Predictive Analytics for HR and Workforce (10/19/2016)

well welcome everybody and thank you for taking the time with us to be with us here today on our webinar this webinar is being brought to you by UC Irvine and today we'll be talking about predictive analytics for HRM workforce but before we get going I just want to give you a little idea how the webinar works we do have the audio lines on mute to kind of cut down on the background noise but we definitely want you guys ask questions so you should see on the right side of your screen either a Q&A area or a chat area we monitor both of those so please feel free to type questions in there we will be answering those along the way if you don't see those there tabs up on the upper right hand side of the screen on your screen you should have the control screen visible to you so please use those and there's there she is and Jon just confirmed you can hear me okay well we are very fortunate today to have greta roberts with us greta roberts is an acknowledged influencer in this field of predictive workforce analytics she has used her continued vision to bridge that gap between business predictive analytics and the workforce communities and since she founded talent analytics in 2001 she has established the company as a globally recognized leader in predicting an individual's performance pre-hire and that's a really good time to assess how well somebody's going to do before you make that commitment and again naturally that leads to dramatic and measurable business our o is she uses this greta uses predictive analytics to solve not just HR issues but really line of business challenges which affect the entire organization and this is proven to be a very effective way to approach defining goals for predictive analytics and workforce she leads her company in developing predictive solutions that can really honestly be easily and elegantly deployed into employee operations even when those teams may not have backgrounds and statistics or math or analytics and that that's where the real power is in addition to being a contributing author to numerous textbooks she regularly is invited to comment in the media and speak about high end predictive analytics and business events around the world she is also the chair of predictive analytics world for workforce which is an innovative and it's an annual predictive analytics event dedicated to solving these challenges as well as all that she is also an instructor and faculty member at the Institute of analytics in forms and also here with us at the University of California at Irvine Greta it's a pleasure to have you here thank you so much Dave and thank you for inviting me to present today on one of my favorite subjects obviously that would be using predictive approaches in HR and the workforce and I like it so much that I even agreed to do this webinar on my birthday so that says something so I wouldn't all say thank you yes yes it's a cheap it's a cheap way and get some happy birthdays obviously interest in this area is growing by leaps and bounds and I love being a part of helping to advance this area so again thank you for inviting me here I thought it would maybe make sense to begin by kind of defining what is predictive analytics for those that might be joining us that you know this is kind of the first time try to really understand what it is so and I like breaking us down into a way that we can kind of all understand so predictive analytics just is quite simply using facts to help make decisions and I sometimes say future decisions but then I realize all decisions or future decisions so but using facts to make those decisions and when you're talking to predictive people or to your data scientists sometimes you might hear the words like outcome and the first thing that you would do when you're talking about predictive analytics is to ID if I the outcome that you want to predict and because of how predictive analytics works it's very important for you to be very specific about the outcome that you want to predict but that is a word that you would hear a lot is outcome what is it very specifically that you want to predict the second thing would be to say hey you know if I could look into the future and I could predict the outcome that I'm looking to predict then I wouldn't need to do any of this predictive analytics but we can't do that so instead of that what we say and really this process to say let's look at the data that's around the outcome that seems like it exists a lot when that outcome also exists that are kind of proxy data that we would call inputs to say let's look at the data around the outcomes and is there a way that we can see that when these inputs exist it tends to have a higher probability of the outcome also happening and so that would be something called inputs once you've defined your outcome and then you've defined what is the data that we have that might exist as inputs that we can also study then there's techniques that are used to find and validate those patterns in the data that actually do predict that outcome and the next two points are extremely important predictive analytics is the process of not only you know identifying inputs and outputs but also looking to say we want to deploy these as well I don't want to just put this in a PowerPoint I don't want to just talk to people about the predictive model I want to deploy that model and something to think about early on is how are you going to deploy that into your daily decision-making your entire organization is not filled with data scientists who have degrees and statistics etc and so is there a way that you can easy easily deploy that and the last point that we have here is that you want your model to learn it's called machine learning based on how it does in real life again with predictive analytics you don't just deploy a model and say we found this prediction let's let it run you know till the end of time you want it to make sure that based on how it does in real life it goes up I got that wrong I got this right let's let the model get smarter and smarter as we go forward so that's kind of a general definition if you will of predictive analytics and I have kind of a metaphor here that might even bring it home a little more closely the decision well first of all predictions are everywhere we may not be as aware of all of the predictions that are happening around us that include us actually or that we are actually making but one that we make a lot is you know is it going to rain unless you're in Southern California but is it going to rain and do I want to take an umbrella with me so the outcome that we want to predict is is it going to rain and so we tend to look around us for these different inputs are there great clouds you know is it raining do we hear thunder are the weather people on TV or on the internet saying yes or no it's not going to rain or not and so eventually what happens the more input that you have that are pointing to yes yes yes yes yes you know you eventually get to the decision that you say you know umbrella or no umbrella but this is just a very classic you know the outcome is I want to predict if it's going to rain the inputs you know what are all the things that could exist as well when it actually does rain and I just think this is a great and very easy and elegant way to kind of understand how predictions and predictive analytics work so this is one that really affects kind of a thing rain but the other thing that also is everywhere is predictions about human behavior and I think a lot of times people in HR