the interesting observation is that the shadow price for 101 assembly and 102 assembly, these are the final assembly steps, are zero, right? Why is that? So if you -- there are a couple of ways to think about this. If I go back to my original sheet, here you see that for 101 assembly I had 5,000 hours that I could have used. How many hours did I use? 4,000. I did not use everything I had. Right? So the -- if I came to you and said, "Hey, why don't we, why don't I sell you more of this capacity? Would you pay anything for it? You are not using what you have, right? So you wouldn't pay anything for it, it's not worth anything to you currently." So that is why its shadow price is zero. Similarly, you are not using everything here, right? So the -- we have 4,500 hours available. We are only using 3,000. So this is not really worth anything. Like having an extra unit of capacity there is not really worth your while. Let me summarize this in my slide here. This is the, this is the sensitivity report we had. And here are the shadow prices. Shadow price is positive only when the constraint is tight. You are using everything you have. We also have this, we call these resources the "bottlenecks." And we spent a great deal of time discussing bottlenecks, how to identify them and so forth -- in the operations management class, especially. So, well, here is the answer: If you know that the shadow price is positive, you've got a bottleneck. Now, here is... our... I think we are going right on time here. So here's the question I wanna look at, I want you to think hard about. And this is a challenging question, right? So maybe it's unfair that I'm putting you through this, but it's also fun. So imagine there's a new product : 103, right? Again, I think they could have picked like Tesla and this and that. But OK, we have Model 103. So now, the question is: Should we introduce this product? But obviously by now, we know better. We need to look at both the financial numbers as well as the production numbers. So the assumption is that this is gonna go through the same steps as the 101: engine assembly, stamping and 101. And it's gonna take this long: 0.8 hours at engine assembly, that's good; 1.5 hours at stamping; and only half an hour in 101 assembly. And the marginal contribution of one unit of this product is $2,000. The question is: Should they produce it along with the others? What do you think? Now I think I, I have another poll here. I'd like to...

Lisa: And once you launch that poll, I do have another question I'm going to pose.

Baris: Sure. Yep. Go ahead, please. I launched a poll.

Lisa: Perfect. While we're doing the poll, Umang would like to know, "Does it also mean that removing the engine assembly bottleneck will make more of an objective impact versus stamping, say through an OEM or weekend work?"

Baris: So...

Lisa: And I can repeat the question if you'd like.

Baris: One more time.

Lisa: Yep.

Baris: Just to make sure I understand it fully.

Lisa: Does it also mean that removing the engine assembly bottleneck will make more of an objective impact versus stamping, say through an OEM or weekend work? And Umang, if you have anything to clarify, I can ask that as well as a follow up.

Baris: That is right.

Lisa: Umang is saying, based on the shadow price of engine assembly versus stamping.

Baris: Sure. So if, you know, if I can do overtime to add hours and it is equally costly, I would put that into the engine assembly because my gain in the objective is higher as indicated by the shadow price. So she is absolutely right.

Lisa: OK, great.

Baris: That being said, there's always more. I wanna go deeper.

Lisa: All right. It depends.

Baris: So not in this case but in general, as you add more and more, things change. You always wanna think marginally because at some point, the bottlenecks may shift. When that happens, those marginal shadow prices will be changing. So you wanna think this through marginally, right? So here in this case, allowable increase is like 500. You probably wanna push it and now once you get there, the question is: OK, what's next? Should I go for more 105s -- I'm sorry, the engine assembly -- or should I try the other one? At that point, more thinking is needed but I'm gonna save that for the actual class.

Lisa: Right. But Umang is joining XP, so you'll see him when he joins XP-92.

Baris: I'll look forward to that. The... now, we have pretty good participation but this is a challenging question, also. Let me give you a few more seconds and then I'm gonna close the poll, and kind of discuss the question. Now there are multiple ways to answer this question. You could even build a model: go back to Excel, add a column there, revise your objective constraints and so forth. Some people do it that way. There is another way to answer that. So that's the one I will do but so let me end the poll here and share the results. Again, most people said, "yes." It's like a comparable number: 62 percent versus 38 percent said, "no." Let me stop sharing this one. So most people said yes. Let's see how we should think about it. So, the answer is no. And here's how we think about it. Imagine -- again sort of, as I was answering the earlier question, Umang's question, I said, "We gotta think marginally." Let's think about the very first 103 that we may, we might be producing. That's gonna take 0.8 hours of engine assembly, 1.5 hours of stamping. And we know that, if I do that, I am taking hours away from my current portfolio. So I can think of, break down this question into two. One is: Well, if I add the 103, I make the extra $2,000 but what happens to my current portfolio? It is, the question then is as though I am just deleting hours from my capacity of those various resources, how many hours am I deleting? It is -- what is given here? -- 0.8 hours from engine assembly, 1.5 hours from stamping, and 0.5 hours from 101. What does that cost me? Per hour? That's given by the shadow price. So for engine assembly, that's $2,000. For stamping, that is $500. And I need 0.8 hours here. 1.5 hours there, and nothing for the 101 assembly, 'cause there I hit excess capacity so it doesn't cost me anything. But when I add these up, I am losing more than $2,000. I'm losing $2,350 by lowering the capacity that's available for my existing resources. So although I'm making an additional $2,000, what I'm giving up here is actually much more so I should, I should not do it. OK? Let me pause here to see if there's any questions. Usually, I get a lot of questions on this one when we do it in person or over Zoom.

