 All right. Good morning. There's a small technical difficulty. I Wanted to change the slides in the last second, but Didn't work so welcome to the what is it to the sixth lecture and in our course before we start as I promised I would like to Show you the results from the survey that we had about 42 42 42 people out of 120 Registered students voted so that's less than it's about 30% it could have been somehow better, but I'll take these results is representative. Nevertheless All right, so quickly All in all the feedback was Somehow positive as far as My personality goals I was trying to identify areas for improvement for for the course and also for myself and I think I found some I'm going to show you those But before that the quick quick aggregated results. So the questions most of you attended all the exercises I appreciate that they were honest people Who simply didn't care and they were honest enough to to say it? So most of you seem to be satisfied with With the exercises so far this could get a little bit better I hope with the vensim with the vensim models it will get better Well, I actually haven't zoomed in so much actually, yeah So I'm happy. I'm happy to see that Somehow I managed to To explain at least the most important things and Yeah, but that's a good thing Now this is a question which I believe could be improved and I was thinking about how to improve it It's about how the exercises are coordinated with the lecture and and there is a good 10% Or let's say aggregate 25% of people who are somehow on the border between Thinking that they fit and thinking that they don't fit I Will come back to this question when we go through the manual feedback. So this is my favorite question Oh, yeah after after the survey it became my favorite question Yeah, so I mean thanks for that. I certainly do try but This makes things actually more difficult for me because you know with with great power comes great responsibility, so It's very easy from here to go Deep into the ground But I'll try for this not to happen, of course The airport cases The majority things they were they were interesting Still there are people 8% in total who don't agree at all and I was looking for some kind of manual comments to see why I Think I know why So we'll get back to this as well Um This is all right the topics are interesting I believe this could be improved. I did not write all the text in the self-studies, but Now that I read it again and again, I believe there is something to be improved purely in the presentation style So that's gonna happen This is also kind of a good feedback But then again, these were just for lectures So I've been interested in this interesting interested to see how this thing develops this question in particular I see value in attending the the lecture This is good the sales studies should not be too difficult should not take you too much time and This is the most concerning question actually for for me The this this question was not present in previous years I came up with it this time and it seems that I mean I see this as kind of a my fault because There is a good percentage of people who don't see What the purpose of this course is and how each lecture builds on top of the previous one I See this as basically failure to communicate this effectively from my side Yeah, I mean what can I say my goal my goal here is every time to reiterate The overall structure of the course and where we are currently and I thought that would be enough To give you an impression of where we're going where we're heading, but it seems not to be enough So I'll be I'll be thinking how to Maybe change the introduction of the lecture so that you get more kind of Memories from previous lectures the pace of the course seems to be alright Okay, and I will not show you this. I mean I will show you this but kind of a stripped down version of it So these were the the multiple choice questions now we go back to To the lecture and to the manual feedback questions I'm using a crew but reader because something is wrong with my With my graphics driver and other PDF readers don't work for some reason So let's discuss the manual input questions and The feedback that you gave me there. I tried to find out the main themes or the main Kind of concerns that you have the first one is This one no formal way to solve problems is taught actually it's a quotation You see This is not This was not the purpose of the course to teach formal ways to solve problems The purpose of the course is To introduce the problem-solving part Before we go to the systems dynamics part system dynamics is basically about trying to find out How systems work and why they don't work but you don't start from here from that part You start before the system is being built by identifying the problem trying to come up with the better Solution and then designing your system according to the better solution that you hopefully you have found So the purpose of the first part of the lecture was just to give you a flavor that this thing exists That you need to do problem-solving There was a general guideline in terms of the problem-solving cycle how you go about this Situation analysis Selection of solution so on and so forth, but this course is not about formal ways for problem-solving In in in particular given the fact that the problems we tackle they don't have a formal way of solving them It's all about you as problem solvers individual problem solvers You have to go through the discussions. You have to go Through all these reiterations of the problem-solving cycle situation analysis try to find out what your client wants so It's actually difficult to teach a formal way to solve this kind of problems. How do you teach a formal way? To solve the self-study with the airport. I Cannot think of anything like this. So first of all this course Was about giving you the first part of the lecture was about giving you an idea that? Problem-solving needs to be done before you start modeling with Ventsim or writing differential equations and stuff like this So if you expect if you expected this then probably I I didn't communicate this well enough in the beginning and Second of all there is no form away For problems there are general guidelines or frameworks, but not to form away second Some of you wanted more real-life examples of the problem-solving cycle so more case studies like the like the airport This is a good point I'm thinking maybe we can drop one of the self-studies in particular the fourth one with the SWAT Or maybe modify it in a in a way that There is a real business case behind it where you can apply the SWAT and the problem-solving cycle But that would not affect you actually that would affect the next the next batch of students, but it's a good point because I Mean we had two self-studies for the problem-solving cycle Either we can make them more precise so Reduce the scope of the problem or we can make three self-studies for the problem-solving cycle I have to discuss this with with professor Schweitzer and get his opinion, but It's certainly a point that that that I will keep in in mind next one. There was a lot of Suggestions complaints about the self-study presentations in particular You wanted some of you those who gave the feedbacks wanted Less presentations and more discussions so If we have two or three groups presenting it for 45 minutes naturally this cuts down on the time we have for discussion And there were suggestions actually I appreciate this not just criticism, but also suggestions how we can do this maybe just part of the problem is The problems are split among the groups and I was thinking about this actually for a couple of days and This is this is what I think now you see for the first part of the lecture with the problem-solving cycle and this kind of a discussion oriented self-studies I Think we had discussions. I mean we certainly discussed the airport example We we tried to discuss the SWAT the SWAT case But importantly now when we go to the Vensim models and I will address this in the lecture as well Modeling is a lot about intuition It's a lot about developing doing it yourself basically learning by doing there is no way that that you can Modeling can be taught in a way that you you get a feeling How to identify the proper feedback feedback loops how to identify the critical parameters there is no formal way For this there is no theory For this there is intuition. It's like learning to program Some computer language learning to program without ever doing it just by somebody discussing no