John Kelly Live on Shally's Alley

Hey everybody, you are live. On Shelly's Alley. Wake up, it's Friday. It is Friday, April 12th. Geez, the year has gone by. What does that make it? January, February, March, April. We're in the second quarter, right? January, February, March, January, February, March, April. So this is already the beginning of the second quarter. So for those of you out there who are looking for a job and your recruiter said we won't be hiring until the second quarter, that's now. Oh, so, uh, very interesting guest today with us here live on Shelly's Alley. It's completely unscripted. The gentleman that you see in front of me with the very classy, uh, grandfather clock and collection of, I'm sure, highly important and very expensive books is John Kelly. John actually invited me to a, uh, well, I don't know if it was a podcast, but he wanted me to do an appearance on his regular show, and I didn't show up. Can you believe that? I had the wrong time, and just, I was like, hey, when's the show? And he's like, you mean the one that was an hour ago? I'm like, uh, whoopsie. So he was very, very kind to me and let me make up for it by recording a video, and then he posted the video on his YouTube channel. Did a really good job with editing and everything. So anyway, he's got a channel and his YouTube channel and his actually also LinkedIn group are about HR science. And of course, that's the fun stuff we're gonna talk about today is we did it all for science. So John is a scientist who specializes in people analytics and a topic that we have not had anybody on Shelly's Alley about, which is compensation. Yeah, that's right. You heard that here first. So, people analytics, org design, and compensation are his specialties. I was really curious to interview John for a couple of reasons. I'll post your channel and your group up there. In a second. But I want, what I wanted folks to kind of hear a little bit about is this other world that as recruiters and people in talent acquisition, we often run into. But when I say run into, it's kind of like you're doing a fast sprint and then out of nowhere comes a wall and you like flat right into it. Right. And that is compensation. You know, we have candidates that are like, well, you know what, the market's really paying this. And you're only offering that. So yeah, good luck. And, you know, as recruiters, we gotta go back to our hiring managers and to our TA leadership and the business leadership and say, can we please have a little bit more money? This is what the marketplace is. And then of course, you know, the compensation team says, well, that's not what our survey says. So here's the guy that told you that's not what our survey says. So can you please tell us about the, like, just tell recruiters what it's like to be in compensation and deal with recruiters. Absolutely.

Like, the story. Absolutely. So yeah, so a lot of my career has been compensation. I did come up through the HR side, not the science side. So I've been doing people analytics in benefits, compensation, and then itself for the past 20 years. But most of that has been in compensation, right? We are sort of that brick wall, right? And like, very honestly, like, I feel terrible going back to recruiters and saying like, hey, you spent a month on this awesome candidate and everything's all ready and now there's a slight tweak and I have to say no to it. Right? That's not a spot that we wanna be in. So with compensation, like you're talking about, there's these market surveys and there's a few different ones out there. There used to be more and they sort of consolidated a little bit. And very honestly, none of them are good. All of them have sort of messy data. They don't give it really to you in the right format. It's sort of a pain in the butt to sort of work with them, but it's the best data that we have, right? So we take all this market data and we sort of say like, okay, so in this location, what is this job paid? And they say the quartiles, they're not literally quartiles anymore, but they sort of give you something of like, you know, what's the bottom 24th percentile? Um, you know, what's the 50th percentile? What's the 75th percentile? And usually a few more depending on where they are. And then we have the conversations with the CEO and the head of HR and sort of say, where do we wanna pay? Right? There's a huge problem right now where since tech roles are so important, everyone's like, well, we gotta pay them at the 75th percentile and everybody else we pay at the 50th or below sometimes, which is not fun. What happens to that, right? We all pay tech jobs a little higher, right? The survey goes around, we submit all of our data to the survey and now, hey, tech jobs were a little bit higher, but now I'm paying 75% on the higher one, right? So then it goes a little higher and a little higher and a little higher and a little higher.

Did they ever come here? Sorry, does it ever come down?

No, it doesn't. Exactly. Exactly right. So there are some roles like this and like, honestly, I love technology. I do coding in the work that I do. I love analytics, right? So like, I love tech jobs and most of my friends are developers. I actually went to school to be a developer for a little while. So like, this is nothing against that field, but if you sort of look at the value added to it, the compensation sort of becomes a little misaligned just due to these sort of HR practices, right? So it's a lot of stuff that we're trying to do within the compensation team to sort of balance these things down. And there are some roles like tech jobs, usually a lot more men applying to those roles than females. Usually a lot more whiter people applying to those roles than non-white people.

