In this video, HireVue Professional Services Consultant Kristen Morgan speaks about how recruiters can join the big data revolution to drive more informed business decisions.
Watch this on-demand webinar now to learn:
Scot: Join me in welcoming Kristen Morgan, professional services consultant at HireVue. Don’t forget to join the conversation at #talentinsights.
Kristen: So thank you so much, Scot. And I’m so excited to be here today talking to everybody about big data. And I feel like the title “Big Data: Go Big or Go Dark” is kind of ominous, but we all know unless you’ve been living in a cave for the last few years, big data is everywhere. It seems like you can’t open your audible feed or look at the scan or any of the three different places that you go to consume news on a daily basis without seeing something about big data. Data analytics using data to make decisions and how to differentiate it. So, HR fortunately or unfortunately has been a little on the back side of this big data revolution because what we’re dealing with people and the uniqueness of people. It’s sometimes hard to push it into the big data world and make it so structured and consumable. But I want to spend a little time today talking to you about the work that we are seeing around big data, what is happening in this world and some of your opportunities to leverage big data within the HR space to really help drive business decisions and make an impact within your organization.
So like Scott said, I’m from the professional services team at HireVue, and I actually lead all of the implementation of our HireVue Insights platform which is our proprietary big data learning analytics engine. And so I’ve worked with a lot of clients and heard a lot of them say just a various challenges they’re facing and if they could predict who their best employees were going to be or who their rock stars we’re going to be or just who could pass the series seven, what ways that it would change their organization and the business that they’re trying to conduct. And so I’m really excited to talk to you about it today and just provide a little perspective.
Now before we get started, I want to really level set what does big data mean when I say it. You may know this. If so, feel free to tune up for a few seconds but a lot of people heard about the words but may not really does delve into what does big data mean when we start talking about it. So it literally is just extremely large sets of data, sets of information so large that one computer can’t process in a timely manner. And our internal data scientist at HireVue kind of blew my mind the day that we sat down and talked about this for the first time because he said the meaning of big data continue to change on a daily basis. Because think back ten years ago, what was considered a large set of data, too large for one machine to process at one time, would now be . . . your phone could do that in a split second.
So it’s going to change and what’s considered large will continue to shrink as technology progresses, but for the conversation today, it’s basically taking data points, so many thousands and millions of data points, that are so significantly large and predicting trends from that information and using a variety of assessment, different machines and multiple machine to get answers. I go to the non-science speak version of this when it really means take all the vast information that we have available with us and doing something useful with it.
So when I say big data, that’s what I’m talking about. And it’s not that wild of a concept. Before I was with HireVue and learning about the topics, I didn’t really realized that big data and predicting outcomes from all the available data really are in our world all day, every day around us. So every time you go into Google and you start typing a search and it starts to predict what you’re going to search for, that’s big data and predictive analytics. They’re predicting your search results based on what they’ve seen in the past from their vast sets of data. Every time you go to apply for a loan or maybe you take your kids to get a car and get your credit pulled, another version of big data and the prediction power, essentially folks are looking at past information associated with what you’ve done credit-wise and the time to pay and your ratios, and they’re predicting how likely you’ll be able to pay that back.
Amazon, Pandora, Netflix, what we would consider recommendation engines that take your feedback and your buying habit and things that you purchase in the past maybe with Pandora. Things that you’ve done thumbs up or thumbs down on and adjusting their recommendations based on the data you provided to them in the past. Though all of it is real life versions of big data and data analytics, so it’s not . . . As we talked to the HR community about HireVue Insights and the work that we’re doing around this sometimes it seems like, “Oh, you guys are a little bit magicians. Maybe you’re just . . . and you’re the first ones to have come up with it.” Nope, it’s just masked.
It’s being done everywhere. We’re just not seeing it done in that human context that much yet because it’s really hard to take really unstructured people data and put it into something meaningful that a machine can then consume. So, I want you to go with me on a story, if you will, for a moment. Go on a journey with me. So my husband and I just watched Interstellar last weekend with Matthew McConaughey and if you haven’t seen the movie basically earth is dying and they have to go find a new planet. And they’re evaluating all these different planets that 13 scientists have gone out to see way outside of our universe and one of the planets that were on the sense about that may work but it may not. It’s right next to a black hole. If you’re a nerd like me and into science fiction at all, black holes can suck everything into them, send them to the event horizon never to be seen again potentially.
