 So I think this is the inevitable result of being the last presentation of the last day. Ok, so the reason I am doing this is because I am from an analytics background and I am very passionate about the application of analytics in functions that are not traditionally using them. So if you think about Binance, if you think about Binance, if you think about retailing, the connection is very obvious and a lot of innovation and a lot of stuff around analytics happens in those areas but where I think analytics can be of immense value is human resources and that is the topic of my talk today. So just before that, so I represent Bridge I2I, this is the start of the new start of a year ago because our sole mission in life is to embed data and decision making as a practice within organizations embed isn't the DNA so that it becomes as important as having a HR team or having a finance team. So that is precisely what we are working towards and we think because that becomes a practice where the amount of business outcomes that can be gained and doing things better can improve by in all the magnitudes. So that is where I am from. So if you think about, so startups don't have this problem because you know everybody pretty intimately but I think most startups want to become large organizations because that's the dream that you have when you do a startup and I will also become 100 million, 1 million organizations having hours of employees. I will be the next Facebook, next Microsoft. So the thing is there is one thing common across large companies is we have a lot of people and we have a lot of people, all of them say that people are the most important assets which is true but there is too much of it. It's an art in itself to kind of manage it. So and I have seen this across service organizations in India which typically have thousands of people where the pressure of the HR team to hire is so immense that basically whoever comes and applies they are very glad that somebody applied and they get the name. So the problem is of quantity but the problem, even bigger problem is of quality and they simply don't have the time to focus on the quality problem which is am I hiring the right profile for the job? Am I even thinking about where this guy is from? Am I even thinking about what job am I hiring for? It's the first problem. So the second problem is typical of large organizations. Nobody thinks that they are paid enough. Large organizations think that they are paying you more than what you deserve and therefore how do you reconcile this problem because you have so much of money that you need to give and that's always fixed and then you have a lot of ways by which you need to divide up that money and you get totally confused on what and where should I go by in terms of dividing this money. It's a huge problem. From what I've seen it is like there are so many venues that play that you but most HR folks either just give up and say it's like everybody just grab what you want or they go by one dimension totally ignoring the other dimensions. The third is of course the obvious problem which is how do I keep people back because most organizations fail because critical people leave them and therefore it's a huge lift in their performance if they're able to keep good people back. So that is the third and these three things pretty much would keep most HR managers. This is what they think about all the time. So I don't think there's anything else that really captures their imagination more than these three. So if you think about what is HR analytics, it is mostly nothing great. You don't have to invest in a lot of stuff. It is just a decision that you will use whatever data that you're already capturing. You're capturing because there is a organizational requirement to do so. Your parent in the headquarters told you that these data has to be captured or the government requires you to capture such data. You have the data with you. So it is just a decision to use that data to take people-related decisions which are as simple as that. So HR analytics means I will use the data that you already have to take better decisions. And second is basically realizing that most problems that seem really difficult to solve seem really impractical actually can be solved, can be addressed to a large extent if you just look at the data that you have. So let me go to the first problem which is the higher effectiveness problem. So these are from assignments that we have done. Within 30 minutes, I'm just presenting an overall view of how we went about solving the problem. It doesn't need, unlike other sessions where it's sat on, it doesn't need Python, it doesn't need Adobe, it doesn't need anything. It's just need because it's for people who are using the HR profession or leaders who are worried about their people who can do this pretty easily with very rudimentary tools. So the first one is the Fritman conundrum which is, I have roles that I need to fill in my organization. What are the things that I consider when people apply? They have a certain qualification. They are from a certain college or an institution. And location could be two things. Location could be where they are from. Location could be where, which location you're hiring for. So these are typically things that are attributes that you consider for filling the job role. And these are typically things that you have data on when you're filling these job roles. Now the point I'm making is when you've done this hiring, if you've captured this data on which job role I filled with what kind of people, that is enough to kind of solve this problem. Because after that, once these people come into your organization, they start a journey in your organization. So there are things that are positive variables, things that they're supposed to do, which is you expect them to stay for a longer while because once you train them, they have to contribute back something and if they leave before that it's a loss. The second is the rating is another measure of how good they are. Third is how fast they get promoted is a fairly good indicator of what kind of leadership potential you have. Against that are the cost variables, which is the money that you're paying them, attrition, which is how fast they leave and all the other things that you do to keep them in. Now all you have to do is if you have this mapping done, in terms of which role did I fill for which kind of profile, you fixed up the matrix and then you kind of track these variables individually for each person. You can pretty easily quantify these because if you look at these variables, all of them are pretty quantified. Promotion, you can quantify them as an average promotion rate, which is during their tenure, how many times they got promoted in how many years or another variable that will help is time for first promotion, which is after I hired them, how long did it take for me to promote them. It's a good indicator of how good they are because of the background that they came in because once as you spend more and more time in the organization, the organization's variables