 Hello global supply chainers and welcome to the wrap up event in SC0x supply chain analytics. My name is Ahmad Ammati and today it means Dr. Chris Capples. So today we're gonna wrap up the SC0x and we're gonna talk about what we learned in this course and how we want to use these techniques that we learned in SC0x in the next courses that we offer for supply chain and after that we're gonna talk about the projects that we have done using these techniques. But before passing the stage to you Chris I am gonna share with you learn some stats. In this round of the course we had 27,000 of students where 1500 of them verified the students. The median age was 29 and they are from 172 countries. Where in India at the most learners 20% well then it was the first time yeah in your past US. So US with 19% and Brazil with 4%. So Chris stage is yours. Okay excellent thanks Ahmad. So what I wanted was four quick things here. I'm gonna talk about what I think you've learned hopefully what you've learned during the course and how you'll use these within the course and then how you might use these in your real in your career. I'll give some examples and Ahmad will give an example as well and then we'll do a breakout and in the breakout what I'm gonna ask you is to think about a problem or a challenge that you've had in your job in your career and how one of these tools or multiple of these tools you think might be able to be used to help solve that. So I'll give you more details on that but you can start thinking about that now. So this is the second run of SC0X and the first runs if any of you have taken any first run courses they're always a little rough because we never know how deep to go or how broad to go. So what you guys just finished or just finishing because they haven't taken the final yet right? Okay now. This went much more focused. I'm much happier with this course. Ahmad's done great things shaping it and it's been turning I think into a very valuable course and so the things that I think you should have learned I put them into four big buckets. The first one is one we've never talked about and that is that hopefully you've gained or recovered your mathematical dexterity and so what I mean by that is most of the things we do later on you use basic mathematical skills that you might have been taught back when you were in grade school or high school or maybe first years of college but you've forgotten and so just getting that dexterity back is kind of like doing calisthenics and getting back in shape so the whole idea of using functions should hopefully you should be really familiar with that now because they're everywhere even if you don't see them immediately they're there and so mathematical functions just mathematical prowess and being able to handle these things simple algebra a lot of people need to brush that rust off also we introduce you to the idea of approximations and approximating things which is exceptionally valuable if you take nothing else out of this the skill of being able to approximate your orders of magnitude is something that will serve you very well over your career just having that in your back pocket but the three big things we did we taught you three classes of tools and these are tools you now have in your tool belt and therefore the you should be sick of me saying that's descriptive descriptive and predictive types of analysis so we started with optimization and we started with the very simplest unconstrained minimizing or maximizing a function with no constraints and you know min max first order conditions with second-order conditions for proof and these are prescriptive models they will tell you a solution an answer that is the best minimum cost maximum profit whatever that is and then we moved into constraint with the linear programming mixed in your linear programming integer programming and a little bit of non-linear programming but the real workhorses here are linear programming where you have minimizer or maximize a function subject to some constraints and everything's a linear function to mix in your linear programs you will use these throughout your career you might not know you're using them but you're using them and hopefully you have an appreciation for those so those are the optimization of prescriptive tools then we taught you a whole bunch of descriptive tools and this is when we introduce probability and statistics because what you learn here is to take data and how to characterize it and with the two big things we talked about you know central tendency and dispersion or spread and that helps you understand the shape of that distribution whether it's demand transit time anything that has lots of data points you can now shape this as a distribution and then we introduce some theoretical distributions that your empirical or you actually observe distributions can map to and the two big ones that we're going to use over and over again are normal and Poisson with a normal distribution right that bell curve with a central tendency with the mean the median the mode the same value and it's symmetric around that and so you have a mean and standard deviation and we'll use normal distributions a lot throughout the courses and throughout your career Poisson's a little bit different in that it can't take negative values it's a discrete distribution you can so it predicts things on a screen and has that nice curve or tail to the right right and so that's usually for lower valued distributions and it has one frame or lambda which is the mean and the standard and the variability so these two distributions of the workhorses there are many many many other ones that you'll use if you go deeper but these are the two workhorses you'll see the most in inventory and the other things in supply chain finally we did regression and simulation and lump those together because these are both predictive tools you use regression to form a relationship between a dependent variable y and a bunch of independent variables so y is a function of x another function that we talked about where the dependent variable say it's demand it's a function of you know the number of people in an area their demand all these other things and so you create a relationship based off that and you're