 Dear participants, welcome to the course of Supply Chain Digitization. It is jointly taught by Professor Piyanka Burma, Professor Sushmita Narayana and myself Professor Devapratadas from Indian Institute of Management, Mumbai. So, in the last lecture of module 3, which is part of analytics in Supply Chain Management module, we talked about what is analytics, what are the various characteristics of analytics. We also talked about big data, what are its characteristics, various ways of big data. So, now in this session, we will focus mainly on various types of analytics, like how do we define various types of analytics and lot of example from supply chain management domain. Now, these are the main four types of analytics which you can see over here. The first one is descriptive analytics and then we have diagnostic analytics, then we have predictive analytics and then we have prescriptive analytics. So, these are the four important analytics which we have. So, in that throughout the next like few minutes, we will see like where these kind of analytics can be applied specifically in supply chain management domain because this course is on supply chain digitization. So, our focus would be in supply chain management and specifically where analytics can be used. So, let us understand these four types of analytics in detail. So, the first one is descriptive analytics. So, that means specifically if you have heard about descriptive statistics. So, this is nothing but in a simple manner descriptive statistics. It talks about what has happened in the past. So, I have data from the past. If I can capture it properly, I will get to know what is the average value. So, let us say if you have a demand data of various stock keeping units from the last 2-3 years, I can actually find out what was the average demand per year, what is the average demand per month, what is the median demand. Now, how the demand is changing month to month. So, all of these are possible through descriptive analytics based on the past data. That means whatever has happened, you have captured data from those points and then you try to find out how data is changing like what is the average value, what is the median value, what is the standard deviation of the data, is there any variability and so on. So, the good example could be I can use bar chart, I can use pie chart and nowadays lot of good software are coming which is actually creating dashboard. So, I have power BI, Tableau. So, descriptive analytics can be very nicely plotted in the dashboard and as a CEO level. So, let us see you are managing the supply chain of a big FMCG company. I need to know like in which DC, how many units are there, in which warehouse, how many products are lying and it is not being sold. I also want to know in which zone products are sold at a very fast rate. I also want to know in which month, how much was my sale. So, all of these can be easily captured using a dashboard and descriptive analytics can be useful for creating such dashboard. So, that is specifically about descriptive analytics. Now, we have second one called diagnostic analytics and as you can see the image of like microscope, we also have lens. So, this is basically saying like if something has happened, why it has happened, like why did it happen? So, let us say from here the bar chart, I can see that for this product A, my sale was low. So, I need to get to know why the demand went low, what was the reason? Similarly, I can see for this product B, the demand went up. So, what was the reason why the demand of product B went up? Obviously, I will be happy if demand goes up, but I also need to know why the demand went up. If I can know the reason, obviously next time I will try to like do the similar kind of activity. So, that sales of product B keeps on increasing. On the other hand, if I get to know the reason of product A, why it demand, why its demand went down, then obviously, I will take care of it. I will make sure that I will not do that activity in the future. I will take some corrective action. So, maybe some competitors have come in the market and they have taken away my demand. So, I have to make sure that they are like if I have to make sure that I am tracking my competitors movement and accordingly I am positioning myself. So, the demand of A again goes up. So, that is being done through diagnostic analysis. So, if some event happened which are not as per my expectation, if I want to diagnose it, I will be able to do it through diagnostic analysis. So, root cause analysis also I can do to find out why there is an issue. Then the next part of analytics is called predictive analytics. It talks about what will happen. So, descriptive tells what has happened in the past. So, suppose from the past data, I can see that demand of product B is going up. Now, I have to predict whether the demand of product B will keep on going up in the future or what will happen. So, descriptive analytics is focusing only on the past data. Whereas, predictive analytics taking the past data and modeling it and predicting what will happen in the future. So, for product A, we saw that demand has gone down in the past. So, will it still go down or not. So, that is what the predictive analytics will do. Now, the fourth one is prescriptive analytics. So, what is it? It tells me what should be done. So, suppose for product B, descriptive analytics says the demand was