Okay, let's continue [COUGH]. So now we're going to talk about business analytics. I'm pretty sure many of you have some experience about business analytics. You heard about this term, but maybe you don't exactly know what that actually it has a tight connection with operations research. Okay, so today everybody talks about business analytics, big data analytics bra bra analytics. And you may even see that many MBA programs has changed their names, traditionally MBA means master of business administrations. You go there you study accounting, finance, marketing, operations, management and so on. But today, those MBA programs generally becomes Masters of Business Analytics, still called MBA abbreviated as MBA, but they are changing their curriculum. They tell you how to use Excel. They tell you how to write Python. They tell you how to write Java. They tell you how to run a database. And then they tell you how to do statistics, something, something and something and use those quantitative methods to solve business problems. Sometimes you see they are called Master of Science in Business Analytics. So, the name may not be the most important thing. The most important thing is what's inside business analytics. So some people use this framework and I have also my own framework, but basically that's pretty much the same. So let's see what is this. Many people say business analytics has three things, three parts, or three levels. The first thing is about description. I want to see what happened in my company. So in that case, I'm going to take some historical data. I'm going to make some plus Calculate some statistics. I want to see what's the average daily sales in the past ten days, something like that. If I do that, that's descriptive analytics. I can use those information to help me do some more decisions. If I see the sales is declining, if I see a plot something like this, This is time, first day second day, and this is sales. If I see a trend like this, the trend is declining. I know I need to do something that's descriptive analytics. Predictive Analytics is the second level If I see a trend like this, as a human being, I do some forecast. And I'm going to say okay tomorrow, the day after tomorrow, I'm going to predict that things will happen. It will keep declining. So I may need to do something but in this particular estimation or prediction may not be accurate. Sometimes they are just too many information. The weather may also affect my sales, whether there will be a holiday may also have some effect, whether my competitor is going to increase or decrease their price may also have some impact. So in that case, people may use predictive analytics to have some prediction module, forecasting module to help me know what's going to happen in the future, the module will tell me what will happen, with some historical data. And of course, there must be some prediction algorithms, prediction models inside and so that's predictive analytics. So today if you hear about machine learning, big data, AI, whatever, pretty much they are doing predictive analytics. And lastly, there is something called prescriptive analytics. So what's that? If I see tomorrow, the day after tomorrow, the sales is going to decline. And if I want to measure my profit, probably I will decide, okay, I'm going to decrease by price, because that may have some help with my quantity. That makes sense, but when we are thinking about what should we do. When we are thinking about decreasing the price, should I decrease my price by $1 $2 $3 ?We don't know. Then if we have a module that do some optimization, do some calculations, combining historical data future prediction and the some kind of optimization tools. That module may help us determine, what should we do? What's the best amount of price reduction we should take for tomorrow? We may make that decision by ourselves, but If you have that prescriptive module is going to make some prescription for you. And it will make some suggestions. That decision model is the heart of prescriptive analytics and the probably you already see that's pretty much operations research. With that operations research tool, people can use that algorithm to help you make some management make some decisions and help you determine something you cannot determine. Help you decide was the price reduction, how much to purchase costs to be sold, and so on, so on and so on. That's a decision making tool. So, if that's the case, then very quickly we may understand some examples here. A typical process of doing those operations thing or business analytics things is two steps. You first collect some data, input them into descriptive or predictive modules those information is going to help you understand the problem. Understand the environment that you are in, understand today, yesterday or understand the future. Those two modules will help you do this. And then operations research is going to help you allocate resources and solve your problem. Most data analysis things are important, because they help you know what's going to happen. And then operations research is also important, because it helps you to know what should we do. So let's see this very simple example. Suppose I'm running a retail store and I sell a lot of things. I sell drinks, I sell cookies, I sell coffee, I sell stationery and so on. In that case I need to determine the inventory level, okay? For example, if I sell Coke, I need to determine how many bottle of Cokes I want to put in my shelf space or in a warehouse. Because my