looking at correlations not causation hopefully made that point but what you're really doing is saying this dependent variable depends on these other independent variables and then we introduce simulation and I lump them together because they're used in the same way both regression and simulation will never tell you an answer they won't say you need to do six that's what optimization is for what it will tell you is if you do a certain thing here's what we think will happen so if you create a predictive model regression model of you know the cost of an apartment based on the square footage if I come in I want to know the cost of an 800 square foot or a thousand square foot apartment I can find that cost because I'm I'm now able to predict same thing with forecasting simulation does the same thing if I have a policy and I see how well it performs so it rates ranks or evaluates it doesn't prescribe a solution so we touched our toe in the water for simulation and I showed simple monkey car with simulation in a spreadsheet and then we showed an extended example using sophisticated software for discrete event simulation both have a role and you have to decide which tool which way you want to go but the real key difference there is how you advance the time right for what I showed in the spreadsheet time advanced every day I did a simple simulation based on the demand inventory and sales for the discrete event simulation we did for the airport kiosk that was every event where their passenger came in they left they were serviced by the kiosk so these are the key differences in the types of simulation that you used and they each have a role one is much simpler and easier to implement the discrete event simulation you can do it in Excel or a spreadsheet but boy it's hard you really try to need to use sophisticated software there so these are the four big buckets mathematical dexterity optimization probability and statistics regression simulation so these are now in your tool belt right so yeah so yeah so we have the tool box you know we have the techniques mm-hmm but very in software change we can actually use this that's a great question a lot yeah that's right so so we teach them for a reason we just don't teach them to teach them because there are a bunch of other things that people wanted to add in and we decided not to because we only wanted to teach the things that you're going to use mainly in the rest of the courses in fcx so let me just quickly run down because the next course you guys should take is sc1x which starts this week or next week very very quickly it gets long sc1x just to start just out of week zero zero just yeah so you have time you can take a week off and hit it next week but we do three big things in sc1x and that's where we look at forecasting of demand managing inventory and planning transportation these are the big three in supply chain in logistics so you'll use all these tools that we talked to you into in these three big buckets for forecasting demand you'll use regression because demand many times is treated as a dependence right my demand depends on these other things so we'll use regression a lot there you'll use your statistics to measure the accuracy of your forecast how well you can look at the accuracy and the bias of that forecast you'll use functions all the time that mathematical dexterity will help you a lot when getting to exponential smoothing because it's really bookkeeping but if you don't understand how a function works then you can get lost so we use all those tools for forecasting and inventory it's all about optimization the most common two most common inventory models are something called economic order quantity and a news vendor a single period and each of them are optimization models EOQ or economic order quantity does the trade-off between fixed and variable costs and you're going to see this in every inventory system the single period problem makes a trade-off between cost of having too much inventorying and cost of having too little excess and shortage and so each of these are optimization problems and you use optimization to call with an answer because you're making a trade-off that's the whole thing with sc1x it's all about the trade-offs do I have increased my service versus cost having too much and too little fixed versus variable optimization is used to make those trade-offs to come up with the optimal or best solution and every time also for inventory you're going to deal with a lot of uncertainty because demand is uncertain so you're going to dive into normal distribution more you need to know what is the probability that I will stock out if I have a certain amount of inventory so you'll get a lot of use of the probability work that we did before and then for transportation it's similar to inventory it's a lot of optimization to make sure mode selection you're making trade-offs should I do the faster mode should I send things by truck the cost so certain amount or by rail which costs a different amount and you're making again trade-off between time and money and you'll use optimization for that as well as probability and statistics when you have uncertainty because your demand will be uncertain and your transit time is uncertain so how I combine those together is a little tricky but we'll show you how to do that and you'll be relying on the probability tools that we taught you in sc0x in sc2x we'll dive into supply chain design and you'll use a ton of the mixed integer linear programs we'll use them for network design for facility location production planning as well as aggregate planning all these different tools use mixed integer linear programs in my opinion it's the most widely used tool in supply chain because it lets you make an optimal decision based on a lot of inputs and you'll see it no matter where you go in transportation you'll also excuse me in supply chain you'll also see it in procurement because more and more companies are using optimization based models for procurement so mixed integer linear programs that's a big tool you'll use a lot going throughout the course sc3x we you'll use some simulation both for processes because we talked about queuing a little bit more we talked about that process but also we introduced