Oh, so this is all stuff that you're calculating. Like you've got, okay, so you're not just looking at like internal equity, you're actually looking at external gender diversity and racial diversity and tenure and okay. So how far behind when we get like For example, I was at a company and they said that they were gonna be doing this. They hired a company to do salary reviews and then they bought some data from a couple of companies and they kept telling us it's gonna come out, it's gonna come out, it's gonna come out. It took like 6 months and then it finally came out and we finally had our bands, our pay bands. By the time that comes out, how obsolete is it?

Very obsolete. Honestly, something that I've implemented at a few different companies is just getting rid of pay bands entirely. Right? Because they never work, right? Because if I give you a pay band and I say, you know, as your recruiter, your pay bands for this role is gonna be $80,000 to $100,000, right? You're gonna be looking for around $100,000, right? Mm-hmm. Because you wanna get the best candidates in the quickest, right? You know, your incentive is to push it and you're probably looking a little over $100,000 too, cuz you know, you can sort of push that a little bit, right? So what's the point of the pay band? Why, why do you even say $80,000, right? That was a waste of time, right? A waste of effort, right? So something I've been trying to implement to sort of make this a little bit easier to say like, our target for somebody who's doing a good job in this role is $100,000. You find somebody more junior, they're gonna be a little bit below that. You find somebody a little more senior, they're a little above that. But now we can have this conversation. You come back with $150,000, is that person really that amazing? We can sort of have that a little bit, right? But like if we put in all these fake pay bands and then people break through them anyway, what was the point? We're fooling ourselves and we're wasting a lot of time.

Right. Okay. So they're pretty much already obsolete. And how are those, so how do we arrive at these pay bands? Is it, a calculation based on what people are currently making? Or, you know, because you hear about these companies that sell this information and there are consulting firms that actually do this. So there must be a whole process to it. Where does that come from?

Definitely, definitely. So pretty much the data is our data, right? So we work with different companies, we take all of our data, we send it to one central spot. The big ones right now were Radford, Mercer, Gloggen used to be a big one for finance. There used to be a lot of other ones that sort of got consolidated into a lot of these, right?

I heard about Radford and Mercer a lot, right?

Yeah, they're the two big ones. Radford's really big in the technology space. Mercer's really big in the global space. I used to work for Mercer. I had an internship there. I love the company. It's a really cool spot. But, but yeah, so, you know, we send all of our data to them and it's pretty much every employee and their compensation amount and some other stuff too, sort of based on their department, their job title, and whatever. This job sucks. Sending that data—

And then you don't look externally?

Sorry?

And you don't look externally, so it's just our data.

Just our data. So we send all of that to them, right? But the problem is that since it's so annoying to sort of like tweak all this stuff, it goes to the lowest level compensation person on the team.

Right.

That person doesn't know a ton about jobs, right? They're getting sort of this thing. So they send in this really not great data. We sort of throw it together with all this other sort of not great data. And then Radfruit sends like an overview of that back to us. And that's the things that we build things like the job bands off of, right? So it's a lot of not really perfect data that people are submitting. Sometimes they're sort of playing games with it, right? Like I'm sure all of us have had a time when somebody said their job title was data scientist when they were really doing reporting, right? Yeah. Because they know data scientists make more money, right? So there's a lot of that sort of gamesmanship.

Got it.

Theoretically with the law of large numbers, right? Like if we get a big enough sample set and stuff is sort of, you know, equally balanced, a little high and a little low, it should come out to the right number. It's not balanced, high or low, it's always going high. Nobody ever is underestimating their types of stuff. The sample sets just aren't big enough a lot of the time. The data we're starting with is not really great to begin with. What we have to do is try to say, here's this data and it's giving us some insights. It's the best insights we have. There's not a better data source or else we'd be using it. How can we build a model influenced by that data but not just based on it?

But that's also realistic with what's going on in the market.

Exactly, exactly right. So that functionally is the market, at least the best we can do to it. The only other way to really do this in what I would sort of say was more a scientific way is look at government data. So the US has a lot of really great stuff going for it, but especially a lot of the data collection that we have is surprisingly good in the US and just not as good in other places. So we have the census, which is written into the Constitution, so we have to do it in a certain way. There's something else called the American Community Survey. And it's not written into the Constitution. So they're a little more freedom to use more like more contemporary data methods in that. So like sampling and stuff like that. That data is free and publicly available. So you can go on right now and get huge amounts of information of how people are being paid across the country. The jobs aren't as specific as we really need, right? We really like to look at these very zoomed in jobs.