Lisa: Let me take a look here. Alexander is asking, "How would you know the minimum marginal contribution of M103 to make its production worthwhile? Would Solver help optimize the marginal contribution, not just the total contribution?"

Baris: Here we have it, Alexander. It is, if it was worth more than this number I would go for it, at least the first unit. So it has to be at least this much. This is exactly trying to get at your question and how would you find that pricing? What's the minimum that you would need? And this type of reasoning -- by the way, it's so tempting to get into, to get into this -- it's very relevant to, say, airline revenue management: how they price different seats and their capacity, and so forth, in shuttle prices. And they call it, they have a different name for it. They call it the "hurdle rate," et cetera. But this is exactly how they are thinking about it. That's ex -- Alexander is right on.

Lisa: Are you using shadow prices rather than remodeling because you are at capacity?

Baris: No. I mean, if, even if I wasn't at capacity, I could still use the shadow prices because it's easier. It's getting late here. I don't wanna work very hard. So if, if you weren't at capacity, these shadow prices would've been zero and the answer would come out to zero here, and I would've said yes. But I could also, let me, because the question came up, I could build that. So what I would do is I would add a column here and say, "OK, here's my third thing," and I would add the data for, like, how we had that 0.8, 1.5. I'd add a column there. I'd have to revise these two formulas here and what's used, and total contribution margin, re-solve it -- and I would get a zero as the answer. Right? So I, I would encourage people to do that to convince themselves if they want to see. I think that's good exercise. But here, because they're equivalent -- and this is kind of pushing, pushing our intuition a bit more -- I prefer this approach. Are there other questions on this one?

Lisa: There are a lot. I'll ask one more and then we can keep going on.

Baris: OK.

Lisa: Rashida was hoping to swap M101 for M103 but that's not an option. Would that be a good alternative?

Baris: Ah, that's a good, a very good question. I mean, we can test to see. I would say that, no, because you can do the cost implication, you can look at -- so there are two ways to answer this. First of all, I know because I built a model where I gave the options of X1, X2, X3, and model ended up choosing X1 as positive, X3 at zero. But that's not an ideal answer. Right? So you can push through the same reasoning and say, "Oh, do I wanna be backing off from 101? What will I gain? What will I lose?" And you will see that you would be, sort of the reverse problem. Do I wanna back off from 101 a bit? And you go ahead and calculate the, how much you'll be gaining because you are freeing up capacity versus how much you are giving up -- which is the total contribution, I think, $3,000 -- and you would prefer to keep producing 101. I hope this answers the question.

Lisa: Thank you.

Baris: But great, great questions. Again, I really wish we could interact more but hopefully we'll get a chance to do that.

Lisa: When you join Booth and you are in Professor Ata's class you can interact much more.

Baris: All right. So just to sort of wrap up, like, the things that we've covered, we went through sort of basics of the optimization modeling, and you -- I hope you got the sense that there was nothing special about this manufacturing setting, and these could have been the different portfolios I would wanna invest in. I could look into their riskiness, and the returns and whatnot, and kind of try to optimize that. But anyway, so this is much more generalizable. And in terms of modeling constraints, we talked about these binding constraints: bottlenecks and the shadow prices. I think this is sort of fairly deep concept, but it's not that hard to get to using, using the tools we have. So with that, I'm gonna stop here. I think we are just on time with 60-minutes mark here. Thank you very much.

Lisa: We are right on time. Thank you for this wonderful session. Let me see if I have any general questions, if you have another minute or two. There was actually a question that came in early on when I was talking about your introduction and your research area. So the question is, In your experience, what has the success rate been integrating operations management research/ best practices into healthcare services, as opposed to non-human facing industries such as manufacturing?"