itself starts affecting them positively and negatively. So because they walk into a good organization, they might become better inside the organization but that's not the problem we're trying to solve for hiring effectiveness. So time for first promotion is a good indicator for how good they were when they came in and if you kind of set it up against the cost that I paid them or all the things, how long were they available before they left or in which qualification group or an institutional group did I see attrition most. It's pretty simple to then construct a very effective data-driven hiring effectiveness engine. It's as simple as this. That's it. You just combine these metrics to form a score. Obviously you have to set off the positives against negatives. You can do it however you want. That is why you have weighted some organizations, value the cost now, the money more some organizations, value the tenure more some organizations, value the rating more so you can give advantages as you feel like and then a good way of normalizing them is you can use what's called a Z score or average dispersion around the mean. So because what happens is not all of them are comparable. So tenure could range from 2 to 7 years whereas salaries could range from 2 lakhs to 20 lakhs. So they're not comparable as units and therefore you should normalize them in some way a very effective, easy way of normalizing them is using what's called a Z score and average dispersion around its own average mean and you apply the weightages and then you're done. So you use this. You can pretty much know for which role, which kind of background works and that's it. So then you can see that the kind of people that you've seen that you've hired for the role you can save money at the same time you can also increase the positive variables and minimize the negative variables. The second one is the compensation problem which is how much do I pay? Now as I said how much do I pay has got various elements inside that most large organizations are not complete metacoresis nor are they complete socialist republics. There is a good measure of both. There are obviously goals that you start with. You're supposed to achieve these goals before you become before you are ready for getting that much amount of increment or whatever and talking about increment because an initial salary is pretty easy to scope and most organizations think of how much you've got before where you're from and they use a light grid and that's very easy. The problem is how much do I give as an increment from a certain pool that I have. So there's goals and those goals are twofold. The business itself has a goal which determines the pool that the business gets as well as the individual has his own performance which is given by the rating and these two can be against each other. The business could have done badly, the individual could have done well and that's when it gets interesting. Then you have tolerance by gold type. So there are some goals of which you might not brook any slacking. There are other goals which you would understand if an employee doesn't do well and therefore it's important to measure the tolerance that you're willing to live with against each goal and like I said there's a performance rating which is the employee's performance rating and then you have the differentiation philosophy which is some organizations believe that the best performer should get far higher than the average. Some organizations believe that it's like a linear scale and some organizations believe that there should be no differentiation. So this can very easily be coded in terms of when you combine all of these performances with the score these can be very easily numerically coded. So for example differentiation philosophy you can either say a red gets 0 average performance which is knowledge gets 10 and a green which is a great performance gets 20 or you can say it's 0, 50 and 100 or you can say 0, 10 and 100. So it depends on how you want to differentiate. So as long as you do that you can very easily code or numerically derive and combine the performance against goals both for the company as well for the individual. Then you have the budget where again the constraints are typically whatever you've done or you've not done, the amount of money available to the business principles as increment is pretty fixed and that can't change. And of course then you have typically a direction from the larger organization saying no matter how bad a business does you should not give less than 3% actually or however well you do you should not give anybody more than 12, 13%. So such allowable spreads are always there and then finally against all this is the problem of how do I make sure that people that I have to keep are satisfied. I have a certain brand image in the market as far as salary is concerned. I usually target to be a 15% higher, if you target to be at 75% higher market etc. So that is another consideration. So typically what happens is in most organizations they will kind of sacrifice or downplay one against what they want to go after. The point being that if you take the time and energy to code them upfront in terms of numbers and if you combine them numerically then you don't need to do anything like that because all you have to do is you put all these inputs and Excel has a smaller engine into which you can put the constraints saying you know what happens the money should not increase so much, what happens somebody should not get so and so and then you can also have constraints around I need to have top performance of a certain proportion average performance of a certain proportion, bottom performance should not get anything all these constraints can be input and then you can pretty easily solve this problem and I am surprised as to why organizations don't do this they kind of, I mean in the HR function especially they pretty much go along one dimension kind of sacrificing the other one. So this is very easily done using Solver provided you do the upfront work to measure or capture all of these things in a numerically way because typically what happens is in the end if somebody is done not well they will say I don't have the data regarding codes or I don't know where to find that it's all in PowerPoint, it can't be coded now in even Excel so you have to decide to set it up, capture them numerically, have them have the tolerances agreed and then it's after that the biggest advantage of this is in the organization that we implemented is you can give this as a guidance to the supervisor who is sometimes if you have large teams you can have a team of say 35 people and sometimes confused about what I should do about some inductions because some individuals stand out in your mind you know that they are either very good or very bad the others you are at a loss as to what should I do with them and then this recency effect and then whether I like their face or whether they are related to my wife things start happening so this is pretty easily defended you can say this is why I did what I did for you and therefore there is no, the