And instead they're like, you know, construction foreman, which could be anything from Houses to buildings to nuclear plants or— Okay.

Exactly. Exactly.

What about BLS, the Bureau of Labor Statistics? Do they have also— Is it different data or—

Sort of the American Community Survey is sort of the best with sort of compensation data. The ones that— That's the only one that I've sort of seen really used in practice pretty successfully. But what's really good with them is that they have data down to the zip code, which you never really get very well from Bradford or Mercer or salary.com, right? So there's sort of some balance, right? Do you care about the jobs more? Do you care about the location more? You know, I've done it before where we sort of merged both datasets and did some sort of interesting stuff by sort of looking at it that way. But a lot of compensation was really just like taking something from Radford. It says this job should be $125,000, and then we set our range to $125,000, and that's getting really outdated.

Yeah.

Right. Okay. We need to use this data and sort of understand like the problems with it and where it's successful, and then have it inform what we're doing. Based on a lot of different factors. Like, you know, which are the teams that we want to grow really quickly, right? What are the areas that are really critical to us?

You want to pay more for the areas that are more critical, even if it's paying more than the band. Okay. So, so definitely not the PLS and census, if there hasn't been a census, might be quite dated. What about like W-2s? What about the information we collect from W-2s? Centralite, like, does the government somehow have an analysis of actual tax filings? Because that would tell you what the job title and the salary is, wouldn't it?

Right, right. That would be amazing if we can sort of get at that. So to my knowledge, yeah, that's not public information to my knowledge. I do know that sometimes there are certain researchers that are allowed into that. But even if we took a step back a little bit, just at the census data, right? So with the census, it was like the goal of it is to literally get everybody in the country and like they get surprisingly close, right? Not there, but surprisingly close. But if I just, you know, sort of gave you this like very high-level census data of, you know, these are some averages and these are some medians and this type of stuff, right? People have ways to de-anonymize even that, which is pretty amazing, right? So the census itself actually has to tweak and lie about some of their numbers very slightly to make it really hard to sort of de-anonymize that type of stuff. So even at that level, it's really hard to keep it confidential. When we look at stuff like W-2s and things like that, they're not going to release that because, you know, you know, you could figure stuff out, right?

Like if you happen to be the only person with that job title at the company, then it's probably you. Yeah, that makes sense. What about like, so survey, you know, surveys where they, like, for example, I'm thinking it out loud here, but I remember typing this and like LinkedIn is asking you now, what do you make? And Glassdoor and Indeed and PayScale and places like that where they actually ask you for whatever. Services that you're like for LinkedIn, they tell you, oh, it's to show you better jobs or whatever. And of course they tell you stuff that you wanna hear. But what about that? Is that even remotely accurate or do a lot of people just kind of lie on that?

You know, that like, there is always the concern of people lying. In my practice, I very rarely see that. And I know this isn't answering your question, but to sort of take a quick aside, I ran an employee survey for a hedge fund years ago for, I think it was 1,200 people, something like that. So a reasonable size organization. And the first year I ran it, I continued to run it what was done previously, fully anonymous, right? And the next year I was like, you know what? We're gonna not anonymize this, right? So when you write that you didn't like this company, your name is gonna be attached to that. And what I said to everybody is like, only the function of people analytics team is gonna be looking at this, right? We're not gonna tell the CEO that you don't like, you know, the color of his hair or something like that. Right?

Okay.

But what we can do with it then is that we can sort of see, look who answered the survey this way. And then 6 months from now, who left? Right? So what are those questions that really indicated who was gonna leave? So we have a lot more leverage on sort of survey data. And look, I did the survey, you know, 2 years in a row with this one anonymous and one not anonymous. There was no dropout from participation and there was no shift in our engagement scores. Right? So I don't think anonymization is really the thing that most people sort of say that like people answer a lot more truthfully. I don't think they do. I think a lot of employees are looking for avenues to sort of share that. The same thing with exit interviews, right? People say like, oh, we need to, a form for an exit interview because people aren't truthful during the exit. People are incredibly truthful during the exit. I used to be head of HR and I did every exit interview. I thought it was super important. And people told me the worst things very truthfully because they're going out the door, right? They've been bottling this up for a while. They want to share with somebody. They want to be heard, right? So the sort of the anonymous side of it, I don't think that that's really there in a lot of this type of stuff. And I think we can be a little more public and a little more open with it and still get sort of the same results for the data.