very high, predictive analytics said that demand will keep on going up in the next 2, 3 years, then I have to think what should I do. So, let us say demand of product B is going up, going up, going up and I have a capacity very limited. So, obviously demand is more than capacity, then I will not be able to serve. So, I have to expand my capacity. So, the problem should be like how do I expand my capacity? Should I add a new plant or should I outsource it to some third party? So, all of this decision has to be taken through prescriptive analytics. So, this is basically in simple term descriptive, diagnostic, predictive and prescriptive. I am sure like you got some idea about this. So, in the next few slides what I will do will give example of each of these from the supply chain lanes. So, let us take this example part of descriptive analytics. So, first I will analyze the past demand of various products. So, I can use bar chart, I can use pie chart, I can also use dashboard as you talked about in the last slide. So, I will get to know what has happened in the past and as I can see. So, let us say this is for product B, the demand is up and demand is very high. So, through predictive analytics I will be able to predict whether demand will keep on going up or not. If demand keeps on going up as we discussed in the last slide. So, I have to decide first of all that whether should I have another plant or not first decision because if demand is more than my capacity I have to plan for it. So, if because plant cannot be installed in a day. So, I have to plan for it. So, prescriptive analytics helps me to plan. That means if demand goes up and crosses my capacity then the next question would be where should I have my next plant. Whether plant should be near to the demand location, plant should be near to the supplier location and where the plant location should be or to be the capacity of the next plant. If I do not want to have plant I have to outsourced. So, whom should I outsource like who will manufacture my products. I also have to take that decision and if my demand goes up I need more raw material more component. So, who will supply me this component like if there are multiple suppliers if there are multiple suppliers with me then the question is whom should I select. So, out of so many suppliers which are empaneled with me who will give me the extra material and then I have to select decide how much to order from them. So, let us say I selected 10 suppliers from each of these suppliers how many quantities of products should I take it from that is again decision. So, basically optimization. So, in a nutshell we do optimization over here as a part of prescriptive analysis which will minimize my cost which will maximize my profit minimize my risk. So, that all other constants are satisfied and it becomes efficient for my organization to do so. Then we will give another example let us say again on example of descriptive analytics from supply chain point of view there are cases where quality failure happened. So, many a time in manufacturing setup quality failure happens and although digitization happened all the automation has come, but still there are products which fails in the quality inspection. So, let us say I have product x, y, z. So, and I have found out that the product z finished product z which is being manufactured in my factory are having lot of quality issues although x and y quality is perfect, but for product j there are lot of quality issues specific in the last 4, 5 months. So, I am producing 3 products, but only product j is having lot lot quality issues. So, that I get to know through descriptive analytics. So, now once I get to know that product z is having lot of quality issues then I can do diagnostic analytics to check for it like why there is a quality issues I need to find out the root cause for the failure. So, I will see through a microscope and see like what is the reason. So, I can go to the root cause and find out like which part or which component is having a problem. If the product is failed which part is giving issue, which part is not correct. So, after doing lot of this diagnostic analytics suppose I find out that there is a component C 1, there is a component C 1 which is there in part which is there in product j which is causing me problem. So, the part the component C 1 is giving me a problem. So, component C 1 is not of good quality. So, because of since component C 1 is part of product j because of that product j is also getting failed. So, now I need to find out who are my supplier who is supplying me component 1. So, let us say I found out that supplier 1 is giving me component 1. So, then I will talk to the supplier 1 and give them the data. Then the last 3, 4 months my product z is getting failed repeatedly and you have component C 1 which I got it from you is causing the problem. So, then the supplier will be notified will have a talk with the suppliers and then obviously, depending upon the contract you will take a action. Now the question is obviously, I cannot carry forward with the supply supplier 1. So, whom to select as a supplier. So, supplier 1 obviously, I will not go with I have supplier 2, I have supplier 3, I have supplier 4, I have supplier 5 and so on. So, out of this suppliers supplier 1 is cancelled obviously, because of the quality. So, now out of this so many suppliers whom should I select who will