scope for argument is reduced significantly their argument will obviously be there because if I have to understand something which is bad for me I would refuse to understand so that is always going to be there but from the organization standpoint you can pretty easily defend your decision and you can give this as a guidance for the supervisor the last minute they can do some pretty, they can do some tweets they can say you know I know that this way is in deep trouble he's been in this for seven years, I want to give him some more money all that you can allow but to start with if you have an institutionalized system that is data driven in its very design it is very, very easy to allocate a large sum of money over a huge amount of money the third is the whole question of trying to understand why people are leaving your organization and I think the problem is as old as HR itself where you try to understand why people are leaving so typically the problem looks like this which is shit everybody seems to be leaving here I don't know why they are leaving but generally everybody is leaving here or you would have a hypothesis that would have formed in your mind saying oh they are leaving because we are paying them enough or they are leaving because some competitor who is working for me has gone and tried another company and they are acting like this so your gut feeling or emotions would drive your reasoning behind why everybody is leaving and then the scope for taking needs of reactions increases the problem with needs of reactions in attrition is it tends to exacerbate the problem the more you employ needs of reactions the more people will leave so you will catch five people and say I want to give you back yours money don't tell anybody obviously all the other people would know and then the five other people who you didn't catch will leave otherwise you would not have left so what in the work that we did it's very simple again I mean all the things that I do are pretty simple and basically all you have to do is cut the data by various dimensions the dimensions could be any dimension could be designation the dimensions could be location the dimensions could be work stream the dimensions could be qualification any dimension that you have data on cut the data using the dimension to then study the attrition problem in 99 cases out of 100 you will feel that not everybody is leaving everybody is obviously not leaving what is happening is that problem is pretty much present in some dimension so you will have here where it's at average I mean this is either something that's happening at average or something that doesn't is not too important here this one segment where you would find that nobody is leaving they're pretty happy on this dimension there's one group which is cleaning in a pretty high proportion compared to the overall average and this could be typically it could be people who are you know analysts who are at the lowest level where it could be people who are you know not getting promoted anything like that any dimension that you capture would have and it will pretty much be a picture like this now what you need to do is once here it's easy to come to the conclusion that okay yes all of these people are a problem we need to take that we need to solve this problem for them but that's usually not the case the story is usually a little more nuanced where the minute you find that there is a certain segment having a bigger attrition proportion than the rest of the population you need to drill down into that segment what you need to do is you need to cut that block into other variables that you can split them into for example if there is this illustration is I've just first cut them by the rating that they receive and you'll see that the bottom performers obviously are leaving and it's a good thing because I want people who are not performing well to leave the average people are staying the people who have rated a little less than the topmost rating are also staying which is a good thing but most alarming is the people who I rate the best are leaving a huge proportion so that among these four this is where my biggest problem is so then if you cut that using other variables like among the people who have given higher rating if I cut them into compensation quartiles what do I see you see that people who have got the best rating who have been paid in the fourth quartile are staying back and the people who are in the bottom quartiles in the topmost rating are going you understand that among the people who have rated the best the people who feel that they are not getting paid as much they are used to or that they deserve are leaving second is among the people who have rated best people who have not got awards so obviously there is an entitlement issue they've got the best rating but they have not got any awards therefore they are dissatisfied and they are leaving second is promotion people who have not got the best rating are not promoted are leaving so if you cut it if you just keep drilling into the data and you find that you will find that where is the problem localized and the interaction of which variables is the problem localized and therefore this enables you to take action this you can't take action everybody seems to be leaving it's like you know just run this enables you to take action the last what helps is if you do some simple workloads around exit interviews that you conducted you will find nuggets of information that are never there in your hypothesis you would have most people if you thought that they are leaving because of the money they would not have left they would have said I am leaving because I am relocating I am leaving because of the management so my point being some of them could be a very obvious hypothesis in your mind but this will enable you to get things some things that you never even knew so this is the last slide basically it's an idea of what we are trying to build for companies which is we want which are to be a data-driven organization like every other function and therefore I think the lift that it gives more than the outline and the bottom line is more fundamental I believe it will give you better people I believe it will give you more engaged people and you will have the satisfaction of having done the best to the last available rupee or dollar that you had expected precisely see if you see this who has gone out of the organization I can't do anything you've left, you've left so what I am trying to say is over here the green and the red capture attrition proportions in the sense of block which is the top rating people the attrition proportion over there is much higher than ours absolutely so now I know people who got the best rating and were not getting paid so well are the people who are leaving so I'll find others who are like that in my group and do something about it so that is precisely what you raise precisely the objective of this exercise which is if I find groups within my organization where attrition is happening more than average I can capture other people like that and address that problem so