share interactive transcript request transcript/captions live captions
|
MyPlaylist
SPEAKER: The following is a presentation of the ILR School at Cornell University. ILR, advancing the world of work.
[MUSIC PLAYING]
Good morning, good afternoon, and good evening. Welcome to our ongoing series of webcasts from CAHRS. Today, I'm going to talk about this idea of innovations in HR analytics, at least that's the title. Innovations may be too strong a word, so we've been out doing working groups on this topic for about six years. And myself and John Hausknecht have pulled together some ideas of what we think constitute what is trending to be best practices in the area.
The first one I want to start with is this notion of structure. And to me, structure is really important if we're actually going to deliver better results through our HR analytics teams. And so what we've seen emerge as the new direction, or maybe a best practice, is this notion of splitting up analytics into at least three broad categories. And so one is the group or team that's just doing data integrity. So that's the team that's collecting the data, is making sure that data is accurate and correct.
That, then, can feed into the second group. And the second group is the group that we think of as doing reporting. So they may be producing ongoing monthly reports, they may be creating live dashboards, they may be doing or helping to support either some form of customized reports for separate groups within the organization or customized reporting for the top of the house. But that group is just doing those reports. So what they're really responsible for is making sure they understand the clients and how they're using it and what data is important, and making sure they're pushing that out on a regular basis.
And then, the third group-- what we think of as HR analytics or the real magic of decision-driven HR as happening-- is that team that's really doing the exploration. And they're really looking at questions that can be answered through data. But the three really need to go together and work well coherently if we're going to do better data-driven decisions across the organization.
So even in the data reporting team, that data reporting is helping generalists or specialists on the ground see where there is changes in the business, see where there's changes in the HR metrics and understand if potentially there is a problem, to which they can then go to the HR analytics team and dive into what, I think, are really the next set of questions that we can really answer better through a strong HR analytics team.
And the questions, to me, start with why? So why is something happening? So we've seen a change in business performance, we've seen a change in an HR metric. It's really trying to do the root cause analysis and using data to understand what may be driving these changes, what may be driving a problem inside the organization.
Second question that we might then think about answering with data is the so what? So we've seen this change in an HR metric. Let's say it's engagement, or turnover, or time and position. And what we want to know is, well, does that matter? Before we try and enact some kind of change or do something different, what we want to know through data is, has this change impacted the business in some kind of negative or positive direction?
Third question we might then ask is how? So can we model through data or, with data alternative solutions to this, try and understand which one might be more efficacious or have a bigger impact on the organization? We might think of that in terms of structuring some case studies or structuring experiments where we try different solutions across different populations and see which works best.
A fourth potential one, which I didn't list on here, is also thinking about using data to answer HR planning questions. So what are the kind of capabilities that are most important to a particular business or working group or unit? Which of those competencies or capabilities are the biggest gaps, currently, toward the future? Try and align some HR planning questions around competencies and direction of the business.
I think what all four of those really have in common is that often, we need more data to answer those questions. The data within our existing HRIS systems may be a starting place, but it tends to come up short in terms of answering any of those four questions. So again, that goes back to the structure question of, do we have the right team in place that can think about these questions, think about how to get new data, think about how to aggregate and investigate that data?
And as long as I'm talking about questions, I do want to remind you that you can submit your questions. So I've got a few prepared thoughts I'm trying to go through here, but I really want to make this as interactive as possible. So feel free to submit those questions as they come. But again, when I think about the core of the HR analytics team being the data scientists, the team that's actually answering the harder questions for the business, it really falls-- for me-- around those three big ones.
And then, potentially, the fourth one, that the HR planning question is, can we really answer, as a group, why things are happening? What's the root cause? Can we answer the question of so what? Did this thing really matter for the business? And then, how might we do something differently? And when we think about those things, it really comes down, to me, to understanding, well, then, what are the key capabilities of the HR analytics portion of the team? The team that's doing the deeper analytics dive.
And so part of that is, I think on that team, we have to have someone who's got deep knowledge of the business, how roles and responsibilities and jobs play out in a particular business. So they really need to know how that business works. What do they produce? How do they produce it? How they impact the customer. How do they drive financial results for the organization?
Part of that team and the deep capabilities they need is really around research design and data modeling. Particularly, if they're going to answer these tougher questions, they have to understand the latest techniques of how do you take this data? How do you analyze it? What's the best way to try and get a causality? What's the best way to really try and understand what's happening in the best way to both design the study, and then, analyze the data?