and the workforce believe that this is really new trying to predict something in the human behavior area but it's just simply not so as an example for a very long time insurance companies have been predicting the probability of you wrecking your car or and that really goes to how much they actually ask you for your insurance premium right because maybe I have a much higher probability based on how old I am or where I live if I live in the city if I park in a garage and all those different things so those would be inputs versus if somebody has very different input so they've been predicting human behavior for a long time paying a mortgage obviously the mortgage lender or the bank wants to figure out are you going to pay your mortgage or are you going to default on that that's also predicting human behavior performing well in a job we want to predict that when we extend an invitation to somebody to join our company and hire them and so we're looking at all these inputs it's the reason we asked them for their CV we asked them for that you know come in for an interview for references those are all inputs that we're looking for to see you know let's gather them all together hoping that in somebody's brain were able to you know consistently come up with a prediction whether using predictive analytics or not when you hire somebody you're predicting that they will be a high performer in that job and so you know even compliance taking medicine in the healthcare area again those are humans and that's human behavior so you know can we predict that somebody will take their medications and if not is there a different behavior that we would enact for that individual to make sure that they're well taken care of so predict this is just the smallest predictions about human behavior are all around us it is not new in terms of us bringing this into employee or the workforce space so there's a lot of excitement and I hear from people all over the world really saying you know I've been told by my executives we need to do something predictive but they haven't given me sort of any definition about what we should do so you know what's the excitement about you know why would the executives being put the pressure on the folks in their organizations there's really only one thing that generates this kind of excitement and that is there's better results that they've seen that intuition there's a single reason and that's the only reason better results better business results and so if you're going to do a predictive project what you want to do is deliver back to the executives what they're looking for which is better results better business results than intuition or or then chance alone so I wanted to talk a little bit about intuition and sort of how we're using that today because intuition really is looking at info as well we don't just make a decision to you know let's say get that umbrella just jump out of bed and go yep today an umbrella we look at things like I you know told you earlier and when we're making a decision about hiring somebody or promoting them or making other employing decisions our intuition favors think people that we like people that have common experience with us that went to the same schools or have a similar background or grew up in the same neighborhood and the problem with this bias or intuition is that it has a lot of bias in it that isn't backed up by data it isn't backed up by necessarily results and it's got this inconsistent set of rules that it applies unlike predictive model models our intuition also it's really hard to track results I know for myself when I make a decision I don't write it down and then say you know when the result happens I'm going to come back and check and see how my decision was because I need to be accountable for that and kind of tweak my decision going forward we just don't do that and so from an intuition perspective we are pulling in all of those inputs it's just that there's no accountability we don't traditionally track our results and the final thing here the final point is that we traditionally don't learn and improve our decisions over time there's really not a consistent way to give that feedback back to us so that our model inside of our head our intuition model doesn't learn and improve over time I wanted to talk a little bit about a really interesting bias study that was done that kind of goes to our intuition and the question was really is there bias in musical auditions for symphonies and you know to me I always thought you know what I wouldn't imagine that what they're looking for is really these professional musicians that you know based on their musical ability they would be selected to play in a professional symphony and so there was some research that was done and published in the princeton weekly bulletin to see to see if there was any bias so what they had was everybody played their behind a curtain and they noticed them that when they would come in to play their audition they were actually still wearing shoes so that you could tell if was male or female or somebody wearing heels are not wearing heels so they had them take their shoes off as well so that they couldn't tell if it was male or female and here's what they found there was significant unconscious bias with Symphony musicians it happened so frequently and it continued until the auditions were held with the musician behind the curtain and also with their shoes off blind auditions then increase the probability that women would advance from the first round it increased it by fifty percent and it wasn't conscious bias the people that were there were just using their intuition and taking in all of these different inputs and for whatever reason gender was one of the inputs that they were looking at and so if a predictive model were running there and if they you know that model was able to just really look at great musician and how they paid you know played the music that was something that would have been screened out because predictive models don't care about those sorts of things some people when they're talking about predictive analytics in the workforce or with HR tend to get really scared and they think it's this great harbinger of doom and we don't see it that way at all we think see if it see it as a real forward-looking innovation for one of the reasons that I already mentioned or a couple actually one is that it just doesn't care a data and algorithms don't care about gender they don't care about skin color education or any other personal details unless that's an input that you're including in your predictive model and we highly recommend that you don't but as long as that is not you know put in there they don't care about that they only care about the ability to predict the outcome that you've said I want you to predict the outcome and then that's what your model is going to go looking for I also love the idea since the workforce and the HR domain is running a little behind in terms of maturity some of the other human prediction arenas or domains we've always loved to go to some of those other domains and see what we can learn from them and I think it's a great example to look at the mortgage lending and see what happens there because it can really mirror what happens really in a talent acquisition example I think it's a really nice metaphor that we can consider so and again the source is down here on the right-hand side if you want to read more so in a mortgage example the very first thing that you would come up with would be the outcome and the outcome for a mortgage lender would