store is limited, the space is limited I need to make that decision that as soon as I only have one product, in that case, the idea is simple. I don't want to store too much. I don't want to store too few. Because if some customer comes in but I don't have product to sale, then that's understocking and I lose some money. If I prepare too much, and no one comes, that's overstocking I also lose something. So I want to find the best balance between these two things. Data Analysis is going to help us estimate those random demand is going to tell me okay with probability 10% I guess tomorrow we will see 20 customers with probability 20% we guess tomorrow we will see 100 customers and so on. Data analysis or those prediction modules is going to help us understand what's going to happen what may happen with some probabilities for tomorrow. Then operations research plays the role of decision making. Is going to consider the procurement cost, the shelf space cost, personnel cost, product price probability for among of some amount of people to show up, combining all these information and then tell us what's the maximum. Now what's the best inventory level to take? So prediction or decision, data analysis or optimization, or operations research? That problem can certainly becomes even more complicated if you have multiple products right? In that case you consider one thing called demand substitution, what's that? Some people may show up into your store and say, I want to buy a bottle of Coke. My god, there's no Coke, then some people will go away, but some people will change to buy something else for example, Pepsi Alright? So in that case, you will know the amount of coke and the amount of Pepsi that decisions are not independent, they should be connected in some way. In that case, the problem is very difficult, because if you want to do data analysis, you not just estimate how many people are going to buy your coke. You also need to estimate if someone cannot get a coke what is the probability for her, for him or she to buy a Pepsi. Okay, that probability is not so easy to estimate. Even with today's technology you need to do a lot of things. And operations research in this case is also very difficult, because the typical retail store have hundreds of items there. You need to have a module to decide the inventory level for all these items at the same time. If that's the case, that's very, very difficult. So I don't want to solve these problems today here because, as I mentioned, is too difficult. This example is just to show you the relationship between prediction and optimization. All kinds of machine learning modules, AI modules is doing predictions. They try to predict what the consumers need, what's the demand, what's going to happen in the future, whether there is a cat or a dog in your photo, and so on and so on. With those information Operations Research try to arrange your resources, and make the best decision that's operations research. Okay. So, basically we have tried to tell you about the relationship between two things. Either you do data analysis or you do operations research, right? And for data analysis, sometimes you do description, sometimes you do prediction, all these things are very important. So actually in practice, if you really want to solve something, then all these process in the whole must be done. Some people are more, interested in doing data analysis. Some people are more interested in doing a operations research, but basically may help each other. So in that sense, when people really want to contact the research, conduct a study to solve a business problem. There are some process to go through. And in particular if you want to do OR studies, then let's take a look at the typical process. So, your project always start by collecting data. You look at the data you have you try to understand the current environment. Try to understand whether the sales trend is increasing or decreasing. That helps you to define the problem that you want to solve. Now, if you see a downward trend, you see that your products getting unpopular among customers, you need to get attention from customers. You need to refine your products, you need to do something like that. If you see an upward trend, then you consider different things. They're going to decide, well, I want to open a new market, I want to sell some other things, something like that. So you take a look at the environment you are in, and then you define your problems. After that, you then you then ask yourself, do you need to know more about the environment? For example, if you see an upward trend, everything goes fine. You want to go to a new market, you need to know something about the new market. You need to know how many people living there, are they rich or not so rich what's the product they are used to buy and so on, will collect more data and then try to solve problems. When you start to solve the problem, we basically do two things. We formulate a model, something that we're going to show you very quickly. After we formulate a model, we use some mathematics to describe our problem. Then we solve the model which is also something that is very important in this course. By formulating a model, solving a model then we can get a solution. Something suggested by the model. The model is going to suggest, okay there are 10 different candidate companies and 10 different countries. Which one should you go first? There are 10 different new products that you may push into the market. Which one should you do first? If