A third big capability, or critical capability, is this notion of translation. So once you have their results, can you really tell the story of what the data is saying? Can you really get more people on board in terms of why this is important, why do we want to work on it, how big is this challenge? And really be able to draw people in. So just putting up a chart with numbers, often, isn't enough. We really want to think about how does this analytics team tell a story to draw the organization in?
The fourth one we really want to think about in terms of capabilities is what I call collaborative consulting. So it's not just, here's the data, good luck. But how do I, now, partner with a generalist on the ground? How do I partner with the business leaders in the unit? To really think through how to solve the problem that's in front of them. Not just giving them the here's what's wrong, good luck. Or here's what our data say, here's a nice file folder or a nice booklet full of charts, but how do I move them from identifying an underlying cause, showing them that it matters for the business, and then, helping them think through potential solutions.
And then the fifth one I have is, there's a world of data, there's a world of knowledge that exists out in the world of academia. So how do I have a team that can also look not just in our own data, but be able to compare that to what exists through hundreds, if not thousands, of studies that exist out there, in the academic literature. So can I compare what's happening here to somewhere else? Has there already been solutions or knowledge built up around this topic? And how do I leverage that to maybe study it more intimately or more deeply in the context of my own company?
And I think the problem that exists here is that for most organizations, what we tend to be looking for is what I would call the unicorn in the forest. Every organization wants to hire their analytics person or lead person and expect them to have this wide ranging skills. To me, these five things are really hard to find in a single person-- someone that has that deep knowledge of the literature, that's also good at being able to do the collaborative consulting, that still knows the latest and greatest methodologies for research design and for data modeling, that also deeply knows the business.
And so to me, I think what we have to start doing is thinking about, again, structure-- what's the structure of our HR analytics team-- so that we've got some combination of those skills. And that might mean that we bring in some HR business partners and generalists that have the deep business knowledge and the deep knowledge of how the organization works and the key roles. It might mean we have a couple, or one or two, data scientists that really get the data modeling and the analysis part. It might mean that we have another person, maybe that has consulting background, that knows how to take this and move towards convincing the client that something needs to be done, presenting the data in a way that draws them in, to help think through the next steps in designing solutions.
Or at the very least, we have to create an analytics team that knows how to partner on either side of the analysis with the generalists out in the field, to have those generalists who are asking really good questions. And then, helping them to think through what are potential ways that we can think through how to roll out solutions and test out which one is better. So again, to me, this capability thing is really difficult to find in a single person. So we have to think about the structure of our team and balancing those capabilities, whether it's all within the team or are some kind of partnership, either internally or externally, to get this work done.
The last real big issue that I wanted to put out on the table before we take some questions is this notion of the underlying challenges of the function. And so what we've heard for years now-- and more importantly now than ever-- is there are some big issues that are impacting our ability to drive better decisions through data. And the first one is probably not a surprise to many of you, is that we've got some problems inside the organization, typically with quality of the existing data.
A lot of us have heard that, we're not even sure we have the right number of employees counted, we don't have all the right information on those employees. Things change quickly in the business, and is that data up to speed? And more often than not, do we even have the data to answer the question that's in front of us? And so I think we have to get better at thinking about how do we get new data? Whether that's specific to a particular challenge, specific to a particular business. I think a lot of what that HR analytics team needs to start thinking about is what new data can we measure to answer specific questions?
And then, how do we get access to it? How do we get access to business data, to customer data, to operational performance data? And to combine that with the people data that we either have or are going to collect to make sure we've got a richer understanding of what's going on inside the organization. So a big challenge we have is access to more and more data, rather than being content with what we have in our current HRIS system.
Second problem I see a lot is that a lot of the studies we've seen presented to us at meetings and at working groups have flaws to it. That we've got an assumption that, well, we've measured this, and it's not really what we wanted to get out, but it's kind of close, so we'll use it. As an example, I've seen a lot of companies talking about things like, well, we're going to use years of experience to really model the extent of knowledge in the job. And we know that's a really poor predictor of how much that person actually understands and knows and has all the key capabilities for a particular job.
So we've got to be really careful in how we're thinking about this notion of data and making sure we're designing our studies more effectively. And so that's a great way to think about either bringing in experts inside the company or even renting that expertise. So again, I mention this a lot, but there's hundreds of great universities around the world with hundreds of really well trained academics who would love to partner with you and think through, how do you effectively design a study using the latest and greatest techniques? Again, they can help you analyze that data. So there's ways to think about bringing in some expertise around these questions.