be what is the probability of defaulting on payments and so you know they want to know where you're going to pay or they want to know you know are you going to default that's the thing they care about so it's very clear to them you know there's nothing wishy-washy so the first thing they would do would be to gather input data points they would look at very specific things like your credit history or your duration of accounts that you've had them or you know how much you use your credit cards or any collections activity have you been bankrupt and other things that you know would seem to be normal I wanted to also point out that by law they cannot use things like religion gender national origin marital status and other personal information which is very similar to by law what Human Resources is not allowed to use for making decisions and I bring this up because I encounter a lot of fear with human resources who believe that their privacy issues I think they believe that they're greater somehow than other privacy issues that others human areas have had to deal with but every other to every other domain that is predicting something about human behavior has had to deal with these same privacy issues and we don't want to get we don't want HR to really either use it as an excuse or to get paralyzed and not you know and say ah there's privacy issues or there's disparate impact issues and we just can't move forward so again don't use this as an excuse and realize that every other domain that deals with predicting human outcomes has had to sit down and figure it out individually what this means for them and how they're going to move forward so once they've figured out and gather the data points the next thing they would do is say okay we have our outcomes that we know we have the data points there are input data points and now let's use machine learning to look for patterns between those two really to answer the question what is it about who pays and who defaults is there something different about the people that default and the people that pay which again would be very similar to what you would do in a talent acquisition environment and sometimes it's going to be a lot of what we see on the left hand side that there is going to be something related to you know disinformation that's excuse me this information that's here the credit history the number of accounts I mean this is really your core data however there might be surprises and it may not always be what you think it could be that things come out and this is a very interesting article like if people start to ask for cash advances or they begin to buy lottery tickets when they didn't before or they begin to retread their tires or they begin to pay the IRS with a credit card there's also some surprises that are going to be in there which sometimes I think when you're working with your data scientists on Workforce Analytics predictions it's interesting a lot to talk to them about the data inputs that you're going to be gathering so once you have the prediction and your machine learning has produced a prediction what you definitely want to do is go through a process called validation to answer the question is the model going to perform for you outside of the laboratory and that's really what you care about you don't care if it performs in the lab but they need to prove to you and on any work that you're doing in you know in your work force you want to make sure that they go through this step it's a rigorous validation process with real data to say you know prove to me that this model is going to work outside of the outside of the laboratory and at another time I could get into a lot of detail about how you go through the validation process and then the final step is deploying the model and again if you go all the way up to validation and then you don't deploy the model to us we believe this is you know not a success of the project finding a prediction who cares about that it's a lot of wasted time and so selecting the project you know that you want to move forward with and then actually deploying the model in the hands of in this example the lenders where the probability is published for a new borrower or a new creditor the commercial lender would make that daily decision and then they would have that continuous tracking and improvement to learn from real life life results so that that model gets better and better at predicting over time who is going to default and who's actually going to pay their mortgage okay now a lot of people focus on what we call two desirable outcomes and they do this with something that we call correlations is there a correlation and the two desirable outcomes would be where they approved a loan that was successfully paid and so people would say that was excellent because it was successful and so they're like woohoo this is great that's the desirable outcome another desirable outcome would be you know I denied that would have defaulted and that's an excellent outcome as well because the bank avoids a loss I just want to mention to you that there's two other outcomes as well and they need to be paid attention to so you want to go beyond what we call correlation and so one outcome would be if you approve a loan that what it was defaulted on obviously that's a negative outcome or you deny a loan where somebody actually would have been a really great lender for you or you can actually apply this same thing it's something called the confusion matrix where you can apply it for hiring or you know Workforce Planning or any of the other predictions that you tend to do so a lot of people just keep it at these two outcomes but we like to look at all four and again at another time and actually at UCI we're talking a little bit more about all four of those outcomes and what they cost to your organization so now let's take what we've seen with another domain which is on the mortgage application side and let's talk about how you can lay that on to HR and the workforce first of all there's two different kinds of predictions that can happen inside of HR in the workforce one is at what we call the aggregate level and one is called the individual level at the aggregate level this would be where you have a prediction that would predict things like staffing levels things like you know in the next three years we need an increase of 25 percent of skilled engineers and that really allows you to you know know how you need to staff up between now and three years or other workforce planning areas that you need to pay attention to right now or it could be the best time to contact candidates again if you have a prediction about you know we should do it five days after they apply or three days or again that's a prediction that you know kind of affects policy or affects you know processes or the ideal times for reviews or promotions or raises or which training delivers best results that's at an aggregate level and again it affects policies and it affects processes the aggregate side also tends to feel and a little less personal because these are policies and these are processes that are being put in place and you can definitely use predictive analytics for aggregate or trends level predictions the other kinds of predictions are ones that are done at the individual level and they would affect candidates and they would affect employees to an individual candidate or to an individual employee so they're going to feel more personal but sometimes the ROI is significantly more when you're looking at the individual level so the kinds of predictions that you would