be able to give me a component C 1. So, I have to decide that for that I will use prescriptive analytics. So, prescriptive analytics is nothing, but optimization as we talked about in the last slide. So, I need to optimize and find out which suppliers would I select out of so many suppliers which are listed with me who can give me component 1. I have to see what I have to see, I have to see their cost, I have to see their delivery time, I have to see their responsiveness and so on. So, there are multiple factors based on which I would select the supplier. So, I will take this data and optimize and find out who would be my base supplier who can give me the C 1 at a minimum cost you know. So, shortest delivery time having good responsiveness and the last, but not the least the quality and there will be some other parameters also. So, I am not listing it down all the parameters, but let us see these are 4 major parameters based on that you have to decide who would be my supplier. So, this is a second example of starting from descriptive analytics we are doing diagnostic analytics and finally, prescriptive analytics. So, this is another set of problem which is related to the quality failure in manufacturing setup. Then I will give another example from supply chain again and it can be extended to any other domain also this is related to the performance of worker. Basically in Indian context like lot of manual labor work is there although automation is happening, but still like we have lot of like labor who are working in the factories warehouses. So, we need to check their performance. So, how can I check their performance? So, first based on the past data I can see like how they were utilized, how much efficient they were if you go to a let us say sorting facility of an e-commerce company you will see lot of pick like pickers, packers are there who are doing like sorting activity. So, everybody has some goal in mind because e-commerce company wants like 2 hours delivery, 12 hours delivery, 1 day delivery, 48 hours delivery. So, if I have to fulfill that then in the back end I also have to sort the product in that rate, I also have to pack it in that rate, I have to bag it in that rate. So, therefore, these worker who are working in the sorting facility has to make sure that they are following the timeline. So, if I have the past data I know who has followed the timeline, who has not followed the time, who is efficient, who is not efficient. So, the script analytics will give me that idea based on their past performance, who is efficient, who is not so efficient. Then I can develop and predictive analytics model, let us prediction model to find out what would be their efficiency in the future. So, although they are efficient in the past some of them, but will they remain efficient. So, I can predict that and find out in the next 6 months who are the worker, who is most efficient, who are the worker, who are not so efficient, who are the worker, who are mediocre efficient. I can get the efficiency score which is the predicted efficiency score, not the actual, but predicted efficiency score. So, if I have the predicted efficiency score, then whenever the next work happen, I can allocate them, I can do allocation of work based on the predicted efficiency score. So, now, let us say when the Diwali season comes in India, the demand of electronics item, demand of e-commerce products goes up like anything. So, that time I need the employee who is very very efficient, because if you make some delay, obviously, will fall behind. So, I have to make sure the best of the best resources are there available to me, who can do peaking process very efficient, who can do bagging of items very fast and efficient manner, who cannot make mistake, who should not like, who does not make much mistake. So, what happens in the specifically in the e-commerce industry, you will have their customers address and pin code, that is how the sorting happens. And if I pack wrong items and send it to wrong person, obviously, the customers will not be happy and satisfied. Therefore, I have to make sure that at sorting facility, I am sorting the correct items and sending it to the correct pin code and that error is minimized. So, if that is the objective, I can use these till starting with descriptive analytics, which will help me to find out, who are the best performer, who are the worst performers, who are not so good performers. I can predict their efficiency score in the future, I can use this score and find out and allocate, who would be the workers, who would be efficient and which job to be given to which worker. So, these are a flow of like different types of analytics which we use in the context of supply chain and specifically in the worker allocations. So, lot of employees are there, how to allocate the work to the employees, so that they can do the job more efficiently. So, these are few examples. Now, we will move to the next slide, where we will summarize starting from procurement to the last mail, since till it reaches to the customers, how different kind of analytics are being used. So, we have procurement. So, here how analytics are being used, as we saw in the last few slides also. So, I have many suppliers. So, out of these suppliers, I have to select which supplier would be the best one for me. So, I have to see like their cost component, like some supplier may give me the