Third big underlying question or challenge I see is that a lot of companies are getting better and better at the analytics part, and then just dumping that across the transom. And what I mean by that is, they've got a smart analytics team who's gone out and they've researched something, they've come up with some good answers. And then, they pa-- [CLEARING THROAT] excuse me-- pass that off either in a binder, in a PowerPoint deck, or in some kind of presentation, for someone else to roll with it.
And so I think that's a big missed opportunity, that HR analytics team doesn't get more involved in measuring the effectiveness of the action that's taken, help think through alternative ways to roll out a solution so that we can test how that might work better or worse under different conditions. So I think we have to get better and more involved in the actual design execution of the solution to measure how well and how effective it's being.
And then, I think the fourth challenge is the better the analytics team gets, the stronger the reputation is they build inside the organization. It really becomes a funnel issue. You're getting dozens, if not hundreds, of questions in a year, and you don't have the capacity to take all those on. So it really starts to become the question of how do you screen? How do you prioritize? How do you funnel them down into the few problems that you can really test out in a serious way in a given year? And then, look to then either build greater capacity or help the generalists out in the field to answer some of those questions on their own.
So I think we've got a lot of questions to wrestle with and a lot of challenges, but again, I think we've got to get better at these four things. Bringing in new data, getting access to more of the business data, improving the design of our studies to make sure they're really answering the questions we think they are, connecting those results back to solutions, and then, thinking about, once we're really successful, how do we winnow down the possible questions to the ones that have the maximum impact on the business.
So with that, I'm going to turn to some questions that have been sent in. So one question that seems to be coming in pretty frequently is, are there any examples of how HR analytics solved a business question? I think there is a ton out there right now. So we've seen lots of companies present this. One of the recent ones we looked at was really trying to understand how a company can increase innovation.
And so what we were able to do with a partner company is go out and take a look at the different innovation groups that sit inside the organization. Look at the patterns of behavior inside those teams, look at how knowledge was shared, look at who was on the team, the diversity of the team. Look at how well that team was connected to other teams inside the organization, outside the organization.
And we collected, with this organization, a really rich pool of data to say, what do the most successful teams look like? How do they compare to the average team? How do they compare to the lowest performing teams? And really start to understand that the makeup of those teams mattered in terms of the diversity of skills and backgrounds, the amount of connections they had inside the organization to other teams, to leverage learning that had taken place somewhere else and bring it back, were the two key factors.
And so that organization was then able to take that and say, well, how do we recreate that? How do we make sure we get the right balance on these other teams in terms of diversity of skills and backgrounds? How do we make sure there's better connections across all the teams so that this natural knowledge sharing across the business matters?
So that's, I think, one great example. But a lot of interesting things have emerged over the last couple of years as teams have presented out in these working groups on analytics. So if you haven't joined us for one of those yet, I would encourage you to keep looking at the website for upcoming working groups.
Another question in on this example of greatest and latest analytical methods. So how do we do these things better? I think one of the things that's really emerging as a great tool is this notion of natural experiments. And this really gets to the question of how do we change something, and how do we change it more effectively?
And really seeing companies lay out, here is three possible alternatives. We're not really sure which one is best, so instead of just guessing which one might be better, actually taking the time to lay out each of the three in fairly similar samples inside the organization and seeing which of the three works best. And then, using that data to then drive a broader launch across the organization as a whole. But really that notion of natural experiments is really good.
I think what we've also started to see is some combination of what I think of as qualitative and quantitative research at the same time. And so the great example of a company that was really looking at productivity and looking at teams and which teams were more productive than others, and they tried to study this all through existing HRIS data. Right They looked at the diversity of the team in terms of language and skills. They looked at the length of the time the team had been together. They looked at the shift. And weren't really finding anything to be big drivers of differences in team performance, and they were able to couple that with going out into the field and actually observing the teams at work.
And what they actually found was those teams that were most productive stood closer to the equipment. So no body of data would have existed inside the HRIS system to actually see what the real cause of this was. So I think another thing is to think about, don't be stuck just in your office with the data that you have. There is always a need to go get more, there is a need to go observe what's actually happening, and a richer understanding of the context and the situation. So a couple different examples, potentially, for you to think about of new and different techniques. But there's always new things happening.