get at an individual level would be you know what is the probability that if we hire this candidate they are likely to be a top performer for us or you know identify of the employees that we have which ones maybe if we have to go through a layoff you know are the least likely to be top performers for us going forward or which of our current employees should be put on target for succession planning or put into a predictive job padding scenario or you know predicting employee lifetime value or what and whom to invest in again even as I'm describing I'm sure that you know the folks are thinking wow that's really to an individual level and so when people are talking about I'm a little nervous about predictive analytics and HR and workforce issues it tends to be more at the individual level versus at the aggregate level and I'll talk in a little bit about some ways that I would recommend as you're going to go forward might be some good first projects to begin you know rather than going directly to you know who are we going to layoff that might not be the best first project but start working on ones that feel a little less personal as people get comfortable with modeling and how it works an ideal predictive analytics outcome on any project and especially on your first project remember that the reason that the executives are putting pressure on your team's to do something predictive is because they want better business results and so if you do a predictive project that doesn't deliver better business results there's going to be a missed and they're going to be you know unsatisfied with the results that come from that so an ideal predictive analytics outcome is going to be one that minimizes current expensive mistakes or maximizes business outcomes where they're currently going up we're just our employees are not performing in this area or we're losing so much money in this area a less effective predictive outcome especially for a first project and I would say a large percentage of the customers that we talk to do these projects and make these mistakes so it might be something that you may want to think about I think because people are nervous and it's kind of new and I get that you don't want to sort of throw yourself completely into something you know is there a way to just sort of test you know a predictive project and there really isn't you kind you can't really just do a half predictive project so I think one thing you don't want to do is to do something that's not really actionable but it's just kind of interesting so as an example for people to you know people to predict what someone's engagement score is going to be next year you know you could go through the process you could go through the whole predictive process you could have the outcomes the inputs you know the whole thing and then you deliver that and people are like yeah all right whatever and again if you think about the reason the executives are putting pressure on your company right now it's because they want better business results and predicting somebody's engagement next year is not going to deliver that to them I'm going to do something a little controversial as well and put in here nothing actionable just interesting and include predicting what someone's flight risk will be next year because a lot of companies are engaged in predicting the current employees flight risk either this year or next year or three years from now or whatever it is the reason I include this is because it is interesting data and a lot of people are wanting to do that I have at this point I have seen one project where people predicted somebody's flight risk and actually implemented a program to do something about that you know for their existing employees so there's a lot of money being spent and consulting and projects and a lot of people spending time on predicting what somebody's flight risk is if that's existing employees flight risk but it's the wrong time to predict somebody's flight risk you need to predict that before you hire them it's too late once you've actually brought them on board and again I've only seen one project and they did a very good job of it so that's a little controversial but something you want to think about a left effective predictive outcome or predictive project – is and people do this a lot and this particular project to discover insights that you don't want to know a lot of people for whatever reason and I think it's because HR is told to do a predictive project and there's really two kinds of predictive projects as well one that solves an HR problem and one that solves a workforce problem which is why we call this HR and workforce predictive workforce and HR analytics so predicting you know setting up a project that is going to predict you know is our diversity and compliance you know what's it what's it going to be if we continue to hire like we do and have the results come back the outcomes saying that yeah as you do we can predict this going to be out of control if you keep hiring like you are here's what happens HR buries that as fast as you can possibly bury it and then the executives that gave the money or the resources or the time or the processes they wonder where that went HR doesn't want this to be shown to anybody and then it's just really bad because people that are even more suspect of predictive projects this is obviously absolutely something that you want to do about but you're you need to be ready to say I'm open to whatever the outcome is on this I have to have a lot of openness that if it does come back showing that our diversity and compliance is going to be out of control we're going to still be very transparent about the results and we're going to do something about it so I cannot tell you the amount of people that have this as their first predictive project that they do in HR and then they bury the results when it comes out so I would stay away from this one as being your first project unless you're an extremely open organization that's going to take that action and be very transparent the results just a couple of things the predictive analytics is not it's not a PowerPoint presentation I've already talked about this you know about the model because who cares you can talk about it even if it has close to 100 percent success unless you can implement it no one cares it's not general insider guidance it's not like what we see on the right hand side where there's just all these dashboards and charts and graphs and things you can eyeball the raziel requires somebody to look at all of this and say well you know we've got these key performance indicators on q1 and then we've got purchases by region and then we pull this all together and therefore that means no that is not predictive analytics this is an example actually from our system called talent analytics advisor it's predictive platform and it's a talent acquisition example and what it shows on the left is just a probability of somebody making their sales quota pre-hire so for somebody let's say that's in the talent acquisition role if they would just very quickly see that this person nathan has a 70 72 percent probability of being a successful sales rep and making their sales quota versus gene has a 27 percent probability of being successful in making their sales quota it's very easy to fit to the current workflow the data is added to other candidate data that they know about gene and Nathan and they just move on they don't have to look at all what we saw on the prior screen screen it's a very specific directive we also get questions about you know will