pointer exceptions and stuff like that so The self-studies until the last the end of the lecture are Going. Oh, let's say What you can get from these self-studies depends a lot on you doing stuff not so much in discussing Because you can have 40-minute discussion on this feedback loop and this critical parameter And you will think you've understood it, but when you get to do it yourself you won't have this kind of Feeling what to do because you've never done it So the important thing for the coming self-studies is that you do stuff And even though we may not have enough time to discuss every single feedback loop every single Parameter that you can vary If you've done it yourself That's good enough Having said that I'm pretty sure we're going to have time to address the most important things in all the models that we're going to see What are the most critical parameters? What is the most important thing that you should get from this model? So this is my my my goal for the coming self-studies But discussions so bottom line is discussions will not help you For learning how to model It's really learning by doing and the purpose of the self-studies is to give you an opportunity to learn this To develop this intuition All we ask for is one presentation and one active discussion. We think that this is the minimum Activity required at least to get you interested and they give you some kind of practical sense. So I Don't see how discussions more discussions will be useful for the coming self-studies and other thing groups were concerned that The value of their presentation is greatly diminished if they're not first meaning that The other group would have probably mentioned everything that they wanted to mention. So they see this as a waste of their effort It's wrong because even if your presentation is identical Well, not identical, of course, but has the the same messages as the previous groups The important thing is that you did it and You learned by doing it and and this is the point of the exercise the point of the exercise is not to polish up your presentation skills But to learn how to do it. So This will not only help you you've done it But it will also help the audience because if we reiterate the point It's just more likely that the audience gets it All right, so don't be concerned if your presentations have huge overlaps So far there haven't been huge overlaps I have to say But even if that's the case that's not the important thing you learn by doing it and that's all that matters Okay, this is related to One of the questions. It's actually a quotation. I don't see where this course is going I'm kind of concerned about this because it's a big failure from my side if you don't see where this course is going and I'll try to tell you where this course is going when we go to the next slide lots of Things about the exam sample solutions sample problems more hints about the exam And then so on and so forth the point is I've already told you more than enough about the exam Everything that is in the slides and nothing more Will be on the exam the self-studies especially the coming self self studies will not be relevant for the exam All right, you will not be asked To sketch events in model and draw lines and stuff like this They're not directly relevant for the exam But indirectly there are because if you get this feeling about feedback loops and their importance and all this kind of stuff Then you will be able to answer conceptual questions Like why is it important that the negative feedback loop has a goal or something like this? So indirectly they are relevant for the exam as As far as your understanding goes and you can only get this understanding from doing stuff Nothing more that is on the self on the slides will be on the exam. There would be no questions solve this Here is a problem apply the problem so solving cycle to it and Write two pages of text. No, it would be if you know the slides if you understand every slide or every second slide You will get a six. It's as simple as that and There can't be any sample solutions to anything I Do I have this point? No It just it just Can't be we cannot have a sample solutions for the airport study or for the predator prey model Okay, last point you don't want to receive model emails Well, nothing can be done about that so You better create the filter or something and then put it in a directory. Yes That's true. That is true. You can answer unsubscribe unsubscribed, but then I Believe when I created the forum I forced everyone to be subscribed because I want you to get all the emails So if you unsubscribe I will just subscribe you back No, see it's not to flood you with emails But I think there have been so far interesting Posts on the forum and even if you don't care about them That's fine, but there may be one person who does So that's why I wanted everyone to be subscribed and it's not a big toll actually on on on you Okay, that's it Where this course is going again? I will try to reiterate and probably I'll change this introduction in the coming lecture We looked at finding solutions to problems. Oh, I have to speed up How to first of all what kind of problems we're dealing with you know, we already know there is no sample solution There is no formal algorithm how to solve these problems. We need human decision makers to find these solutions discussions negotiations stakeholder Analysis all this kind of stuff Implementing solutions come second. So these first two points just give you a flavor What needs to be done before you even go to controlling these solutions? Right, so it's not that you start from Vincent. No you start from from the blackboard talking to your clients What is systems dynamics about let me reiterate? The core goal of systems dynamics is to reveal interdependences between system elements and Having done that to highlight Important feedback loops. That's all systems dynamics is about how the elements are interdependent and What kind of feedback loops exist between these these system elements from a control perspective it's simply We consider the system is a white box as opposed to black box. We put some input in and we want to understand How this input gets transformed to an output in the system right so contrary to a black box we don't We don't just take the system for granted like a weather forecast model You just put some some stuff and you get the probability How do you call this probability density of the weather I guess No, we want to understand exactly what's going on how this input is transformed to an output I spent a great deal of time last lecture talking about modeling We actually didn't have so much time to talk about the actual model from last lecture, which was a predator prey model But this was not important the important thing was to understand what kind of models we're dealing with and What to expect from these models what you should expect is not quantitative or real-life Answers you should expect just highlights of things that are important to you. Remember this picture with the X-ray and the and Ultrasound right You just get a highlight of a little bit of your system So in that sense every model is wrong because you can always get a different highlight and claim that this is more important Then what some other guy did? All right We looked at the population dynamics model. It's a very important model It's a prototypical model which can be applied to biology economic social systems And it's I think one of the first systems dynamics models developed We looked at the rabbits and foxes how both populations Fair in isolation. So the eigen dynamics of of both systems So the rabbits explode or die and the foxes just die if they don't have enough food And then we coupled them together and we saw the interesting oscillations. So coexistence of the two populations Manifested through through these kind of oscillations. This was what what was also in the cell study Now this is Basically this lecture will be about Going and building a model from scratch. We haven't done this yet. What you had the model Ready for you in the previous self-study, but here we build the model from scratch And that's what you have to do in the self-study as well How do we go about building a model from scratch? And this is basically it tells you first you focus on the most important Effects the effects with the highest weight Don't care about the little things the little system elements or feedback loops that that have some influence, but insignificant If you focus just on one dominant effect, that's already