So it's not really being utilized, like LinkedIn's data is not being utilized, only they use it obviously, but they're— what about like Glassdoor and Indeed and PayScale and stuff? Do they use that or does that ever end up in our comp charts? It does.

So no, definitely, definitely.

Okay.

It sometimes does, but I'm one of the ones that sort of questions against a lot of that.

Yeah.

Right. Because if you sort of think about it, right, assume that we have a lot of really messy data, right? So if your job, let's say you were getting paid $100,000, right? And the data that comes in might be like 90 or might be 110 or something like that because like someone maybe misclassified you a little bit, right? Or something along those lines, right? So everyone has sort of messy data. What I think is an analyst, you might think it's a senior analyst, all this stuff that's like not—

Oh yeah. Right, right.

The only way to sort of get around that would be to sort of just get a ton of it, right? So if you think an analyst, you think this job is an analyst and I think it's a senior analyst and we both do it, the number's gonna be funky if 1,000 people did it. It's probably gonna come to like something at a reasonable spot, right? The more people sort of doing it, the more data you have, the better. The problem with a lot of these is they have very, very small datasets. So people ask me, is there like a boutique compensation survey just for a very small area? We don't want a small area, right? You want the biggest area possible, right? When I give them the data, they're like, oh, can we cut it this way and this way and this way? Can you just find people who are in this type of job in this location?

It gets too specific.

I can, but then you have 15 data points, right? Where if I cut it another way, we can have 5,000 data points, right? So that's the only real like push we have against sort of this like crummy data. So the best thing that we can do is sort of like zoom out as much as possible and then be very careful about where we zoom in and use different methods, right? So before I talked about the American Community Survey, which is again, sort of like the census, but it's done more often. We can sort of take some of the data from that and have it influence some of our data. So if I felt that we wanted to look at, let's say I'm in New York City, right? Let's say we had a huge office in New York City and only a few people in other locations, right? And we had like maybe like 5 people in Iowa. The data in Iowa is probably not very good 'cause there's probably not a lot of people in Iowa doing what we're doing. So we can mark on Radford the New York City rate and really get a good idea of that. 'Cause there's a lot of people in New York City, there's a lot of people in our company in New York City, get a good idea of that. And then we can use something like the American Community Survey to sort of say like, well, what's the difference between pay rates in New York and Iowa? Okay.

Right?

And sort of leverage these different data sets together a little bit. So yeah, so the biggest thing to sort of remember in compensation is like, as much as we can zoom out, we get better data, and then we zoom in only exactly where we need to.

Okay, that makes more sense. So let's pivot this over to the other really interesting thing, and that you and I kind of touched on a little bit, is org design. And so you said that you didn't, you don't usually get access to be able to work on that, but you've recently been able to get into org design. First of all, define org design for those of us that don't know what it is.

No, definitely, definitely. It's a really good question. Cause I think there are a few different ideas sort of going out there. The way that I would decide to define it, right, is that if you have a recruiter, right, they're trying to sort of get that job. You have the step before that, the sourcer, and you're trying to get a lot of different candidates and trying to like find the right people for that job. The step before that should be organization design, which is sort of saying like, what should this job be? Right. If we have a marketing team, should we have 5 managers and 15 employees, or should we have 3 managers and 20 employees? Right? You know, how do we wanna set that up? How do we look at, you know, connecting that to compensation data? All of that's sort of like really developing what's gonna be in the job rec should be the org design stuff. And it serves sourcers really well, right? Because like, I would wager there's a few times that you've sourced a candidate perfectly and the hiring manager's like, well, I changed my mind. Right? We wanna do something else. We wanna have someone a little higher level, you know, whatever. Right? This is sort of trying to have that conversation really early, but not having it by candidates, right? Not even having it by role, but having it by team. So if we can sort of go back to the hiring manager, right? And sort of say, hey, you run the marketing department, tell us how marketing works, right? What are the areas you want to grow over the next 5 years? How is it working right now? Where are the stumbling points? What do you want to sort of do? Really develop a really good hierarchy to that and a really good plan to it. And then come back and put employees into it. Because you can develop this hierarchy without employees and then sort of say like, okay, I think Harry would actually be really good to get a promotion over here. I think Sue would be okay over there. And hey, look, there's an open rec here, right? And sort of go from there. So it's sort of really trying to sort of get in front of what we're doing in sourcing. And also, of course, in things like compensation and performance and all of these areas to have a really strong idea of like, how can HR bring our special data, bring our special knowledge to help develop a team and really sort of help the manager, excuse me, help the manager understand what are the methods to sort of make a team better.