And so this one planet that they’re evaluating is right next to a black hole, and there’s this whole conversation about is it viable? And can we go there? And it made me think as I was watching, HR kind of sits on the edge of a black hole sometimes. We have all this data available to us. We got the resumes that we get from candidates. We’ve got the interviews that are conducted with them. We’ve got the pay that they’re offered. We’ve got their performance data. We’ve got their attrition, whether or not they stay. We’ve got their exit surveys and they don’t like it. But there’s not really any methodology that take it all and make sense of it.
It seems like it kind of just goes into the black hole, never to be seen again. Funny part, spoiler alert, they end up going to the planet with the black hole at the very end because that’s actually the one most viable. But side bar. So that visual struck me while I was watching the movie, and I thought this is the problem. There just hasn’t been a good methodology for have to take all that information out that’s out there, all that big data, extremely large data set and make it mean something.
And this is kind of the same thing that we’ve been talking about at HireVue for years. So for those of you who aren’t familiar, HireVue started out as a digital interview platform. So candidates could come in and instead of you doing a phone screen, you can have them record a video interview so you could watch them give the answers to your interview versus having to talk to them on the phone or have them face to face. And so our team sort of conceptualizing with our data scientists, what can we do with all these data that we have? We have thousands and thousands, hundreds of thousands of megabytes of data from videos and what people are saying and how they’re saying it. And so we started to think through some of the major issues that we face within our industry and there’s capacity issues. So you’re opening up a requisition for a new employee and maybe you have this problem but maybe you have a volume problem where you have 50 applicants or you have thousands of applicants and your team only have so much time to fill the role. They’ve only got so much capacity to handle it and so maybe you don’t get everybody.
And how are you sure that the person you placed, who may have been the tenth interview isn’t . . . that you’re not missing out on somebody, who maybe was 27th or 230th. How can we take this data and address the capacity issue? The same line of thinking, how could you take all this data that we have and address consistency issues. Over and over again my clients tell me they are concerned about what their managers are doing in interviews, in coaching, interactions, is the consistency there? Are they all identifying talent in the same way? So again what if we’re able to take all the data that we have in the evaluations that we’ve gathered and use it to help predict which interviewers are the best of taking talent. And then maybe use those interviewer evaluation skills and overlay it against candidates so that you could match up. You could basically have your best interviewer in every interview.
Lastly, what is efficiency issues? What if you’re losing candidates just because of maybe their resume isn’t that great just yet. They’re recently out of college and your applicant tracking system which is awesome and goes through and searches for particular keywords, you might not have them. Because though they’ve got the passion and they’ve got the charisma and all those things that you want from somebody, they’re that candidate who just is dying to work for your company, they’re don’t look as good on paper as they should. And so they go into the resume black hole, if you will, where they’re never actually even given the opportunity and you don’t get to see them. So some of the things that we were thinking about as an organization.
And so HireVue decided to bring on a full time data scientist because why not jump into the deep end of the pool. But the thing about big data and about how you’re going to take all these data is it has to be structured. So in the world of big data you’ve got unstructured data, things that are not numbers, ones and zeros, things that are not machine ready, and then you’ve got structured data. And baseball is probably the best example of structured data. All of you that saw money ball I’m sure conceptually get this. But in money ball, they were able to predict who is going to be a good fit for their team based on the statistics gathered over the course of their baseball career.
And so that is when you think about big data and you think about what to do with all of that information that you have, we have to really consider it has to be structured. It has to be consumable by a machine so that it can take it all, kind of do its thing and put it. So these are some of the challenges that have been associated in a human resources and big data in particular.
And you’re maybe at the point in this presentation where you have some burning questions. So I want to go ahead and say I do have a Twitter handle, I would love if you would tweet at me, it’s morgan_kristenl. Make sure to use the #talentinsight and send me your questions. I’ll be responding that way. I really want this to be an open conversation because the first time I heard this information, it kind of blew my mind. And then I started to get really excited about what could happen. So if you think you want to embark on a big data journey, tweet me and use the hashtag, and we’ll find answers to your questions. Even if they’re not, you should come and work with HireVue. So just a pause and throw it out there.
So, I feel like we’ve laid out a pretty good foundation for what is big data, a vast sets of data used to predict information. We know that it needs to be structured versus unstructured and we’ll talk a little bit more about that in a minute. One of the first things that started to come about during what we’ll call the big data HR revolution is HR resume modeling. Because somebody said, “Well, resume have all these information. What if we could predict performance from resumes?” They’ve got work history, they’ve got education, they’ve got skills and competencies, they’ve got the candidate’s accomplishments. Should we in theory be able to consume all of that and make a prediction of who our best employees are going to be. And so the way that it works essentially is you take what we consider all that unstructured data. The candidate’s name and their GPA, what they did before, what school they went to, what’s their burning passions are, and you structure it. So that just means make it machine consumable.