good enough Remember you we take the system as a given and we just have to model its elements try to Basically based on your intuition develop your system elements and their feedbacks We do not care about the welfare of individual agents. We actually don't care about the welfare of any of the system elements Individually we only care about the welfare of the whole system, right? So in the population dynamics model It happens that if the rabbits die The rabbit population dies out the whole system basically cease to exist So in that case we would care But there are other situations where it wouldn't be so concerned about how each system element is doing individually, but only collectively What I mentioned with the discussions itself studies a Lot about modeling is is about intuition heuristic approaches so rules of thumb How to identify feedback loops how to identify system elements this cannot be thought As a from a theoretical perspective, this is what you have to do now do it It's it's just really based on doing and learning through doing And the same system the same phenomenon can be described In different ways remember my example about cooperation from last lecture We can explain it with People sense of fairness, but we can also explain it with punishing people so both of them would probably work There is no formal way to model so If you'd like to take this opportunity to learn how to do it, you really have to sit down and do it You cannot rely on 15 minute or 45 minute of presentation about this And we start from the simplest from the simplest Systems and we built Incrementally right we also talked about this last lecture Start with something very simple if it doesn't work if we doesn't generate what we expect build more and more And actually when you have your system, this is an important point when you have your system you can test That system just as you can test the theory Right you say well now If we apply this kind of input we should get this kind of output And if you don't then your model is probably wrong or it needs improvements Just like in any theory you you come up with a theory and then you come up with ways With experiments to test it if you can't test it. So if the theory is not falsifiable It's not a good theory It's the same thing with modeling. So let's start with the model without further ado The model is about workforce and inventory or production and inventory and the situation is the following You observe in your company and and people actually observe in in companies that There is a kind of a cycle between periods of low capacity utilization and High capacity utilization and even not just high but overload of capacities Right. So what is the reason for this? Well, you made say you made say the diet diamond the demand is very Volatile so if the demand is very volatile my sales will be very volatile Air go my production will be very volatile So that's the economic argument Is it it's basically caused by the by the market? however, if you look at Data basically it empirical data what you see is that sales or demand they fluctuate a lot less than production So there is something internally happening that causes this amplification of of Volatility from from sales to production and We want to understand. What is this? What is that internal thing that causes the effect? We build the model starting from the simplest possible scenario. So let's see how to do it first The problem what is actually the problem? Well, the problem is that production is less stable than sales or demand and Here I've shown you I think I've also provided the source. Okay. No, you should have the source for these figures in the notes What you see here is two industries, this is the oil industry and this is some kind of Machine tool industry. Well, let's look at this one This is kind of the oil value chain, right? So you start from the drilling you go through production of Actually the refining refining this thing and then this is basically the petroleum that the gas that goes into your car So look at the drilling and this is kind of growth rates fractional growth rate over Yes per year The drilling fluctuates a lot Right, then you go further the value chain further down the value chain. So this is upstream You know, this is very much upstream you go down you go to the oil and gas production So it's basically that curve This one. All right, it fluctuates slightly less and eventually if you go to the to the demand to the actual demand for In the gas stations It doesn't fluctuate at all compared to to this upstream So that's one example why production Along the value chain so upstream fluctuates a lot Sorry Yeah, production fluctuates a lot less than than sales or demand another example is from this machine tool industry Here we have the So here we have the actual car sales They fluctuate somehow, but if you go up the value chain the ordering of tools for these cars The demand for the production of these tools fluctuate a lot more So basically we're going from upstream downstream and then finally we look at the GDP and Then it almost doesn't fluctuate So that's a problem. It's an empirical fact So something is going on. All right Let's start What would what do we want what an important thing is to Before even sitting down and developing a model you need to define for yourself what you expect from that model When you plot, let's say production and sales, what do you expect? What do you expect them to look like? Otherwise, you cannot test your model at least not in the first at first glance Well, we want to see something like this demand fluctuates a little bit and Production well, it's a very ugly sketch, but you get the point demand fluctuates a little bit and then production fluctuates a lot more We want to get this from our model Well, let's start building it What dynamics is? Capable of producing such a behavior. So let's basically start thinking. Okay. What kind of system elements do I need? Obviously, we have production on the one hand and we have demand or sales on the other hand So we identify two variables production and sales How are they connected through inventory? So we connect them Through inventory now you see You can already spot what are your stock variables and what are your flow variables remember stock is Something that accumulates value over time Flow is something that that is just like instantaneous flow like energy and power right power is just a flow It of energy at any given time and energy is actually accumulation of total consumption of Yeah, total consumption. So basically we connect them like this Production goes to inventory and then we take this is our factory or our warehouse We take stuff from here and we sell it at the market very simple In a different well, this is not really differential question But if you want to put it in a in a mathematical form, it's simply like this The inventory or the stock variable is simply the accumulation the accumulation of What you cannot sell basically Over time right you sum up all the time Simple as that if you want to put it in a differential form You just say the rate of change of that thing. So di dt is basically the difference So the rate of change at any given point of time of inventory is the instantaneous difference Between production and sales at this point of time But it don't if you don't like math just scratch that you don't need it. See there. We can have this nice diagram here All right, but what else do we need so this was This was basically connecting our two most important variables, but What does production depend on now you can think of I'm sure you can think of many many things what production depends on but in the simplest case we start with humans production needs people producing stuff Even though we have a lot of machines nowadays, but we still need workforce so we introduce workforce here and the workforce Mind you we have not connected the workforce Element system element to our system yet. It's just an isolation the workforce Depends in turn on a flow and we call this flow net hiring rate This aggregates all kinds of turnover that you have in your workforce layoffs The retirements new hires all this kind of Turnover in work in workforce is capturing here is captured here. So in a differential form Your workforce changes according to a net hiring rate, which which is constant at any given point of time But of course in general it can be a function It's not necessary that it's constant Let's say 2% over time it could fluctuate and this is basically a function how it fluctuates