our organizational behavior change you know our culture isn't ready for this I don't know this is really you know this is something big and scary and I just don't think we can move from this to that so how do you know if your organizational behavior can change or what will it take we definitely see definitely see organizational behavior change but a couple of things have to happen one is people have to be ready to move from ok you know when there's a challenge that they need to solve currently people start from ok let's get into a meeting and talk about what do we think okay everybody go around what do you think and you need to move from that concept to saying well let's start out what data do we have and what is the data tell us and a great example I think is with competencies people spend a ton of time and a ton of money either internally or with consultants talking about you know for this particular role what are the competencies that need to you know that relate to somebody that's a top performer in the role but nobody ever proves that it's a well-meaning committee that has a lot of opinions about it but nobody comes back and says is that true can we actually see it or people actually borrow those competencies or buy those competencies libraries from somebody else but I guarantee you that consultant has never come back and correlated that and said you can actually predict based on having these competencies that you're going to have a better outcome so instead of the committee deciding what competencies lead to top performance imagine if you could you know use analytics to say as an input to say when these competencies exist we can actually use those to predict top performance it's a completely different way of starting from what does the data tell us instead of what do we think so if people are able to say just chill out a little bit we'll come back and give our advice after we see the data then yes you can begin to move so it's really that openness to challenging the data challenging you know how you do things already today and also an openness to data challenging some widely held beliefs and also once you begin to move to actual behavioral change and predicting business performance then it starts HR starting to be connected a little more to the business and you can move beyond dr. predictive indicators like engagement or what we call middle measures or middle measures like performance reviews to actual business performance because performance reviews are not business performance business performance is what they actually do in their job what they were hired to do there's a couple of concerns and cautions and actually when we talk to somebody new the first thing they tend to do is to throw up a lot of the concerns and cautions and I think again sometimes you know there's a lot of fear but I don't let this be paralyzing for you and don't just throw this up so fast that people just stop and don't even explore you know is there a way that you can logically think through some of these concerns and cautions because they're you know you can you can allow it to you know suicide line you or you can say no let's just work through it logically because there are ways to work through these logically one thing to be concerned about is that the inputs are too close to ethical boundaries and this is really interesting to me because HR is the one that has a lot of the ethical concerns and yet one of the first thing everybody wants to see in HR is zip codes like how far from work is you know is that a in put that actually predicts somebody's performance or how long they lasted a role and that sort of thing that could be an input that comes too close to ethical boundaries because it actually could be a proxy piece of data for somebody's socioeconomic status you know different zip codes or they could be a proxy for you know different kinds of you know origin country of origin things like that and so you know rather than just having HR specify or a business leaders specify I think it's really important to work with your data scientists to closely think about the inputs that you're going to be using to make sure that you know you don't cross those ethical boundaries and again privacy is there's not one rule about privacy your company really needs to decide where you draw the line somebody or months' some companies are much more open with let's say looking at the email streams or you know people's calendars and kind of taking that into as part of their inputs as well other people say no absolutely for our company that's gone a little bit too far again that can be decided just like everything else could be decided there's also concerned that some people have especially in HR and across the organization are people too unique to be predicted and I understand that concern because I don't like to feel like I fit in a boxer that can people can you know predict everything about me and we can't there's no way we can predict everything about me however we are much more predictable than we think we are but this is something that you know people are going to understand and the more you begin to use predictive models you see how they actually do perform and they can be a factor for good and not just this big factor for for evil one thing to also think about when you start your predictive project is again I think people want to tiptoe into predictive analytics or you know and I think a lot of people just like to start project small and I guess I understand that for sure but when you're looking at analytics what you're looking for our patterns and patterns require a certain you know size of data and so if you go to small and say oh let's just look at this population of you know thirty five people to begin with that's too small and you don't want to you know not be rigorous you know there's trade-offs if you want to be very rigorous in what you find or do you just want to be super super tiny and again just let fear sort of drive you know how much of a sample that you get and again I think your workforce analyst can help you with that or your data scientist and the last concern that I have here is that we see a lot of management butterflies who say and data politics that we see where people believe that giving up their data you know to sort of combine with other people's data might also get into giving up some of their power and we always recommend that people focus on you know what is the goal that we have here but you know what is the outcome that we're all trying to solve here and let's just give up our data and move beyond some of the data politics and some of the management butterflies that are there there's also people need to realize concern about not using predictive analytics and these are happening today in your organization what's happening are every you know many many many undocumented decisions we call these mystery factors because when we're making decisions based on intuition again predicting who to hire predicting you know who to promote predicting you know on our nine box who's going to be a good leader predicting what kinds of what kinds of skills people need to have if etc etc a lot of these comes down to undocumented decisions bias and a lot of mystery factors and there's this continued ongoing significant bias that is built into a lot of these decisions based on nothing more than you know opinions really and it isn't it doesn't have this feedback loop about what is really working so that we can update the way that we make our decisions and I guess the final concern about not using predictive analytics is