exit interviews are for people who left so obviously from the beginning of time people who have left there is nothing that you can do about them but if you had an exit interview process and you just capture data and just kept it in a file that's not going to help all I am saying is if you had that information with you and if you mind that information it will tell you why people are leaving that is going to stop you from stop or it will enable you to stop people like them to leave so the attrition problems always manage the people that you that was still there it's going to manage the people who left so yes absolutely so this one it's not just an exit interview it could be an employee satisfaction service and these two things are not at the same time this you can do at any time I am not saying this and this has to happen I am saying this you can do anytime you just take the data that you have and start analyzing it tomorrow and start seeing where people are left the main foundation for this is you should have captured this data because typically the answers that you get is who knows where it is so I am saying it's a waste because you captured it and you are not using it to solve your own problem even when you ask the formal capture the informal data somewhere and then insert it into the data see which is where so this is nothing to do what the employee is feeling right now this is just behavioral data that you can capture from data that you have and it is not so even if no high performance telling you that you are getting poorly paid this one will tell you that there is a problem so you don't need to depend on anybody here exit interviews if you think about them are easier than employee engagement service because if the guy already decided to leave if you put him in a comfortable position where you are not trying to pressurizing him too much he may give you more useful data now that he decided to leave than employment engagement service the employment engagement service there are better ways of doing it you make sure that you communicate and you state to the fact that you are not tracking them individually if you make sure that you don't capture data individually you are only capturing its summarize and if you are and it is amply clear by the process that you are requiring that nobody in the organization is going to know that me and me are thick set so and so is this bad then it's fine I mean it's also got to do the culture that you built up front with the police state kind of culture then you are dead first chance people are going to get to me fine I am glad that somebody is there the audience that I am yeah so I my advice would be start with the things that you think are different for example if you have 230,000 employees spread across 9 countries or so start with segments that you think are different for example you would know that an analyst is different from me right so spread them by back you would know that Romania is pretty different from India so you read them by country start with that and then go into the other segments that you have data because the first ones are usually the ones that you know I have to cut them and then when you combine them with the other things that you that you have with you but you don't have any upfront hypothesis on that is when the story gets interesting and you will find things that you never thought it is that's interesting right but the thing that I tried to do here was not predict predictions I think prediction is a more involved in science which is pretty interesting in itself here I am just doing a post hoc analysis to see can it keep people back so that is your in attrition prediction you are trying to be prepared for how many I should hire to back up the people there definitely this data would get written back yeah absolutely see that's what I am saying so most people depending on the culture that you build would think that I should not tell them the organization I am going to or I should not tell them that the manager is a problem because he is going to get me he is going to invite me to my new company so it's also a cultural thing I think it also starts with if you convey the fact to the employees that you are generally trying to do something about a problem I think that is when you will get good information over there and it's like I said exit interview is probably the last hope because he has I would think some of the companies would think of being open because they decided not but obviously there are people who wouldn't come out even then and then this and they would help you sorry no so I am not used psychometric data to begin with because my first premise was let me see the thing is the first especially if you are going into a client organization and saying do these red things before I start doing something you are going to go out after the problem so my first premise was what all can I do with data that they already have sitting in if they have done some so for example some of them have leadership development programs some of them have this training programs where they would have collected psychometric data and if that is there it will help but what I work with were typically things that everybody has and therefore I thought let me start with this but what you are saying is fairly interesting because then you can bring in the psychological angle of how they are feeling to what the data is saying and that is pretty interesting because then you can start correlating what they are feeling and what has happened and there is a linkage they are behaving in a dr. Jethy and Mr. Hyde kind of way where they are saying something and doing something so that you certainly I think social media that somebody talked about is pretty interesting you can see what people who have gone out of saying in social media try to mine that information correlate that with what has happened here so those are interesting things that you are thinking of doing for the subsequent assignments try to interrupt the closing and feedback questions about the science and all that yes you do have this data right even if it is a fresher you have these right even if it is a fresher you will certainly have these so I am saying you will need more information for example I would love to have what is your overall work experience what is the company that you have had before so that is all more better information the first thing I started to work with this is something that everybody will have and therefore what can I do with this so this itself if you just slot them into a matrix and start doing a linkage matrix wise as to who works for what role you will find that some of your assumptions large bands back office the master of arts is doing on all these parameters they are doing much better than a CA or a commerce graduate why because they are happy that they got this role it is something new and interesting for them and therefore they are performing better this is something that you will never start off believing right that is what these things can help you it is very simple analysis that can help you get these insights you