All right Any questions so far? I think that's that's quite easy The model is very simple Yes, oh Okay, I forgot to mention Now you can think that production depends in the long term So let's say in five years or something it depends on the amount you've invested in factories in machines the amount of capacity to build up or Outsourced or something that's true. That's certainly true and then you can well develop a model which incorporates this But in the short term these kind of fluctuations that we observe We want to explain them by short term measures in a sense so we ignore the long-term effects and We only concentrate in the short term and in the short term what production depends on is is mostly people So this is why we focus on the workforce here Let's connect them now. We simply connect the workforce to production in the following way we say well All these people that we've hired They can produce Stuff with a given efficiency or we with a given productivity, right? So we say the production is Linked to our people to the workforce in that way. So we simply have a productivity measure Which is a constant and we multiply it by our workforce and we get a production right for example if if we have 200 people and each of them can produce two units per day Then we have our production for one day's 400 units simple simple Assumption It's probably a lot more complicated in real life But we want to focus on the most important effects not on the little details remember that And we link them so Our production now depends on the workforce. There it is But it also depends on the productivity of our workforce this little e See that's how they're linked And this is our model Right is anything missing conceptually I'm sure you can add more elements We've already talked about this. You see there is a there is a negative feedback loop here if your sales increase Your production Has to increase but your production would increase by increasing the workforce Your workforce increases by increasing the net hiring rate So in essence you will be able to satisfy this demand this huge Sales demand or this huge demand. So if on the other hand the sales decrease Your production will decrease by basically laying off some people So the workforce is like the balancing feedback, right? It tries to I'll get back to it tries to get Your production to the level that it has to be You know that your production doesn't explode or die down, but it tries to match your sales So it's a negative feedback loop, but what's missing? Yes The assault yes, so why is it? Why is there no feedback between sales? sales and And the hiring the workforce actually there is a feedback We just haven't sketched it the assumption is that managers they observe sales Or they observe the demand and then they they change the the workforce accordingly So there is an implicit feedback that that we just haven't sketched Yeah, of course, yeah, but something is missing and We talked about this When I was introducing feedback loops what is missing is the goal of the feedback loop So that's a that's an important guideline for you developing your models every negative feedback loop should have a goal to what do we want to balance our system and This is how we factor in the goal Let me explain it now we introduce target production So to what we want to match our production. What is the target production? Well, that's basically the demand, right? Our target production must be equal to the demand or to the sales This is the goal of the feedback loop and we call the target production PZ. I Guess Z comes from till or something. It's it's historically in the slides, but it's PZ all right and From here you can derive the target Quantities of all the other system elements So for example, the workforce will be simply the target production Divided by the productivity of your workforce. So WZ and now we can try to Become a little bit more realistic. Of course when you need to hire more people to lay off some people You cannot do this instantaneously. They'll label laws work contracts Especially for Europe, I don't know how it works in in North America, for example, but in Europe You know pretty well that if you want to lay off people that takes some time if you want to hire people interviews Maybe hire the wrong people you need to hire new people. So there is there is time You need basically technical time. You need to adjust your workforce And we call this the workforce adjustment rate or the net hiring rate and We this is basically our towel All right towel is The adjustment time that is needed. So let me explain this equation to you now the net hiring rate are Disregard the towel Imagine we can hire People and we can lay off people instantaneously or just Kind of reproducing from a machine so we get What we need to hire is in fact our target workforce divided Minus the workforce at at this point of time All right, but now if we introduce this kind of adjustment time we simply divide by towel And for instance if we want to hire 20 people For 20 people, but our towel is 100 then we can only hire Well 0.5 Persons at this point of time right so we cannot hire all the 20 people now But we can hire maybe one and then One more two five and so on That's that's the idea of the adjustment adjustment time So that should be all This is the complete picture of the model The only thing we've introduced. Yes, it's also given here, but this is the complete one our sales or the demand Determines our target production Remember PZ the target for so basically they should be equal The target production determines our target workforce WZ, which is basically PZ divided by E Target workforce Together with the time to adjust to that target workforce determines your net hiring rate at this point of time That determines the increase or decrease in workforce which determines your production and so on It's this is how the model looks like and Let's run it you can build this in Vensim I believe you have to for the sales study and when you run it we get that We start well, so let me explain this We start reading this whole thing from here now at that point of time This is maybe 20 time 20. I believe 20 about 20. Imagine. There is a search In demand or in sales so very simple very simple example Your sales increased by 20 and they remain at this level forever Right so from 10 units you suddenly need to sell 30 for basically eternity so if Your demand increases. Let's go back here Your demand increases your target production which needs to be equal to the demand also increases by the same amount And your target workforce would also increase. It's basically production divided by productivity So your target workforce also increases You suddenly want all these people to produce The additional units and you want to keep them forever Because if your sales don't change anymore However, you cannot hire all these people immediately, right? There is a Adjustment time and this is the net hiring rate. So your net hiring rate spikes but then So basically Out of so let's say you were at equilibrium. You were not hiring anything Because everything was matched completely now suddenly you need to hire certain amount of people Given by the difference between target workforce current workforce divided by the adjustment time and then There is some time you can see this time Which needs to pass before you can hire all these people that you wanted, right? So at this point of time you have already hired all the people that you wanted and you stop hiring So you have zero percent hiring Hiring net rate Will continue after the break from from this graph. So And we're right on time that this is good. Okay so we had a nice discussion about some of the survey feedback and Yeah, I'll try to incorporate actually worries. Yeah, I'll try to incorporate this actually in the next next lecture Starting from the next lecture. So I hope that's somehow of an incentive for all of you to fill the second survey Because that really great greatly helps Better discourse in a sense. All right So we looked at The results of our model. So let's continue with discussing the results. Remember, we needed to hire all these people But we can't hire them immediately take some time Until we can hire them and this is the point when we've actually hired all of them This is how the workforce actually looks like Right. So we it's not a pulse It's not a step, but we need some time until we hire all these people like we start from What is that? maybe 90 people up to 150 and We need this time to hire them. This is due to