Got it.

Got it.

So instead of the manager saying willy-nilly, okay, I'm gonna have a team of 5 and here are gonna be This is the HR function, the people function coming in and helping them structure that to get more efficiency and improve communication and essentially drive better results using the, I guess, the psychology of organizational design.

Yeah, definitely. Definitely. So one of the things that's just a really easy example, right? Is that if you're in a startup and you're growing really quickly, a lot of my career has been in startups, right? You start off with one person that does X and they wear a lot of hats and you grow and now you have 10 people that do X and they wear a lot of hats, right? Should we have 5 people that are doing X and 5 people that are doing Y and sort of split up the role that way? Should we let everybody sort of continue to sort of do everything? Is that more efficient? You have to have a call center maybe that is, because you just need a lot of people to answer the phones. Should you have a, like a badging system where you sort of come in a generalist and then you get badged to sort of be a specialist and you get more money that way? So just like conversations like that, that a lot of managers don't really think about. But it has a huge impact on the employees, right? You know, how many levels do we have? You know, how many of them are going to be managerial?

Right.

You want a specialist manager or do you want a generalist manager over the team? You know, every time I have these conversations with some of the business, they come in and they're like, I need X, Y, and Z. They're very firm.

And then you—

what are your concerns? 2 weeks later, it's nothing like that anymore. And they're much happier.

Interesting. Okay. That's great to hear. So is it possible? How do you get into that? How do you get into org design? Is it you go in through analytics? You're going through HR? Like what's the career transition for that?

And I think actually, I think you hit on a great point with this, right? Because like this is one of the problems. There's not a lot of easy ways to get into it, especially if you didn't have sort of a history in HR. So like I never got an org design role. I got really good at compensation and then I always sort of built an org design practice either within—

Outside of that. Okay. Got it.

Because at the end of the day, like the things that you need to think about if we're sort of developing a team are similar to what you need to think about for compensation. What's the thing that needs to get done? What are the jobs that are gonna do that thing? And then have those jobs sort of interact with each other. So getting really good at comp really helps you good with org design, right? To sort of understand that. But then there is also that larger element where in compensation, you know, obviously we're focusing on money and those aspects. And this is really sort of taking a step back and sort of seeing like, how does the team work? You know, you know, what are the different models that we can sort of use for the team and sort of do that? So I do hope that, you know, org design does become just like a normal specialty that people can go into. I hope there's a lot more, schools for it and a lot more written about it that's really useful. As much as I love our HR VPs, I think that this is a specialty itself and I don't think just having an HR VP trying to do this without more training is always the most successful because it's really data-focused, right? Understanding what a team is, being able to look at things like turnover, being able to look at why people have been turning over, right? What aspects of this team are really working and which ones aren't? Look at the compensation. It's sort of a pretty data-heavy focus with, of course, a lot of conversation to sort of make people feel good about it.