you take a look at the model, sometimes you feel, okay, the model is bad, the solution is weird, that's possible. Because every model may be too far from the realistic situation, okay? We always see some real situation, and then make a model. But the model may be too abstract, too single, or sometimes too complicated. If your model does not reflect the realistic situations enough, the suggestion will be weird. In that case, we need to go back to refined our model. Refined our model, solve it again. Take a look at the solution refined. So take a look, refined, solve, take a look until we think okay, the model is good enough, then the suggested solution will be interpreted, and the people can really make some decisions. Okay, so I think this particular flow chart we can see there are so many different things to do and here I highlighted these two circles defining the problem and interpreting the solution and the making suggestions. So in this case, in this sense, these two things are critical. But in this course, instead, I will focus on these two parts. So you may think about well you say this is important, but you want to teach that, what problem do you have? The reason is the following. Suppose you are new to operations research, you need to first, and know his theory, know his underlying principles. Know what it is, know the technical parts of operations research, less about modeling less about problem solution. So that's something to learn at the beginning if you are new to this field or new to this subject, only after you understand what the tool may do, how the tool works. Then you may start to think about how to use the tools, okay. So, some what you may say this part is talking about theory. This part is talking about applications, something like that. If you are currently an undergrad students. Master's students, PhD students, or if you are 25 28 you just go to the society you just go to companies for one or two years. I think this is the time for you to take understanding about the theory part about the tool itself. And then you may start to try to apply the tools. That takes some experience that takes some domain knowledge. That's not something we may teach you in an online course. But the theory is important because only if you have the concrete understanding about all these underlying things, then you may start to do applications. So, if you really have some time, I really encourage you to try to follow our pace to take a look at this course. Then you will have some deep understanding about models. Then after you have understanding about these models, whatever industry you are in whatever companies you are in, you will find something that you may do. You may do better with operations research. And of course, before we really jump into operations research to tell you a lot of good things, we need to give you some short warning. Analytics certainly is not everything or it cannot do everything. The good part is that we use mathematics. The mathematical language is precise and concise. That helps us to define a problem in a way that there is no ambiguity among people, and that there is no ambiguity from the computer's perspective. If you want a computer to solve something, you need to define the problem in a way that computers can understand. Pretty much that's the model or mathematics, okay? So OR is good in this sense, it also facilitates the use of computers to solve a problem. As I just mentioned, if you want the computer to help you solve something you need to first tell computer what the problem is. In a way that a computer can understand that is using modeling, okay? So later we will see some examples. The disadvantage part is that, well sometimes a problem may be too hard to be formulated into mathematical models, or sometimes some critical information may be missing. So let's take that multi product inventory problem as an example. I mentioned to you about demand substitution, right? I say when someone wants to buy Coke, if there's no Coke, someone leaves someone buy Pepsi, someone buy seven up someone buy tapioca, milk tea, something like that. So, in that case, you need to know the probability for dimension session. You need to know well 70% of persons will buy Pepsi, 20% will go away and the 10% will buy blah. If you don't have that information, whatever model you have, whatever OR tool you are using. Is useless because they cannot solve your problem in a way that is realistic. But it's really possible that you just don't have that information, right? So sometimes OR just cannot help you if you don't have that information. So, we know there are something we can do we know there are something we cannot do. The aim of this course basically is that, through various examples, we want to tell you what may be solved by operations research and what cannot be solved by operations research. Hopefully after this course, we all get a concrete idea about this. Then when you go on to your job, when you go to your company, you may take a look at the problems that you are solving, and then tell yourself, this is OR problem, or this is not OR problem. OR cannot solve everything, so we need to have an understanding about what it can do. Then when we apply the tools, it helps us instead of is discriminant everything. So that's our objective. In the series, we have multiple courses about OR, but in the first course, this one, basically we focus on models and applications. So after the whole course, after the six weeks, you should already be able to answer these questions.