product at a lower cost, but the delivery time will be more. Some other customers may be giving me the product at higher price, but they can deliver the product at a shortest time, also quality will be good. Some customers might give me at lower price, but they are not responsive. So, therefore, how do I find out which supplier to be selected, from which I can procure my raw material, I can procure my component and so on. So, here I can select prescript analytics to find out who would be my best suppliers, from which I have to procure my items. Then the next question is, if you select 3, 4 suppliers, I have to decide which supplier, how much quantity should I procure it from, that is also an optimization model, which can be part of prescript analytics. Then also I can predict their behavior, like in future how will they behave, what would be their predicted cost, if I take the products from them. Then next in the context of manufacturing, once you get the products, raw materials component from the various suppliers, in manufacturing facility, I start producing the products, producing the finished goods, for each finished goods, I have to check their quality, I have to inspect, whether each part is good component or bad component or so on. So, as we discussed that image analytics are being used nowadays, for checking quality of the parts. And in the last class of analytics module, we saw the example of cheap manufacturing facility, where image analytics are being used to check whether good quality cheap is being produced or not. In the manufacturing setup, there is another example of predictive maintenance. So, what happens in predictive maintenance? So, in the manufacturing facility, like many a time, like parts that machine gets broken down. So, if I do predictive maintenance, then I will be alerted that this machine might be broken down in future or in next 2, 3 days, I can take action accordingly. So, this is one good example of predictive maintenance, which is a part of predictive analytics in the manufacturing setup. Then we have warehousing. So, once your product is finished product is done, equality is checked, then it comes to the warehouse. So, in warehouse, lot of analytics can be used. So, we will start with slotting optimization. So, what happens here? So, warehousing typically are large warehouses, the 1 lakh square feet, 1.5 lakh, 2 lakh square feet, huge space is there. So, there are multiple slots. So, which slot, which product to be kept so that I can optimize my space, I can also be quick to pick the items and I can find it very easily. So, let us take an example of slow moving versus fast moving. So, if the item is fast moving, I have to keep the slots in such location. So, that quickly I can pick it up and send it for the delivery. If I keep the fast moving item in the back side of the warehouse, where I need to take lot of time to go pick up from there and send it to the customers, obviously it will be not a good idea, because I have to go back all the way to the back end of the warehouse, bring the product and give it to the delivery location. So, therefore, where should I locate my fast moving items is an important decision. Where should I keep the slow moving? Let us say for a particular item, demand occurs once in a month. So, obviously, I will not keep it near the delivery station, I will keep it at the back side, so that the movement is not being hampered. Then, there are options like zone, zoning. So, in warehouse like clustering algorithm are being used for defining like which zone should be used for which items. So, there are many products like specifically those are managing e-commerce warehouses like few products are ordered together. I know based on my past experience, past data, there are few items which are ordered together. So, if the items are ordered together, I will keep them together in the warehouse, so that it becomes easy for me to picking up during the picking up process. So, therefore, I need to zone, create zones in the warehouse, so that I can keep the items which are ordered together, I will keep them together, so that pickers can pick them together and send it. So, my total picking time will be minimized and searching cost also will be minimized inside the warehouse. Then, we also saw the example of video analytics in warehouse. So, we saw this in the context of inventory count in the last lecture. So, if there is inventory discrepancy between ERP inventory count as well as physical inventory count, that can be done using video analytics like a drone can be fitted, a video camera can be fitted on top of drone, then the drone can go along the aisles and scan the products each and every SQ and comes back and gives me the report of inventory discrepancy. So, these are interesting examples of analytics in the context of warehousing. Then, we have logistics and transportation, here also like a lot of opportunities are there. So, I will give one example which is not discussed in the last class related to the delivery. So, of course, there are some manufacturing facility are there in India, their suppliers are located outside India. So, some of the suppliers are coming from Europe, some of them from America, some of them from China. So, obviously, they come through sea route. So, what happens when you bring the products through sea route? Obviously, there will be delay sometimes because of emergency situation like now