You’re laying out, are they male or female. Their GPA is already a member. Yehey, their job can be assigned a number. You basically want to get everything done into ones and zeros. And when it’s been transitioned over from all of the various points in a resume into numbers, we call that munch cleaned data. And the thing about resume modeling, it was an awesome idea. And we see a lot of clients who that is how they’re entering into the world of big data since I’m looking at resume modeling.
And you don’t have to necessarily hire a data scientist to start on this practice. There’s a resume parsing services, where you can send your resumes to, and they’ll take the data and move it from unstructured into structured. And then you can also utilize generalized prediction services like Amazon and Google and DataRobot. So that when you have a structured data, you can feed it to them, and they will help you predict. But one of the things that started to come out with resume modeling is twofold. First of all, all of you that have had a resume professionally written, raise your hands. And I think there’s probably like a thousand hands raised across the country, if not more, because we all have. Or maybe you get to the point in your career where maybe you just need a little help. And so sometimes, it’s easier for folks to give, to deceive a resume parsing, a resume modeling service because it’s not actually them that’s being represented. They got help.
On the flip side, there are those before you could afford to get somebody to write your resume for you, you were just out of college. And maybe your resume were terrible and sparse. But you actually had a lot to offer, a lot of value, a lot of passion. And so on the flip side you guess at a candidate who don’t get that opportunity because they don’t have a sophisticated enough resume. And so while resume modeling and resume prediction of success is good, it’s a good first step, our team at HireVue thought, “You know, is there more? Is there more that we could do?” So as an organization we started to think about everything that we’re gathering, the millions of interaction points beyond just what somebody put on a piece of paper. And how do we take that and feed it through a system that could help us predict performance for our clients? And so when you look at this screen, I consider above the water the things that are publicly available about your candidates or about people that currently work for you, their history, their education. You can gather their references. You can gather their skills in some capacity, assessment, things of that nature. List below the water stuff the things that really make them them, their passion, their emotion, the language that they use, the way they deliver it, their expression, their . . . All of that is a significantly more robust data set. And so we wanted to do big data malling off of that but how the heck would you put all that into the funnel to get to your rock star.
And so a quick and dirty version of big data modeling where we undertook building predictive models using all of that interview interaction data. So we took the raw data that the actual on demand interviews and we’re able to convert that video and audio and the lighting and the cadence and the structure into structured data. This happens with every single big data models that is built. The next thing that happens is you have to pull out what’s important, what do you want the model to predict were called features. So you want to really say out of our gigantic vast data pool, what are we trying to predict and what are those features that are actually aligned with predicting that.
So in the world of HireVue, you may have been a quiet [inaudible 00:19:34], go ahead and take a look at all the interviews of candidates who have gone off to become top performers. Go ahead and take a look at the interviews of candidates who have been with us for five years. They even write a retention, attrition data, performance data. But whatever it was, we’re able to pull out the differentiators between those who our clients would like to predict and everybody else. So once you do this what we call features selection and this happens across the board every time that a pretty good model is built. You then take all the data and you are going to train the model that you’re building on the data. And so basically it kind of like if you are going to run a marathon.
You put together a training. You have a training regimen. And so we’re feeding all the data through the model, through the algorithms, things that can buy our data scientists to train it on what it’s looking for. But the really important thing is you want to hold out a subset of that data, what we call a validation set. Because you want to train the system and then you want to confirm with completely unrelated data that what you thought it was going to tell you, is in fact is telling you. So you’ll hear data scientists talk training set, validation set.
Once the data has been pushed, you pick everything that you want to predict, you train them all on how to find it, you validate that you have it, then you can deploy it. And so this same process will occur whether you wanted to predict using your attrition data, whether you wanted to predict using your performance management formats, whatever it was that made sense to you in terms of your data set. This is what we need to happen for you to have a model that would work. So in the world of HireVue, we did all these, we took all these information and we were able to put together what we call HireVue IRIS, which is just the name for our deep learning analytics engine. And so taking into account all of those data, pulling out the data points, pulling out the features that matter, and then predicting results. It became this output of big data. So clients, in our world at least, can compare incoming candidates to the model of those who have done exceptionally well or gone on to be top performers.