the adjustment to the to the adjustment time that we need All right, so we've hired these people. It's basically the same graph and the production It simply has the same shape. Remember production is workforce times Productivity, so we simply multiply this graph by the productivity and we get our production And I think in this example the productivity is one Right, so they're basically the same So what we see is our sales Jump by 20. I believe so. Yes, our sales jump by 20 And our production adjusted With some time of course and this time is due to the delay in hiring Is this what we expected to see is this what we expected to see? No, exactly. There is no There are no fluctuations right no variability in in the workforce and You have some volatility in sales, but you don't have more volatility in production. This is what we wanted to see So what obviously the model is incomplete, you know, this is how it works you start building a model It's very simple doesn't work you introduce more stuff and now one can think Here where here's where intuition comes into play again one can think well What else do we need? We need qualifications for people. We need this in that But we focus on the most important effects and and one of the most important effects is the inventory we assumed that our inventory is simply like a Instantaneous warehouse, which is just used to ship units from production to sales It's like a tunnel basically right units from production go through this tunnel and then go to sales But this is of course not how real inventories work We most companies I dare say not all but a huge majority have So-called safety stocks or inventory cover inventory coverage This is in a sense the percentage of your sales that you want to have in your inventory. It's it's it resembles the Reserve requirements for banks All right, it's the same thing here So there is a percentage of your sales that you want to be able to satisfy immediately without producing And you store this in your inventory and now the inventory is really like a warehouse and not just like a tunnel So we need to incorporate safety stocks, right and and just is From the previous model. This is how your inventory How your inventory look would look like with a safety stock, right? So you have some inventory Sorry, not with a safety stock, but this is from the previous example If there is a spike in production, then you get your inventory decreases and it doesn't change anymore But this is not what happens in real life. You have to go back to to your safety stock level So let's do this Let's incorporate safety stock or inventory coverage and it's pretty much the same way as before We introduce target inventory I see this one here and the target inventory is simply as I said the part the percentage of the demand of the sales that we want covered Hence the name the inventory coverage and Q this parameter Q is inventory coverage What's what's not clear? You had some questions. All right Yeah It's simple, right? So we have the inventory coverage Q and This is what we want to be in our inventory at all times but then again and By the way, depending on the industry, of course the Q is different, but let's say about 10% is a good Kind of rule of thumb But that's relevant so Just as before we need some time before we can bring our inventory to the desired level, right? So if we suddenly want 100 more units in the inventory, then of course we need to produce them first and that takes time Therefore we introduce again another time delay, which is the Correction time or the inventory the inventory correction time and we call this alpha It's the same mathematical shape as before This is what we want to have in our inventory plus or minus, of course Maybe we need to have 100 more units. Maybe we need to have 100 less units, but it takes some time before we can get there Therefore what is our production now? We want to produce obviously Enough to satisfy the demand or the sales But we also want to produce something in addition to satisfy the inventory coverage ratio So our target demand a target production now is the demand Plus this is what we store in the inventory Just in case there is a huge spike We'll be able to satisfy it and this is how the model looks like This is our demand or sales It determines the target inventory Through the inventory coverage. It's basically this equation The target inventory determines So the inventory correction, which is K So this is the amount of stuff that we add or remove from the inventory at any given point of time K depends of course on the current inventory the target inventory and the correction time current inventory target inventory and correction time This determines your target production and everything else is the same way as before clear, I mean It's basically the same model We just introduced an additional negative feedback loop with a goal. The goal is this one To the model So let's see how it works now We put it in a computer or Vensim or some other software and We have something better Let's we read it the same way as before our sales jump by some amount Yeah, this amount Let's say it's 50 and Then it stays there. It's a very simple case Maybe 50 more people were born and then they want one unit each forever What happens to our inventory correction now inventory correction remember is is This additional Amount of units that we want in our inventory at this point of time, right? So what will happen is Our sales increase Our target inventory or our coverage would increase Therefore, this is what we have right now. This is what we want to have so that difference increases So we need to put more stuff in our inventory because the sales increased, but we cannot put it immediately Right, we cannot put it at this point of time Maybe we can put it over the course of one week due to this correction time. So what will happen is? Depending on this alpha and this is a case where alpha is very small What happens when alpha is very small? Well, we you overreact This is modeling overreaction. So imagine this difference is 20 You had 10 now you want to have 20 30 therefore you need to have 20 more units, but you overreact. This is less than one 0.1 for instance You overreact and you put 200 Right, so you there's a huge spike here then in the next time period you see okay I I have 200 but I wanted to have 20 So I need to remove Let's say 180, but again you overreact a little bit. So again you go like Very very like from here you spike down to here. So you overreact in both both directions When you need more you put a lot more when you need less You take a lot less and then it takes some time until eventually that difference is so small That your overreaction doesn't matter anymore. So there is damped oscillations here. You see and they're still oscillating a Long time after your demand is stable your inventory correction is still oscillating due to this overreaction This overreaction causes your target production to Exhibit this damped oscillations again, right? Remember target production is Where is the target production? It's it basically It's what you need to sell plus this inventory correction, right? So you basically add that thing you add it to your sales Right, so you add this to that basically you add up these two curves and you get this So this is your target production it fluctuates in the same way target workforce is is Given productivity one is the same as your target production and then it takes you some time to hire this workforce. Remember It takes you this time to hire the workforce and finally the workforce looks like this and the workforce is equal to the productivity basically given Sorry production the workforce Equals production remember they were linked with the productivity assuming productivity one Even if you don't assume productivity one they will have the same shape just the scale would change but the point is There is a fluctuation here. So the model reproduces What we needed to see right there is a spike in demand Actually very well behaved spike. I would say And your workforce goes like that or your production Fluctuates like this and what was the cause? Well, two things It was caused by basically two delays delay in your inventory correction and delay in hiring workforce Right, so of course if you play with the delays you can get other things not just damped Oscillations, but you can reproduce Let's say perfect forecasting, right? You don't overreact Right, but you will play with this in the self-study Okay, yes, so this is basically what what I just