that the less favorable business results are going to occur because you're not using predictive versus you know other companies that are that are and the concern that your company might fall behind their competitors just a couple more slides here first of all just you know there's a lot of positive and empowering way so you can start your predictive projects you don't need to jump to the big scary ones right away and the two that I would recommend that feel less scary that are less personal one is talent acquisition and one is workforce planning workforce planning because again it's at the aggregate level and you know it's really you know just feels less scary in general you're just making sort of policy decisions and process decisions and talent acquisition because these are people you don't know yet you don't have a personal relationship with the candidates and so you know this tends to be these tend to be two very good areas to focus on the other thing about you know really reminding people when you move in to do a predictive project is to remind people that data science documents and holds the model accountable it's not that it just is implemented into your your stream of decisions and then nobody comes back to it but that it is held accountable whether it's every month or every three months we really recommend you know typically at least every three months that it was that it's watched all the time but every three months there's a definite report that comes back and says these are the decisions that were made this is what the model said and this is what actually happened you know outside the lab doesn't need to be tweaked at all is anything different and that report given to the management team so that the model is hell accountable and also to remind people that it's only part of the decisions these aren't robots but it's part of the decision that are given to the business people they take it into consideration and then they add that to their other data when they make their decisions moving forward I think HR can really be instrumental with predictive analytics you know they can take a stand for privacy ethics compliance as well as project progress they're not mutually exclusive it really removes significant human bias because machines just don't care and predictive analytics and data science really documents the actions it puts a lot more accountability in with a model kind of what it's doing in the decisions that it's making and I think HR can help others in HR and others in different workforce areas to really right-size the concern and caution instead of running around being super scared and letting things paralyze you you can get over this if financial services people who you know have your all your financial information your credit card numbers all of that or even medical think about hospitals and all of the information they have by comparison the scary information that organizations have about you like your pay or maybe your gender or whatever there's not a lot of really scary information that an organization has about you predictive analytics really is a force for outcomes fairness and diversity one thing I love about predictive analytics is that it's finds value in in subtle factors that drive success that today we're missing one thing that makes me nuts is just knowing in an organization that these high-value talent acquisition people that do the you know the first maybe call of somebody kind of a screener or a pre screener a lot of that is over the phone and who knows what you subtle factors they're not picking up on even when they take a look at the resumes there's something subtle in there that you know just goes nope that's not you're not in or you're in or you're out or you're in who knows what that is and so there's lots of subtle factors that are getting by the screeners and pre screeners and though could those could be top performers and we'll never learn and will ever find out the other thing that predictive analytics does is really there's a lot of biased categories and hidden categories that we don't know but predictive analytics would find that and say hey there's these very cool things you guys don't even know actually helps you know would really help to make people more successful here so we can look for that inside of the organization it's really the mechanism that connects the employees to the business results because there's no other way to do it than to have that full you know that full model of making a decision and then seeing the results tweaked in the model come back and just constantly giving feedback if you're you know I this is a small amount of time to get into a whole conversation about workforce and HR analytics so I just wanted to bring up very briefly and mention that if you're interested in going a little deeper Dave mentioned at the beginning that I am a an instructor at UC Irvine and there is an online course that's beginning November 7th on predictive HR and workforce analytics that's all online is just a 6-week course and I just wanted to mention some of the other topics that we'll be exploring in more detail at that time so we'll be exploring the difference between HR predictions and workforce predictions and getting into kind of what are those differences there's the best data science methodology that we highly recommend for using in your predictive projects we get the question all the time what is the most valuable data to use for creating predictive models so we answer those questions and kind of dig into that I mentioned I think briefly something called the confusion matrix which is very important to know in predictions we get into that talk a little bit more about data politics and bust a lot of predictive analytics myths we get into and this is a lot to cover in six weeks but we do we talk about calculating employee cost employee breakeven and employee lifetime value which is getting a lot of press recently as is something called survival analytics and we've been waving the flag for survival analytics for quite a while now because it really is the most valuable way to understand and predict turnover we talked about five predictive tools for predicting turnover and how to use cost information during the modeling process and then walk through to talent acquisition case studies I have the URL here for that course and again it begins November 7 so it's just coming up in a couple of weeks and just a little bit about our firm we as you can probably tell predict employee turnover and performance and we talk about performance it could be you know its business performance your probability of making your sales probability of you know having accidents if you're a truck driver probability of anything that you can measure we do that pre and post higher for something we call higher volume roles so that would be a role that has at least let's say 200 people in the role and then we deploy our models into advisor it's extremely easy to use you don't have to have stats map anything the cloud platform and machine learning keeps our models fresh so with that that was the end of my formal presentation but I would love to hear from you and take any questions and thank you so much that was excellent there are several questions and please guys bring any of those questions in in the Q&A area but a couple of them I just want to tell everybody I know a lot of you guys came in a bit late so the webinar is recorded and you will get a URL back to that if you have any trouble you can come onto our website and a bunch of our free events are sort of there to help the entire