So again, to me, one great way to think about this is, you don't always have to own the capability. So can you go and partner with a local university? Can you partner with some HR faculty members that may know your business really well? Or faculty members in I/O psychology, organizational behavior, or economics. Whatever it might be that can help come in and say, well, here's a new technique that might work for you, or here's a way we can model that data in a new way with some new technique that's come out. So if you're trying to keep up on all that yourself, I think you're in trouble because things change so quickly.
So another set of questions, and this is a common thing that's emerging is, where should this HR analytics team sit? Should it be inside of HR? Should it be in a broader analytics team? Should it be in a data team or an information team, technology group? And that's a really hard question, too, because I think it depends on how much capacity the HR function has to put resources into this.
So certainly, the larger companies we see are structuring their own HR analytics teams, and they have those three components. They have the data integrity team, they have the reporting team, and they have the analytics team. I can certainly see, though, where it helps to have them at least crossing over or interacting with HR-- or analytics teams, excuse me-- from other parts of the business.
So if we need access to marketing data, do they have some personal connections or line connections to the marketing team? Do they have connections to the operations team? Where they can get access to the data they need and the kind of data they want in a timely manner. So I think having some natural connections, either that's because they've got strong networks or connections to those other analytics teams that exist inside the organization, or because there's a reporting line authority that give them some of that access, I think is important,
And we're also seeing, for some companies, what they've done is, because they don't have the capacity to own or build that team themselves, they leverage the data scientists from other parts of the organization. They already have economists or psychologists or marketing experts that are deeply savvy in different techniques for analyzing data, and they leverage that talent internally.
So I think, again, it depends on size of the company, how many of these questions you think you have, what kind of resources you have to build your own team. But certainly, even if you built your own great HR analytics team, you want them connecting with the data scientists and data people in other parts of the business.
b questions. So lots of questions coming up around how do you build training? How do you develop people to really enhance the capability of the analytics team? And I think of this in at least a couple of ways. So one, I think a lot of companies have seen some huge benefit from upskilling the generalist population, and even some of the specialist populations, to really ask better questions.
So helping them understand what kinds of questions can be answered through data, which of those questions can they answer themselves by just monitoring their own changes and their own dashboards, their own reports, and which kinds of questions might be better answered through the collection of new data through more sophisticated modeling techniques. And really helping them think through, what kinds of questions can I answer on my own? Where do I need the help of the analytics team to get better at this? And then, on the backside of that, helping them take the data analysis that's been done and having those generalists help to design potential solutions.
I think in the middle part on that-- how do you train people to be better at the analytics piece and the data modeling, the research design-- I think the only real solution is, unfortunately, buying that on the open market. Whether that's buying people with PhDs, master's degrees in I/O psych, in HR, in OB, in economics. So you don't necessarily always need a PhD. Sometimes, master's levels is clearly enough to understand the techniques.
But again, you've got the rest of those capabilities to build into the team as well. So that might be bringing in a former consultant to the team that can help through the process or consulting component. It could, again, [? being ?] bring in the generalist or even a business line person who deeply understands the business. So there's a lot of those capabilities that have to connect together, and it's really, how do you manage the personalities and the balance of those capabilities? But the real deep knowledge around the techniques is really hard to train through a short one hour, or five hour, or two week session on anything because it often takes us years to develop a PhD to really think through what's the right method, the right technique, to analyze that data and model something in a key way.
Lot of questions, too, around, how do we do this well, inside an organization? So a lot of times, what I've seen is, companies that are best at this, have natural organizational units that are easy to compare. So that question I had around [INAUDIBLE] a company that's around why are some production teams better than others? All these teams were doing the exact same kind of work. So they may have been slightly different in size-- some might have been five members or four members or six members-- but you can control for size effects by the way you model the data.
But the bigger question is, how comparable is the work? So if you've got-- and we did see this with one of the companies that presented. They presented some data and analysis they did around diversity and the impact of female diversity on performance. And what they'd found, through their research, was that the greater the percentage of females in their manufacturing workforce in a plant, the greater the productivity and performance of that plant.
And they were really excited about this. It was a great story, until you dug a little bit deeper into the data, which they didn't do right away. And to say, OK, but did we control for differences in plants? And it turns out that those plants that were most productive and had the highest percentage of female employees, were also the newest plants in the system. So of course they had newer technology, newer production lines, and that was really the biggest driver of performance.