said By introducing these two time constants now we have two delays and you'll see the delays are important so The model is in essence reproduces things successfully in In the self-study you will have to play with these two parameters See, these are good candidates for critical parameters, right? You can change them in the short term Actually, they are short-term parameters and you see Different effects and this is just an example if you play a bit with the With the alpha and the alpha was the Adjustment of workforce right the time it takes you to adjust your workforce. No, was it let me see Alpha was oh, yeah, it's the inventory correction time So if you play a little bit with it You see here we have huge overreaction This one Right you need Some inventory, but you overreact then you overact again Then you hit we have a little bit more overreaction if you change the parameter and then here we have something more or less stable right, of course if you assume that you can immediately adjust your inventory You would have no oscillations at all, but since that's not what happens in real life We have a good proof a good Case to claim that There is overreaction in real life Yes Yes, yes They are different alphas And okay, this is the the production which is basically go to the workforce remember. I mean they're always equal linked by the productivity This is what the self study is going to be about and again. Let me reiterate this You can only learn and really understand how these feed feedback loops work by doing it It cannot be thought Even though you may understand it you don't develop the intuition So if you're interested, that's an opportunity for you to do it. Let's look now at the real case, however This was just a Mickey Mouse example Let's look at the real case and this is indeed the real case done by McKinsey guys in a company Which is called fast-growing electronics. It's obviously fake name There is a real company behind it though, but due to whatever confidentiality reasons It has not been disclosed. The case is taken from the book of Sherman Systems Dynamics now business. What was it? Yeah, the book Business Dynamics, I think it was called And it's a very interesting case because it happens a lot in real life Let me explain it quickly. We have All right, so we have this company is in the so-called high-velocity industries What does high velocity mean? Well high velocity means that The the demand and the prices are very dynamic. They change a lot in most cases Prices go down functionality is required to go up Feature sizes go down things become smaller with more functionality more complex Product life cycles become a lot shorter Compared to the development times And you can think of industries like that electronic industry computer components Laptops, you know transistors are packed more and more and onto single chips a Product life cycles are in the in the range of eight months And the development time sometimes maybe even more than that. So imagine you have a very short time Where you can make a lot of profits or you you have to make a lot of profits huge margins in order to make up for these huge development costs Which otherwise, I mean you wouldn't be able to operate Right, you have like maybe two or three months to sell your product at the highest margin And then it goes out of fashion It happens with with these things all the time So this is this is the high-velocity industries and the most important Processes in these industries are obviously product development You have to shorten the the product development cycle and supply chains Right, I mean this is basically built from components all over the world So supply chain becomes very important All right, and this is our company It's kind of a Used to be a new company. It experienced the the following Kind of effects huge growth over five years 50% growth per year. So by the way, this is per year Over five years. So 50% per year growth in production 40% growth in revenue and now you can immediately see Declining prices per year, right you produce you ship 50% more, but your revenue or your sales Only increased by 40% why prices go down. That's why and Huge increase in net income as well And then this is a scenario where lots of startups can find themselves in fast growth Cannot be supported by existing workforce by existing processes technology and infrastructure and then they have all these kind of problems long delivery times Lots of inventory build up low predictability of demand What happens is that they give guarantees to their customers Meaning that the customers or they give them the so-called price guarantees meaning that the customer can cancel the order in the last moment You know, that's that's kind of a nice thing you can do to your customers If you cancel the order in the last moment, you're not punished So obviously when they do that This results in low predictability in demand and and all this kind of stuff Quarter volatility that's something very important. Probably the MES guys know a lot about it It's basically you have some targets that you need to fulfill in your quarter And then at the end of the quarter you haven't fulfilled a lot of them So you have this kind of hockey stick phenomenon where you try to Fulfill your targets in units shipped or whatever in two weeks Right and then you start overworking people pushing Trying to find shortcuts how to get these targets The number it's not mentioned here But just think about it the number of stock keeping units SKUs increased by a factor of 35 By a factor of 35 in all these five years So the amount of products that basically they were selling different products and services Increased times 35. It's a huge thing And this was a problem for them In fact by this you can't really pinpoint the problem, right? There is a problem. What is it? We don't know exactly This is where problem-solving cycle comes into play. You'll see it later It's important to say also that Different people different managers had many different ideas about solutions Depending on what they thought the problem was There were so many propositions like let's move to Build to order production Production schedule, right? You only build stuff when you've secured the order like expensive cars are nowadays There were lots of different propositions, but the problem was there was no way to Quantify or to evaluate them Which one would be better than the other ones? You know, it's a similar problem, right? How to objectively quantify these different proposals? even if We agree on some kind of proposals. How do you implement them in what succession? What do you do first? What do you do second? There was there was a Even conflicts between some of the propositions for instance If you there was a proposition to reduce the lead time of Products so to get your materials faster from your suppliers But then of course that conflicts with internal Processes for evaluation of suppliers it takes some time to evaluate suppliers And then if you rush this Obviously this these two things are conflicting. So the base the the main problem was that there was not a problem defined an objective problem and Based on on what every individual thought the problem was there were different solutions. It's a very similar situation for us and what happened is Some consultants from a Kinsey came to help them out and they developed a systems dynamic model for the whole situation and identified What causes all these kind of key problems and what are the main? Levers that you have to pull to address these problems and we'll see this now So these guys came and For two weeks they they spent they spent two weeks talking to people to managers to to engineers seminars workshops This is basically the problem-solving cycle, right? The the first part of this course you have to go in you have to talk to people There is no algorithm for this Yeah, so based on that They found out the very important or very interesting and counter-intuitive result. It seems that all this excess inventory That was building up Was invariant to what the product was in particular Even if you have a slow product Slow product meaning product, which doesn't sell. Well, you've produced a lot of it But now nobody wants to buy it and it's natural that for these products you build up Inventory you build up huge inventory you're unwilling To revise your estimates and to admit that this product is actually a flop and you have to discontinue it So you can understand why inventory can build up for slow products But what these guys find out found