sewage canal incident happened, then Panama canal incident happened. Since this is coming through sea route there might be delay. So, let us take an example that you get to know the supplier told you that the product will be delivered to you in 50 days ok, 50 days the part or component will come to you. If it comes on 50th day exactly I am happy because it is coming as per my plan. So, I do not have to change anything. Suppose it comes on 55 days, suppose the product from the supplier comes 55th day, then 5 days I have to wait for the parts to arrive. So, means sometimes it may happen that I have to stop my production because part is not available I have been delayed by 5 days. So, many a time I may have to outsource it from the local suppliers who will charge me hefty amount. So, therefore, it is not good for me. Similarly, let us say sometimes it may come on 46th days also. Let us see was good everything was smooth and I got the product on 46th day. Is it good for me? Yes, no it is again not good for me because my part has arrived in the facility 4 days before. So, I have to build up the inventory and my space will be lost. So, therefore, if the product comes before it is not good because my inventory holding cost have to pay. If the product comes after the expected delivery day that is also not good. So, therefore, I need to make sure that it is coming exactly on the same day. So, lot of analytics are being used nowadays specifically the predictive analytics to predict when my part will be arriving in my manufacturing facility. If it is coming from foreign suppliers overseas suppliers specifically there will be delays because of long distance travel. So, lot of mathematical model AML based models are being used to predict when the parts will be arriving in our facility. Then we also talked about few examples from transportation in the last lecture related to drivers fatigue. So, through image analytics through image analytics I can check drivers fatigue. So, if I can track their eyes movement continuously take the image and there is no movement in like eyeball movements are not there continuously for sometime I will be able to predict that the driver may fall asleep. So, therefore, I will alarm them and make them aware that you may fall asleep. So, take rest wash your eyes and then again drive. Then another interesting example of analytics in demand planning and lot of energy and time are being spent in demand planning. It is a very important aspect and most of the companies will have a demand planning team itself. So, they have to predict what will be the demand and as you talked about like customer behavior are changing. So, they have been seeing lot of advertisement in the TV media about their competitors product brand customer nowadays are very much aware about all the brand which are coming customers are aware because they are using the social media. So, they are getting information. So, customer is having lot of information about each and every product. So, they can take the decision based on their own understanding and since customer behavior are changing and it is being influenced by social media and then advertisement. So, obviously, very difficult to plan for it. So, nowadays demand sensing is very important topic like FMCG industries are actually like they are putting lot of efforts to make sure that they can estimate their demand properly. Not only FMCG, other industries also are spending equal amount of money and preparing data scientist creating data science team to make sure that their planning is done properly. So, if they can plan properly, if they estimate the demand properly the half of that thing is done. So, therefore, lot of efforts are being given for the demand planning and forecasting. So, for each and every excuse what will be my demand? What will be my demand in 6 months time? What will be my demand in 2 months time? What will be my demand in next 15 days? So, continuously I have to keep on tracking and I may change or update the demand also. So, lot of prediction modeling are being used for demand planning and since consumer behavior are changing the traditional forecasting tool may not be applied in every context ok. So, traditional forecasting tool can capture seasonality, it can capture trend, but since there are lot of variability in the demand and if I have to capture those ML model would be a good fit for that and obviously, companies are going in that direction. So, in the summary like what we did in the last lecture and this lecture we talked about what is analytics? We discussed what are the various components of analytics like I need to have data then from the data I will develop model. The model will help me to give better decision and using the decision it should create value for the organization. Then we also talked about big data and its characteristic, especially the 6 V of big data like volume, variety, velocity, variability, value and veracity. So, these are the 6 V's then we talked about type of analytics, descriptive, diagnostic, predictive and prescriptive and we gave lot of examples from supply chain management domain how these analytics can be used. So, I am sure like you have enjoyed this session. So, thank you and look forward to seeing you in the next lecture. If you have any doubt these write it to us we will be happy to answer your queries. Thank you and see you in the next class. Thank you so much.