workforce get jobs and get better jobs and do a better job of their existing job we have a bunch of free webinars like this one out on our website and they are listed under free events and this one will be up there tomorrow you can share it with other people if needed the other question was how much math and statistics and computer science do I need in order to use any of this stuff Greta and I know you mentioned a little bit about that but I've had these kind of questions before what's the typical HR person's background with and computer science and how much do they really need so yeah it's a great question and I would break it into two questions so one question that I will you know the first question I would have is how you know what kind of background math statistics etc does a person need to create a model and then the second one is you know what kind of background do you need similarly to use a model so to create a model a predictive model I mean you can do some you know if you know Excel and you want to do some you know some regression analysis or whatever on turnover you know as long you know maybe you've got an MBA and you you know have a little bit of background in that you can do some simple work and and modeling but it actually is a very specific it's a very specific science that has a combination of statistics and it even goes beyond analytics so really understanding tools like R and Python and other modeling software or SPSS that you know allows you to really you know look at the inputs and look for the correlations or the connections to the outcomes that you're looking for so the kind of background that I would say would be somebody that has actually some computer you know programming experience somebody that has statistics and somebody that has math is probably a good background it's a really interesting combination the other thing that I wanted to bring up is that if you're inside of an organization and you're like wow you know our workforce analytics person doesn't have that or a lot of times what we see is maybe the person that's been doing reporting or operations for HR is now being pushed into trying to get into this role I think it's a very different person typically that does reporting and dashboards that are done the person that actually does core predictive modeling it's very different but you might be able to look for some help with your marketing analytics people or if you have a data science team elsewhere so you can really work with them and you know maybe break different pieces out and subsequently actually also there's also the person that maybe takes the results and then also does data storytelling so the other thing is to think about and we actually did some work on this what really is the data scientist and so we see different clusters of you know people that actually define the outcomes people that get the data people that actually work with the data people that do the modeling people that do the visualization and the dortmund data storytelling so kind of really depends on where you are you know what pieces of that process you're doing as well however it also for that for the class is there a specific requirement you have to be you know real deep or can you come in just being in either one of those categories for the class yes this is an introduction course so you can be in any of those categories yes this is a we really do a lot of introduction to the concepts and we wanted it to be slightly lightweight to a lot of the I mean you're going to get a lot from the class but you don't need to be a programmer to be in this class or a statistician to take the class great great and one of the other questions and again please guys we'll have a couple more minutes left if there are any other questions you want to answer do you have a expert on the line here so please take advantage of her and again if you took the class you'd also have a whole six weeks with her get a lot of free consulting on the side because there's a lot of great interaction happens in the class and it really is quite valuable as you move along in the career just not not to mention the content itself but the other question that Greta was you know budget wise this person has a hundred and fifty 200 employees it's relatively small company is it something that a smaller company can get into and if so what's it going to cost them at yeah it's a great question so predictive analytics takes data and so you know there's definitely things that smaller organizations an organization of to is you know that's going to be a while that you're building your data but you know an organization of 150 people there are definitely things that you can do like talking to your others to know and really identifying you know one you know great example is Talent assessments I have a real issue with people that offer Talent assessments we offer them as well but that don't give companies their raw data back to them to try to use for correlations for their predictive modeling and so one thing 150 per person firm could begin to do right away is to make sure that with all their vendors their software vendors that are capturing any kind of employee data that they can access that raw data because as employees come and go you're going to have some turnover in you know one hundred and fifty person firm and you may not be any be able to do anything with that data today but maybe a year and a half from now if you continue to gather that data all of a sudden you've got all this great historical data that you can at that point start to do some predictive modeling but if your vendors aren't providing you with that that's a real tragedy so that would be one thing I would say for everybody on the call regardless of size you know your if your software is working great that's one thing but the data that comes out of your software is almost more important so there's definitely things that you can do and for smaller firms sometimes you need to go with more of an industry benchmark so you could start there until you have like your own going for you but there's always something you can do including getting your data that's your IP you should get it next one all right well we are just about out of time so I'm going to end it there this is one of the questions that we should know I guess Burnett has one other question Ram it I couldn't find you there the question was I'd like to get into this industry and I'm not currently there and there's a lot of answers that we typically get because we give these kinds of questions a lot and I'll let Greta chime in as well but one of the biggest things you can do whenever you're trying to transition careers is you know improve your odds by sharpening your skills a course or two like these but also again the biggest thing networking now within a course you can also network of course because there's going to be 20 or 25 people including the instructor many of whom are well integrated into the place that you want to go and that's a really good thing because you get to know these people even though it's online and the other thing is industry events and I want to plug greta's event one more time since she's on the show tonight but there is a great event called predictive analytics world the ones coming up in an early part of next year I believe Greta is pod workforce in April or March it's in it's in May actually yeah predictive analytics world for workforce it's been it's like the May 14th through the 18th and San Francisco they will just gate put some of these things on that calendar guys because when you get to