out was that inventory gets built up also for hot selling products. So very Products in high demand and that is counter-intuitive it shouldn't be the case Actually exactly the opposite is what you would expect you cannot satisfy the demand and You should not have no inventory basically but this problem was invariant to to the product and This was a very interesting question now. How does this excess inventory emerge for hot products? And you can see now there are basically there is a very easy trap that you can fall into right now You can say well, let's develop a model For slow products and let's develop a model for hot products slow-moving products and hot products and Analyze them separately see what is responsible for the inventory build up for slow products and for hot products with the two different models, but that would be wrong because If you tailor made a model, which is very specific to a given situation then you can basically the predictions or Extrap or extra extrapolations that you can make from this model are very limited because it's only build up for this particular specific situation and you can always do this you can always limit your problem scope and and create a model which is Which is basically useless at the end of the day What you want to do is to make predictions right to to try based on the mechanisms that you identify You try to make larger predictions Maybe policies management practices stuff like that, but you cannot do this if you develop two different models what needs to work Or what needs to be done is to develop a general model and just by changing the parameters It can apply to slow-moving products and also to hot To hot products, so this is this is the way and this is what the guys the guys did but before Obviously you start building a model. Remember we have to define what we expect to see just as as we saw with the Oscillations in production what we expect to see is the following This is how the dynamics of a typical hot product look looks like so let me explain this figure Let's start from here. This is time and this is units units produced It's a hot product remember so iPhone 4s for instance There is you announce the product initially and Lots of people are very excited about it. They really want to have it now So what happens is you get a lot a lot of orders Even more than you can then you can possibly handle These are so cold. Well, I'll talk about phantom orders in a second But you for now you get a lot of orders in the very beginning, right? So you're Inventory your back lock gets filled Your production currently is limited and therefore your lead times Your lead times basically start to start to increase. So the basically the lead time is the difference between the time the order was received and The time the order can be actually fulfilled or shipped to the customer. So as you get more and more orders very fast Your lead times that you offer to your customers increase the back lock basically the back lock is Everybody knows what a backlog is The backlog is basically the the amount of orders of the that have already been accumulated in the in the Manufacturer right and there is a delay between the order is made and and it's actually received in a backlog So it looks basically the same with this given delay. Okay, but that's that's all that's all right So there is a different some delay In your backlog But they both increase because you know it's such a hot product and you simply don't have the capacity to Manufacture a lot of it at the moment And this is your inventory Okay, so let's stop here. Oh, no, let's continue. So with time You ramp up your capacity you ramp up your production and you're able to satisfy all these orders So your lead times go back your backlog goes down but You end up with a lot of inventory Right, this is what we want to reproduce for a hot product Now let's look here on the upper on the upper part This is our demand Normally, that's how demand looks like it's like a bell-shaped curve or normally distributed Right, so initially we have huge demand and then this is the peak of the demand and then it goes down Remember in the beginning everybody wants to order a lot Different retailers different wholesalers. They want to order the iPhone 4s for themselves, right? They know that probably Apple will have difficulties Satisfying all the orders for different retailers So they order a lot only for themselves right the order hundred thousand units and they hope that I Mean it's perfectly rational if you think about it if you order a lot you may hope for special treatment from the company You limit the amount of units that can be shipped to your competitors Right, even though your inventory may increase That's good because there is such it there will be such a high demand for this product that you will quickly get rid of this Inventory and most importantly your competitors will suffer They will not be able to sell the iPhone and you will right so you get a lot of these initial orders These are the orders channel orders Right Here and this is the the build rate so how fast you build your components Right, so you you get a lot of orders, but it takes you some time until you build them Right, this is the time it takes you. This is the delay. This is the lag response to the supply chain So this is where you've satisfied at this point of time. You've satisfied This amount of orders, but look what happens at that point of time here The demand peaks And then it starts to go down so just as You have produced a lot With a given delay The demand goes down all these orders that were initially made get cancelled that's why they're called phantom orders and And you end up with a lot of inventory and The cost is the delay this delay here. Oh We assume that this is the cause right so imagine I Don't know how familiar you are with with this kind of supply chain contracts But when you make an order, of course, you can cancel it Given some kind of terms, of course But if you order 100,000 units, you can cancel half of them and maybe pay some kind of Punishment fee or whatever, but it may not be such a big deal for you So all these orders you think they're real you produce them with a delay, but then they get cancelled And you end up and the demand is actually low Gets lower and you end up with a lot of inventory. This is what we want to reproduce So the guys The McKinsey guys they developed this kind of systems dynamics model Of course, it's very calm. I mean, it's a complex thing. We cannot show the whole model But I will I'm going to show you the most important components of it. All right Okay, so you have explanations for for all these balancing feedbacks. Okay, let's start with here Customers make purchases. So you introduce you introduce the product and All the lots of people make Make purchases for it. So when you increase, let's say when when the retailers or the Whole sailors they assume that there would be huge demand for this product. So these guys They assume that there would be a lot of demand for this for this product They would increase their channel orders. They would increase the orders, obviously they would Order more from you from fast-growing electronics as they order more from you Your back lock increases obviously so you get more and more orders to fulfill. It's like a queue, right? That you have to fulfill The channel saw the debate your back lock increases when your back lock increases You cannot Accept so many orders anymore. So you basically you you say, well We're not going to accept 1 million units per month now. We can only accept 100,000 units per month right and then there's this kind of a balancing feedback loop until both of these get equal But let's start from here initially your back lock increases or your queue of orders increases what happens then You're so full of things to do that you can't really Promise concrete delivery times Right, you cannot you cannot promise. Yes, you will have this product in one week I'm still waiting for my laptop from Neptune by the way So I mean that that's that's pretty much here when the back lock increases your delivery predictability goes down When the delivery predictability goes down The retailers They want to have more of the product in store because they think well Apple cannot ship all these products and We assume that Apple cannot ship these products reliably. They cannot promise me you will have 100,000 units next week Therefore, I'd better be on the safe side and order more Right. I want to have more In my shop just in case right. So