So again, I think we have to be really careful, and that's where someone that really understands the business, how the work is done, how comparable that work is across groups, really matters. Because otherwise, you can come up with some false analyses that tell you one thing when, actually, something else is the real driver. So again, of these capabilities, I really stress you've got to have that deep knowledge of the business and how the jobs actually work across those different settings to make sure they're comparable.
I have time for at least a couple more questions. Let me dig through this. So lots, again, questions around what are the questions that matter most? And which questions can impact our business the most? Again, this is where I would push you to get away from comparing yourselves to other organizations because you're all at different stages of this. And I think what you want to do, particularly early on, is work with the senior teams in the business itself to say, what are the biggest issues we're having as a company? Where are we seeing the biggest issues in terms of satisfying our customers, hitting production targets, hitting performance targets?
And then, really work backwards to say, is there data that we can collect, that we can analyze, that can help understand where there's a people component that's really driving this? So I always push-- as my former consulting genes-- is to answer the big questions, but the big questions that are important for your business. Where are you not hitting the right targets? Where are you not hitting your numbers in terms of performance, customer outcomes, productivity outcomes, operational outcomes? To work backward from there to say, OK, now, which of those might have an underlying people problem? That's where I think that funnel thing becomes really important, is to pick off the two or three questions that are most important to you because that's what's going to really establish the credibility for the analytics team as a whole.
Last question I'll try and take on, which I'm not sure I have a great answer for, is this notion of predictive analytics. So same problem, here. Garbage in, garbage out, particularly when it comes to predictive analytics and workforce planning. And we see a lot of companies really try to put a lot of energy into this notion of, how do we start planning and predicting how many people we need, what skills we're going to need, what skills are going to be a gap? Because a lot of that is going to be predicated on having really accurate marketing data.
And I think what we tend to know is that often, that marketing data is really good for three months out, not so good at 12 months out, and really not so good at 36 months out. So when we start thinking about HR planning, we want to plan what the people needs are for the next five years. I think the farther you get out, the vaguer or the trickier that gets because a lot of the inputting data from the marketing team or the operations team is not so accurate because they're just a projection.
So I think when we get to that point, particularly when it comes to predictive analytics, we have to get better at the modeling of different scenarios. So what would it look like at 10% above that, 10% below that, 20% above it, 20% below it? And start to get what we'd call is academics confidence interval. So here's what we're really confident we're going to need. Here's where we are a little less confident we're going to need. And here is if there's a best case scenario, that's the biggest thing. And again, what you're probably looking for is spotting the biggest gaps. Where are we going to be in most trouble in terms of capability shortages, headcount shortages, headcount overages.
So we can start to think about making decisions now to help balance that. So how can we get on a trajectory to start answering some of those questions. But it really becomes predicated on that access to other information other than just HR pieces. So what's marketing saying? What's operations saying? What are managers saying are the key capabilities to drive the new business direction we're headed towards? And so then, wrapping that all up, we can start making projections across a range of scenarios, again, to give us confidence intervals around, here's the lowest level of need, the potential medium level, and then, if all things really clicked, here is the high end of that.
So again, I want to reinforce that this analytics thing is still fairly new. So a lot of these things I'm saying sound great but often come in baby steps. So think through first, do you have the right data? Do you have accurate data? Do you have someone that can help translate this into action at the end of the day? Do you have people that are truly, deeply and understand what's going on in the business so we're not making false predictions because we haven't collected the right kind of data or misinterpreted it somehow? So again, I would push all of you to really think about the balance of that team and again, where you can rent some of it or borrow some of it versus owning it all together.
So thank you, all, for joining us today. We will certainly schedule another one of these. I know there's a bunch of questions I didn't get to, so potentially, what we'll think about doing is scheduling a second one of these in a month or so to follow up on the questions we couldn't get to today. Also, make sure you keep your eye to the CAHRS website for all the other upcoming working groups we have on this and other related topics. And we look forward to seeing you all again soon. Thanks.
This has been a production of the ILR School at Cornell University.
[MUSIC PLAYING]
Chris Collins, associate professor of human resource studies and director of the Center for Advanced Human Resource Studies (CAHRS) discusses how to optimally structure an internal HR analytics department; key questions that the HR analytics team should be able to answer and the data requirements for these questions; and thoughts on future directions for HR analytics.