these events live and you can get to some of the local events there's a digital analytics Association you know dig into some of these events yourself search up some of these things try to get to events live and and meet people and talk to them it gives you great guidance on career paths but also helps you hook up with people that could at some point in the future help you get a job Nathan I want to thank you for bringing that up and one thing I would say is a lot of HR people only go to HR events like they'll go to a Workforce Analytics event or an HR event where they kind of touch on analytics my recommendation is that data science or predictive model err is a predictive model or as a predictive model or as a data science there's a data scientist regardless of the data set that you're working on and so if I were you it's why you'll see our name associated with predictive analytics world because the predictive people are there that notice stuff really really well and where we're on the you know a part of the predictive certificate at UCI because the predictive you know experts are there and so don't hold yourself to just going to HR events where they're just learning typically at HR events they're just learning and I don't have any problem with HR events but the experts are going to be at data science events and so go to meetups where there are other predictive people there go to data science events as well as HR events if you want to learn more about HR but to learn more about predictive go to predictive events x1 excellent and George was asking another question about if Greta could share more details about the one success story she shared about a company that predicted someone's flight risk and then took action and before she gets on to that I know some of you have to drop off if you do again the rest of the webinars being recorded so if you want to listen to the last little bit here you will you will see that so we're going to go ahead and go for another three or four minutes here let Greta answer that question and I'm also going to come back to the other question about how with that November some of the class how you get access to that but I'll let Greta answer the other question first and yep absolutely let me just see that I had the URL here I thought okay the URL is down here at the bottom if you want to make note of that while we're I'm telling you about so the one example was sprint and it was their call centers and they were looking to so I don't know what software they used or process to predict the probability of existing of the flight risk for their existing employees but they came back essentially with the list that had you know everybody's probability and they very systematically typically what happens is that people will share that maybe with the managers and say hey of your direct reports you know Greta is a high flight risk and the problem with that is that you introduce bias or you have potential to and I've typically seen that that happens because now if I as a manager have a really great assignment but I know that you know Greta has a high probability of you know leaving now I start to not give her assignments and you kind of have a self-fulfilling prophecy or I think she's going to be here for a long while I might start giving things to her so it's print what they did was they actually immediately went into I think even before they got you know the list of you know probabilities people flight risk they hit put into place if this is what their scores are then these are the things that we do you know and so they actually had remediation after that that involved the managers and involves the people you know they had all kinds of things ready to go for the managers to contact the people you know and if they were top performers because you don't just want to contact everybody just because they're you know high probability of them you know leaving if they're really a bottom performer so you have to work all of that stuff out beforehand and so they had it all worked out they had their plan and they were able to I forget what the term is but you know essentially you know rescue or keep people on board and really change flight risk or change turnover based on the flight risk initiative that they did I was very impressed they actually presented earlier this year at predictive analytics world for workforce they did a really really great job and it took me a long time to find somebody I'm the program chair there and it took me a long time to find somebody that could present on flight risk that had done a great job on it so that was the back side better thanks so much I'm going to switch over real quick and just show you guys switch over to me I take the ball back change roles to presenter and then share yes all right guys this one quick second here and we're just going to say happy birthday to oh my oh well I had to do that thank you thank you mute everybody but some of these guys really can't sing very well so that's what that word is good and there's no calories so this is awesome now what I'm going to do real quick here guys is my little desktop I'm just going to assert so you want to do the same thing if you want to find that class UCI UC irvine and if you want you can type predict analytics in there and our certificates really popular so it's going to come up pretty high on the list if you just click on that predictive analytics certificate program at UCI and tell you a little bit about it you click this little tab on the right hand side that says course schedule and there's the core courses but Greta's class is an elective in the program that's right down here predictive HR and workforce analytics it's brand new I've never been taught click on that little green button and you can click on this little green button and you can enroll so if you are thinking about doing that do it rather quickly the class is coming up but it is filling up so you want to do that pretty darn quick and with that I'd see I don't see anything else's major and we're just about at the time thing if there are any questions about this at all you can go to our website there's contact information for myself and other people that can help you with courses here you've got greatest information on the last slide again everything was captured and will be sent to you so you don't have to worry about if you missed writing something down that will all be linked back to you please let us know how we can help you can tell both Greta and I are pretty we're very excited about this field it is a very strong and growing field it is something that can help companies do better and compete better not just in a charmin at other places as well a very exciting place with a huge amount of job opportunities out there go look on dice or monster or something and type in data science or predictive or HR analytics and you'll just see tons and tons of hits so you can tell we're excited it's a fun place to be it's very very interesting if you have any interest at all please pursue it it is very cool and with that I just want to say Thank You Greta so much that was absolutely wonderful perfect thank you David little presentation and in a great class and thank you to all of you for taking the time I know a lot of you this was probably West Coast guys you were on your lunch so we really appreciate that probably hungry so now you can go eat and with that just thanks again everybody thanks Greta thank you so much see you in class you guys

Be First to Comment

Leave a Reply

Your email address will not be published. Required fields are marked *