Okay. All right. There's one other question that I wanted to ask you before we run too far out of time and need to wrap things up. But I wanted to ask you about— well, let me preface this question with saying that there are two areas in people analytics. Switching now to people analytics, which is squarely in your, you know, in your long-term career. You've been doing a lot of it in different ways from HRIS to whatever. But it— so I've been toying with people analytics for the last 7 years, not nearly the depth that you have. But I did essentially accidentally stumble into it when I was getting tagged to do workforce planning stuff and talent pool mapping and identification. And that sort of led into workforce analytics and people analytics and so on. And then I actually had a job. Job for 2 years doing people analytics, global head of people analytics. And in my experience, as amateur as that was, there are 2 significant— well, there are at least 2, or at least these are my 2 challenges in doing the job. One of those is figuring out or understanding what to analyze. So, not necessarily the formulas. There's a lot of help on the formulas. There's a lot of help on, you know, what you should be— the questions that you should be answering, what the leadership team wants. And in some cases, the company's leadership team just tells you what they want. But the challenge that I encountered was in figuring out what to analyze. So, you know, what data to pull. And the second challenge of for me was visualizing it, telling the story. And let me explain on both of those. So on understanding what to analyze, under figuring out what data to look at, there are a lot of different micro challenges within that. For example, there's data structure, there's data hygiene, there's like, you know, there's incomplete data and incorrect data, all these other challenges, but just forget about all the little challenges. The big challenge is what am I even, what should I even be pulling, right? Like what data should I be pulling? And then on the other side, on the visualization, it's not that I have a hard time visualizing data. I can tell the story. The problem is what story, because if you tell too many stories, they're not listening. And if you pick the wrong story, they're not listening. So it's almost like how do you figure out what date, what question to ask the data and how do you figure out what story to tell? And I, those are my two challenges that I, so what are, what do you find are the big challenges in people analytics? No, definitely.

You really hit it on the head. So the way that I talked to my analyst about that, so I currently have one analyst. I'm trying to hire two more up for my current role. And you know, I always go to them at the very start and sort of say like, hey, you know, you're joining the team. You're really good for the technical aspects, right? If somebody gives you a request that they need whatever report or dashboard or analytics or whatever, and you do that perfectly, you failed, right? Our job isn't just to create a visualization or report or something like that. Our job is to take our specialty experience and bring it to the business and really help them understand their questions better, right? So what you were talking about before, sort of like understanding where to look in the data, right? I have a pipeline that I try to sort of use, you know, these organizations and it called, you know, the, the people analytics pipeline or this HR pipeline, right? And it starts with the real world because we're real people, right? All the data that we're working on is about people and people exist in the real world, right? And then where we come in is that first step is abstraction, right? What is the data that we're getting? Because the data isn't the people, right? Our headcount report isn't people, right? It's sort of this abstraction of it and sort of make sure that we understand like the faults in that abstraction, in that sort of abstraction, right?

Yeah, actually get that. Yep. The big picture.

Right, right. Definitely. So then the next step is the analytics side where we do sort of the heavier analytics like you're talking about. And then after that, there's the deabstraction, which is usually where the HR team just sort of forgets about it and says like, "Hey, we found a correlation. Here you go." Throws it back to the business, right? And that's a huge failing too. And the last step is something like policy, right? Or something along those lines. And HR analytics needs to be a little more ingrained in all of those steps to do it successfully, right? So we need to be at that beginning to sort of say, "What data are we collecting?" Right? The number of times that somebody will come to me with an employee survey and say, "Here, analyze it." And I'm like, you should have called me when you're making the questions, right? I can help develop questions that are gonna get you better responses, so then I can analyze better, right? Or at the very end they're like, okay, give me this report. I'm like, why? Right? Why do you want this report? Help me understand the business question you're trying to ask, and I'm gonna give you some really data-informed information that's gonna help you answer that question. But it's probably just not gonna be that report, right? So we need to be involved in that policy side too. We're not gonna be the ones that decide every policy, right? But at the end of the day, like if we're not in that conversation, The people who are, are missing out of a lot of this, right? So like, that's the way that I try to sort of tie it all together and sort of make sure like sort of a little bit like the business partners were doing, right? Where before they were sort of generalists and they sort of did a lot of different pieces and it was sort of administrative and they were like, no, we could do more than this, right? We can understand how sales works. We can understand how marketing works. We're gonna be embedded in your team and really do that. That's the exact way that we should be looking at people analytics, right?

Right. And do you like the root cause diagnostic? And yeah, that's the part that I really enjoyed and I was able to do that. Quite well. And, you know, but I get what you're saying about the abstraction and the analysis, but you kind of left me behind, or I fell behind with the de-abstraction. Tell me about that.

Sure, definitely, definitely. So if we sort of think about these different steps, right, the only step that's really computer-focused is that analytical step, right? When we're thinking about like what type of data to collect in an employee survey or an HRS or something like that, like that's a big human decision because it's really important about sort of like thinking about it from a human perspective. Then the de-abstraction would be things like the visualizations you're talking about, reporting dashboards, anything like that. That's also a real human aspect because the numbers themselves, punching the numbers to find if this correlation is statistically significant or something, let the computer handle that. We have computers that do really good jobs at that. They just sort of do it perfectly. I use R very heavily in my analytics and I don't have to worry that I did the math wrong. It does the math perfectly all the time provided I set it up correctly. Where I need to worry is that deabstraction of if I put this number on a piece of paper, is the person who's reading it, is their mind gonna jump to the right conclusion? And this is a huge problem, right? Especially when I did a lot of DEIB reporting where I would do whatever dashboard or report and somebody would see that and say like, oh, we have 51% white people and then 49% Black people. That's a problem. And the next month things slightly change and say, wait, we have 51% Black people, 49% white people, everything's good. And it's like, that's not what the numbers are saying.