something called system dynamics how different systems interact and we'll do some simple simulations to show how behavior can change over time also in sc3x we'll do a lot of optimization probability using risk contracts and so that's the whole idea if i'm a a vendor to a company and i'm selling them product i might be able to form a risk contract that increases the total profit for both of us by sharing some of the risks some of the cost and essentially it's building on some of the models you developed in sc1x as well as the tools that you learned in sc0x and then finally in sc4x the last course systems and technology that's all about data manipulation because we move from these small problems to how do you deal with massive amounts of data so the whole data manipulation everything you learned in 0x about that mathematical dexterity it's through the roof now because you're dealing with millions of records how do you handle that and we introduce relational databases and things like that and also in sc4x we kind of add on to your to your analytical tool belt by teaching you some machine learning and this is something that complements all the basic analytics that you just learned so you see throughout the remaining four courses you will rely on the models and methods that you used in sc0x so now i want to turn to the last one the projects okay before that i got your question from asma which is asking about what's the best use of sample or excel so if you're asking which you know software is better you know for different techniques for different problems i would say you can use excel for most of the you know the topics that we have covered but for optimization part specifically for mid-sinteger programming i do recommend to use ample or sets because they're more robots and they are kind of designed to do so the stuff so but excel is really good for visualization you can you know plot you know the data it's very really good for regression and yeah you can use excel for simulation but not for larger scale not for discrete event simulation but yeah yeah no i agree with everything you're saying i think spreadsheets excel specifically is really good for prototyping so if you get something really quick you do some quick analysis on it spreadsheets are awesome but if you're going to do something and you test something i'll do quick optimization models in it just because i find it easier to visualize it but as soon as it gets beyond a certain scale or if it's something that we're going to be using more than once then move into another system back when i was doing coding i'd go into cplex which is another system but ample and sass are great tools for regression i'm starting to use more and more sass i used to use it back in the 80s but then i went away and started using fpss but the the beautiful thing don't hook into to one brand because what we tried to do is teach you the fundamentals you hopefully should be able to jump between tools because we have no idea what the next tool is going to come out but if you understand the basics behind it you should be fine and your company might dictate what tool you use because they'll have the license fee there's some that they dictate the type of tool you need to use yeah i don't want to tell you that's not that's cold but yeah at the time you are using the sass it doesn't mainly a mainframe you know for you know it's statistical absolutely you know yeah but right now sass is a very general purpose yep it's changed doing a lot of change since the 80s i had hair when i used to use sass yeah and the big manuals the big manuals you have to use yeah things have changed yeah okay uh maybe it's time to go to projects if you have time yeah so um so essentially i broke down just three quick examples and these are examples that i've done either in consulting or with students here for companies so the first one is with a company called ch robinson um so hopefully you guys have heard of ch robinson they're a global company they're the us if not the world's largest brokerage company they're really a transportation management and so they actually started in managing agriculture movements vegetables and fruit but now they're a massive company and one thing they do is manage transportation for companies kind of as a 3 pl and so we've had a number of projects with them and the question because there's always an underlying fundamental question and their question is what are the things that a shipper someone who's buying transportation can do to improve their rates so in other words what are the things that impact the transportation rates at the transportation carriers the truck and companies charge so if you get this question you're thinking okay what influences and what are the correlations you should immediately start thinking of regression or some kind of statistical model and so what we've done for them is we've built a series of models over different projects that we've run to try to see okay the dependent variable is the rate the transportation rate that the carrier charges what drives that what influences that where the correlations and so you can think of developing a series of models where you look at the the distance some of the other characteristics of the project itself the geographies and we found things such as the amount of lead time you provide influences it and it's not always linear but what we did here is develop a series of regression ordinary lead square regression as well as some models that we haven't talked about that build on that things like binary loget models which are essentially regression models where instead of your why your dependent variable being a continuous variable it's binary it's one or zero right and so we develop models that say what influence say for example the having on-time delivery under certain characteristics and that on-time delivery is my dependent variable and it's either yes or no so you have a binary and then you have all these dependent variables or independent variables that influence that and you try to find the correlations so we've got a lot of these kind of models for c.h. Robinson where we use all the statistical tools another project we've done recently is with a company called starbucks and you know they you should