Baris: That's a great question. I mean, there's a lot of activity currently at Booth. I will be participating in a meeting with our counsel, actually, in the coming weeks. There's a healthcare initiative that I encourage everyone to check out. I mean, there's also this applied artificial intelligence-related center. They also have some healthcare work going on. So in terms of process optimization in the healthcare setting, there's a lot that is happening. I have my own research in the healthcare area. I've been speaking with the radiology department, for example, at UChicago. They've been interestingly using these types of tools to assess their capacity of, like, the medical imaging and so forth; how we can boost things up by process improvements. And that has been fun, and fairly close to the level we are discussing. It's just a bit more complex. I can talk about healthcare for a long time. Like, there's also various research projects in the transplant area that I'm, organ transplant area, that I'm interested in, like policy, you know: the optimal policy in the U.S. and so forth. And more recently, I've been sort of interested in renal transplantation. Sort of growing kidneys in, in pigs, and then later transplanting that to human -- which also intersects with, like, genetic engineering, as well as sort of machine learning, artificial intelligence. Sort of you look at the data to pick, you're sort of trying to do the first human clinical trials there, and picking the right candidates. And so there's a lot of activity: not just me, but within Booth. What sort of more traditional process optimization and so forth, but also some sort of cutting edge -- both in terms of data science and the medical science, and so forth. So I think you'll, you'll be exposed to -- there are also courses, like several courses, taught at this intersection of process optimization as well as healthcare analytics.

Lisa: How about a book recommendation? What's a good book you recommend to continue reading about this topic?

Baris: I would say, I guess you could read our textbook.

Lisa: There you go. I'm gonna share two book recommendations as well because we have our graduating students, that we just asked all of our faculty for their favorite book recommendations and Professor Ata, I'll share the two that you submitted. One is "The Code Breaker" by Isaacson and the other is "The Devil in the White City" -- not related to this topic, but these are two of your favorite books. So speaking of book recommendations, I'll share that. In terms of tools, is the most common analytical tool that you use Excel, or are you using R, Python, other tools in the classroom?

Baris: In the classroom, I do use Excel because it is so easily accessible and really visual. But for my own research, the answer is all of the above. I use R, Python among others, like, we use all sorts of ... And then in terms of optimization solvers, we do use state-of-the-art solvers. So you can solve these problems with like millions of variables and constraints and so forth. And we do need to do that occasionally. So for that, also, like: Excel is great for illustrating concepts and so forth. I mean, just a couple of weeks back one of my former students sent me an email saying, "I gotta solve a problem of this size. What would you recommend?" And I sent him a link; and then a few weeks. Then thereafter, he sent me an email saying, "Oh, it worked. Thank you," and so forth. So we use Excel in this class for teaching purposes. That being said, last year one of my colleagues developed a version of this modeling class in Python. It's being taught currently, and it is, it's smaller in terms of scale because it just started but it's very popular. So it's being taught in the spring quarter in our MBA program. So you can do this in Python, MATLAB, whatever you prefer to do.

Lisa: Excellent. I'm scrolling through to see if there are any other questions.

Baris: Maybe...

Lisa: "Real world, do you find that managers are reluctant or open to implementing such prescriptive solutions?"

Baris: I mean, I do think people are open minded, like if they see the value. I think the burden is on us a little bit to illustrate the value, 'cause it's in everyone's interest to do better.

Lisa: We are receiving some really fantastic comments about your session. So I wanna just thank you for taking an hour of your evening in Chicago. Someone said, "This was awesome. Unlike anything I have ever seen before. Looking forward to exploring more of Booth." Someone else said, "This was awesome, Professor Ata. I appreciate the level you have gone to in order to make this session interactive, especially your translucency feature." You got a lot of laughs from your Jedi component...

Lisa: ...so you couldn't hear 'em, but there was laughter in the Q&A. So thank you so much for your time this evening. I'll put our EMBA email address in the chat box here, so if you have any questions from the admissions team to the audience, I would be happy to follow up with that. I'm also going to plug our next Masterclass Series, which is coming up on Monday, May 3rd. We have Professor Scott Meadow, who's teaching a class called "I Love Cashflow: Building Intellectual Proof of Concept." It'll be a really great session. That's at 6 p.m. Central. We will be posting our entire spring lineup of Masterclass options for you to join. So take a look on our website and you'll see that series. Again, I'll put my email address in our chat box. If you have questions for the admissions team, I'm happy to help. Professor Ata, this has been tremendous! Thank you for sharing your time this evening. You're welcoming us to the Booth classroom -- and to our 300 participants and more, stay healthy, be well, and we hope to see you at Chicago Booth. Thank you.

Baris: Thanks everyone. Bye-bye.

Lisa: Bye all.