They actually got a score which is very similar when you think back to a credit score predicting how likely someone is to pay back their loans or not. An insight score in our world for our candidate is predicting how likely they are to become a top performer. And it’s removing some of that inconsistency or efficiency losses where . . . so every candidate can be reviewed by the model. Every manager can be assessed how accurate they are at identifying talent and how aligned they are as a team. So it’s giving you that really powerful data set that an HR organization needs to justify their value, justify the expense for the thing and really sort of take the next step in moving from mostly art because a lot of HR is art, it really is, and it’s nothing that can be completely outsourced to a machine. We see that the best relationship are those where the people and the art getting matched with the science.
And so clients are then able to take that information and use it to make decisions. So I wanted to get set back and just kind of let that soak in and consider this. You as an organization say you want to pursue some big data activities. You’ve got some steps you can take. And as you look at this slide on the screen, you can see obviously things go from good, better, best, cheaper to most expensive. But these are also stacked based on risks. And so the items at the top of the triangle are the riskiest. Items at the bottom are lower risk.
We’ll spend a few minutes just talking through organizationally you decide that you want to pursue big data, you wanted to do some objective decision making. You want to predict using all these vast information that you gathered. You can go about it a few ways. First, most expensive, most aggressive, riskiest, but potentially the best, is you can hire your own HR data scientist. This field is exploding. There is no shortage of HR data scientists available to you. So if you’re the type of organization that says, “We have a strong vision, we know what we want. We’re going to bring somebody in-house full time,” then you can do that. And that’s again what we do at HireVue.
The thing about bringing in your HR data scientist is you really need to know what you’re trying to accomplish. They are scientists, and I am not one. Either we have them on staff, I worked with them. I used to grade with them at times. They are awesome and smart but you have to really get that balance of can they understand the vision of your business while also understanding the science. Do they have the coding skills? Do they have a deep understanding of statistics? Can they take the data you have available and transfer it from being something really unstructured into the structure you need it to be. So that is kind of deep end of the pool. You’re going to go in over your head since . . . Get you an HR data scientist. The next two levels down which are data science consultants and HR analyst are what we see as probably the most common right now.
So if you step down from having a full time HR data scientist on your team, you can have a consultant come in and that consultant can really help you to show the value of this practice. They could help you to chew up the problems you’re facing as an organization. They can identify the issues. They can help basically to get the momentum going around this process and help you get the justification that you need and the dollars basically identified to undertake a big data predictive analytics modeling type of culture within your organization.
So HR consultants, HR data science consultants are really available. They are less expensive than hiring one yourself, but you’re going to get a little bit less as well because they’re not in-house. The next step down would be an HR analyst. So a consultant is fully embedded in statistics. They can code, they can take your unstructured data, they can make it structured, they can build the model, they can do the predictions, pretty much every single thing [SP] in-house, the HR data scientist can do but per hour.
An HR analyst which again is a pretty common place that we see a lot of organization is starting. They had exposure to statistics. They have much more knowledge about it than your typical HR director but they may not have the coding or the programming knowledge that maybe a full pledge HR data scientist would have. But again this is a good place to start if you’re interested in structuring your data, getting all your ducks in a row if you will because you know that you want to embark on your own prediction, your own modeling and undertake it yourself.
And then at the bottom of the triangle, you can obviously leverage third party tools. I can spend a few minutes telling you about what HireVue is doing in terms of our HireVue Insights products, and so that is the prime example of a third party tool that is using data modeling and visualization. Another one might be Tableau. Tableau is a really good data visualization organization, and they can take all of your data and push it out in a consumable way and be a dashboard and make that information consumable.
So a lot of the third party tools again to kind of keep it in scale much like the day the DVD players came out. They were funky and expensive and over time, just like technology, a lot of this big data is going to get easier to consume, lower barrier to entry, just generally friendlier. But there are already a lot of great third party resume parsing tools, HR modeling tools available for you, and you’ll see, they are the least expensive and the least risky that gives out of the box solutions.
I’m going to go ahead and stop talking. I’m coming up on my time, and I just want to remind everybody, if you do have questions, if you do want to continue this discussion, please do tweet at me, @morgan_kristenl. You can send me an e-mail as well, it’s on the screen. Make sure to use #talentinsights and go ahead and send me your questions because I absolutely want to hear from you. I want to know what you’re doing in this space. I want to know the challenges that you’re facing and I would love to talk to you about HR data modeling and big data analytics. I can also call on our data scientist as needed to have the conversations when you get weighed over my head.
All in all, I just want to thank everybody for your time today. I hope that you learned a little bit more about big data, about some of the ways that is being used in our day-to-day life and about the ways that you can start to dip your toes in the pool or dive right in if you’re ready to get started with HR data analytics. Thank you so much. Have a great day.