probably know starbucks if you're anywhere in the world right now there's probably one within a mile of where you are but they're a big coffee house and they they are moving away from not moving away but moving towards being more than just place to get coffee they're bringing in fresh food and so the problem that we worked on is how should can a company provide fresh food on a daily basis at a cost effective manner to a very set of a set of very small stores in an urban environment and so they had different options you could have it coming from the kitchen that makes the fresh food each day to the distribution center that then gets married with the other products that gets delivered to the stores or you could have it as a separate channel or you might have something blended like that so you have all these different options of how you deliver to the stores and what we did there is we developed what's known as a total cost model and that's simply looking at all the cost elements that affect the cost of getting product from a the kitchen to be each of the stores now to estimate some of that we have to do some approximations of local routing and since this was a strategic model you're not going to actually find the exact route that you would deliver between each store because it would change each day so we use continuous approximation similar to what you learned earlier in the course and so we assumed the routing cost based on the density of the stores and the area within each city and so we developed a model and you determined okay if i could get the price down below a certain amount i would run the distribution through the dc or go directly from the store and we helped them determine where the break points when the demand for the fresh food hits a certain point should they shift the distribution method going through the dc or coming straight from the kitchen and so those kind of related questions the tools we used were just analytical cost models but also that approximation method that we taught you before to approximate something that you can't solve exactly last one is a project we did Walmart about a year or two ago and so it's also a replenishment question and their question was how can we deliver product of three different temperature zones efficiently to all of our stores because they have a series of stores that are smaller footprint on these neighborhood markets and they sell frozen goods refrigerated goods a produce which is a little colder than ambient but not frozen not that there's a different temperature zone and then ambient or room temperature goods and they have these trucks that are called multi compartment or three temperature zone trucks where you have an ambient or refrigerating a frozen zone to it or they can have trucks that are all of one zone and the question is which should they use where and how should they plan on using those so as soon as you frame a question like that immediately what you pop up is well we want to know how to do something so it's an optimization problem there is probably a best way to do it and so we developed a mix in germania program to help determine which types of vehicles to allocate to which types of routes and of what quantities and you can see it's subject to making sure the stores get a sufficient amount of each of the temperature zone materials subject to their demand and also to minimize the cost and so that was where we used a large-scale mix in germania program so these three projects use different models and it's all based off of what question you're trying to answer and then think about okay what model what approach should i use for that is this asking for a prescriptive a descriptive or predictive type model so you want to talk about your recent one okay so we just you know we just finished one project for one of the biggest you know multinational manufacturer here in us so together with the master students in supply chain management so what the company asked us was to find a city flow for their you know daily transportation from the plant to distribution center and what we did you know we actually used several techniques that we had in sc0x we use the you know summary statistics so to find the eligible products to to send through the city flow and then use the optimization techniques to figure out how many trucks we need and how to you know build the vehicles and then use a simulation part to to figure out what will happen in the future for the you know what type of uncertainty you know to simulate the uncertainty that we may face in the future and to deal with them so if the model is robust in terms of facing with the uncertainties or not so yeah basically we covered you know different parts of the sc0x in that project so that's a really good point because a lot of these problems they're not just they don't come and say oh i'm a hypothesis testing problem that you're going to use a little bit of everything especially the funny thing about hypothesis testing and all the statistical tests you don't use them explicitly you use them implicitly so when you run a regression you're using them you're looking at t-stats so you're looking at the p values hopefully you have a good grounding on the p values you'll see these all the time but they're kind of part of other analysis but you'll use this the smorgasbord if you will it's an original smorgasbord of these different options this portfolio of different tools and you never know which one you're going to use and that's why we teach them all great so we have the barricades rooms today so yeah so here's where i'd like you to answer discuss the challenge or problem that you faced in work and talk about it with each other and what would be fun is if someone presents a challenge the other people say oh i'd use regression for that and maybe you want to consider that then at the end we'll give them what arthur 10 minutes 15 minutes you have to report that yeah so 15 minutes well we'll send a bell and everything come in send us in and chat what what your problem was and that and we'll talk about also if you have any other questions that for a monitor myself or even for arthur let us know and we'll answer those as well great so please join the barricades rooms and then and you really care great