this is this kind of defensive ordering So you order defensively, you know just in case something goes wrong and this is a reinforcing feedback loop, right? So delivered predictability goes down The retailers they want to have more at their stores Which increases the the orders and so on and remember we analyze the feedback loops Given everything else is held constant, you know, that's that's the important thing So of course you can say well, but I thought that that this feedback loop was was acting So how can how can this thing increase? From that feedback Of course they act at the same time, but when we analyze them we analyze them in isolation So everything else held constant So this is a reinforcing feedback. Let's go from to the left side. Your back clock increases It's full of orders and so you need to produce you need to manufacture these orders what you do You order materials you restage raw materials So you you buy more Let's say CPUs from Samsung If you're Apple, but yeah with this lawsuits, I don't know if they're going to buy From Samsung anymore, we'll see but anyway, you have to order more materials obviously, but of course you can do this with a delay You cannot get the raw materials immediately Right, so there is a delay here You have a huge backlog. There's a delay you get these materials But then there is an additional delay before you can build these new products you got the materials There is a delay before you can build them After you build them your inventory increases as well Your vendor increases if your inventory inventory increases, then of course you can make more shipments If you can make more shipments Then obviously Let's say your lead time would decrease So you will be able to deliver your products faster. This is obviously not the case with my laptop, but Yeah, we're here as your lead time decreases Your customers are happy They don't order defensively anymore They they're happy if they have less in their inventory Actually, it will be profitable for them to have less in their inventory, you know instead of ordering 100,000 units Now they will order 1000 units per week as they need them because they know they can get them in time and Then you have a nice balancing feedback loop, but of course And in this way you can analyze the rest of the loops for instance your back clock increases in the beginning What happens your lead time increases as I said As your lead time increases So I can promise you that you will get this product, but you will not get it in one week. You will get it in two weeks as opposed to here Which basically means I can probably get it plus minus three four weeks or not minus obviously but either in three or four weeks Right, but here you can probably reliably say no you will get the product, but two weeks later Because the lead times increase this causes Your customers to purchase a hat Right, they think a hat. Okay in two weeks. I'm going to need to be needing 200 units more. So let's order them now Okay, and that's a reinforcing feedback loop And I think these are all the loops. Yeah So this this was the systems dynamics model here And the guys run the model and let's see what they got Before we analyze the pictures. Let me again remind you the model formulation itself took two weeks Right so not just the Interviews and the workshops and the different Meetings with managers, but just formulating this model. It took two weeks And most importantly the guys didn't just say, okay, we had our interviews now We go to our company and we come back to you in two weeks. We offer you a model No, they stayed there and they interacted with all the managers all the people involved in Developing this model. They were asking them. Okay. What do you think? Maybe an important feedback that we've missed an Important process that we've missed That just goes back to to tell you about the feedback that the problem solving cycle. It's very important. All right, and this is what they got This is a slow-moving product hot moving product slow-moving product the historical backlog looks like that The model backlog looks like that The inventory historical inventory looks like this the model inventory looks like this this for a slow-moving product very good fit Hot moving product the historical net inventory Looks like that. Right. This is the in the beginning. You cannot satisfy all these orders, but you have a lot of them Right and then just as you're able to satisfy them they get cancelled The phantom orders and you end up with a huge inventory Model reproduced it quite good. So what are we have for about five minutes? I think so. Let's let's speed up a bit What can we do from this now? Well, we can think about policy input policy analysis What can we suggest to these company now that we have the model is it just good? that we've reproduced some historical data or Can we do something more and what you can do is you can you can? You can basically do two things stand alone policy analysis, which means let's focus on This particular feedback loop and and see what we can do with it or you can offer this kind of integrated policies where you can do a lot of things together and You amplify each all of these loops together and we'll have a look. Okay. For example, you have a an example as Stand an isolated policy. Maybe if you improve your forecast accuracy Or product launch predictability what will be the result and you run it in the model. You see well, we have average impact However, if you reduce the delays in response to supply chain changes, so if you reduce these delays here You have high impact You know, this is already a good result. You would think that predictability in demand would be important Huge impact, but actually turns out that it's average Whereas decreasing your internal delays has a more impact. That's that's a good. That's a good insight already Let's look at an integrated policy We can Do these things and together we can reduce our material lead time So the time it takes us to get materials you can we can reduce our planning cycle and We can incorporate this built-to-order policy for instance And this is just an example. These are different policies Shorter lead time material lead time shorter response time and And Shorter order to receive time So this is this basically the time between you receiving the order and being able to issue an invoice and The point is look at this I mean and unfortunately we don't have time to go through the details in all feedback loops But you can understand them yourself. I'm pretty sure look at all this They're only reinforcing feedback loops positive feedback loops the positive feedback loop remember reinforces positive things and negative things it doesn't care But you do so if you Put a positive thing into this system. It will get Reamplified and reamplified this positive thing. So if you do something positive here something positive here or here or at the same time All these different small effects would get amplified through these different feedback loops positive feedback loops and your result would be a Huge huge impact from this integrated policy one more minute So that that is the example of self-reinforcing build-up So this is just a positive feedback loop which can be influenced By you right so you know it exists. Let's put something positive in it. Let's Let's order more materials Before a product launch Right and this will have a nice positive feedback. This is the last slide Yes Product life cycle it's two slides. We can cover it next time but this is the result from What we can do with the model now think about this You start building a product and you encounter different bottlenecks for example you encounter Material acquisition bottlenecks you want to produce these iPhones, but you just can't get the materials fast enough while you solve it Then you are throughput increases then you encounter a new bottleneck you solve it Your throughput increases a new bottleneck you solve it your throughput increases and so on right so this is attacking bottlenecks as they emerge You can use the model for this, but what you can also do is you try to predict the bottlenecks From your model. This is using the model to anticipate the emergence of bottlenecks and By doing this of course, this is an ideal picture, but by doing this your throughput your process throughput Would behave much nicer much smoother? Okay, so We're going to start next lecture with quickly showing you the last two slides from the project life cycle. It's very simple But I hope You got the message today. Thank you