Right, right, right.

That's somebody who didn't understand the dashboard and the numbers.

Oh, and so they just see the numbers and they make their own decision based on that, but that's not necessarily what the data is saying. So it's like, it's the interpretation of— okay, got it.

So we— and it's our job to build a dashboard that's gonna help them come to the right conclusion. And this is why dashboards actually don't work. Pretty well because I can build a perfect dashboard that puts in the correct information and all of that. And somebody reading it is gonna look at numbers that I haven't seen, 'cause the point of the dashboard that it runs without me looking at it, right? And they're gonna come up with a certain conclusion, right? So we need dashboards. I'm not anti-dashboard, but most of the time it has to be a better conversation with it so that they can say like, oh, well, does this mean this? And the human from my team is gonna be like, actually it doesn't, right? Or we can look into that more to see if it is connected or sort of give that sort of context, right? And we can build some dashboards that do that. That really can sort of give you— so if you imagine something that's like a turnover dashboard, right? And it sort of has this chart with like the percent of turnover for one team, right? And then under it has another chart that's like, oh, this is the voluntary turnover. And then under that it has like, this is the involuntary turnover, right? You got some insights from that, right? You were able to sort of see like a little bit of the story of like, oh, turnover is really bad this month. Why? Oh, it's really involuntary turnover, right? And we can build dashboards that sort of help tell the story. Just really hard, right? Because we don't know who's going to be listening to that story. So we have to really, really tweak it really well. And like the, the, the last piece about this that's the most important is that like tweaking that is based on sort of morality more than visualization.

Yeah.

Right.

Okay.

It's sort of a moral imperative to make sure that the person who's making this decision has the information they need to make the right decision or else the data's gonna be pushing things in the wrong way. Yeah. And that's the, the, the worst thing that could happen in analytics.

Well, I thought I had, I thought I had the last question, but I lied. I, you just made me think of something and I, and I, I'm very fond of my, I'm very attached to my definition of the difference between dashboards and scorecards, but I would love to hear yours.

Interesting. Interesting.

So, the difference between a dashboard and a scorecard.

Definitely, definitely. Coming from compensation, so we do executive scorecards a lot.

Yeah.

And with them, if anyone hasn't sort of seen it, the executives will have a few different goals and then they sort of have, you know, increase revenue by 20%, right? And then at the end of the year they'll have, oh, we increased it by 25%, and then there's some formula that sets their executive compensation based off of that, right? So, so I would say that to me, scorecards are sort of making sure that there's some sort of metric on top of it to sort of say, are we meeting or not meeting this metric, right? So it could be turnover. We want to have turnover under 20% this year. So that's the, you know, scorecard aspect of it. And currently turnover is at 22%, you know, we really need to figure something out. Right? We're reporting, we just have a lot of different reporting that doesn't really have that metric. There's a huge push for OKRs. I'm not a fan of OKRs, honestly. I think that it's a solve that people feel is pretty universal, but only works in sort of certain use cases of it. But an OKR, like at the heart of it, right, is trying to sort of say, let's use a number to sort of say if we were successful or not. So I think that there's—

So an OKR would be more akin to a scorecard?

Exactly. Exactly. So I think there's a lot more interest in scorecards now, which is good. Having some metrics gives people something to sort of, you know, tie to, right? So it's sort of foundation point, right? Because like, honestly, if, you know, I know that you've been doing this for a while, but if you got a new manager and you said, hey, your turnover is 17%, is that good? Is that bad, right? You know, how does the manager know what that's going to mean without giving them something else, right? Same thing with compensation. You know, we said, hey, I'm going to give you $150,000, right? You know, you have some guess about how the market's working by talking to your friends, right? That's pretty biased, right? You're in a pretty small bubble. You don't have a ton of information. So like, is that good, right? Is a 3% increase good? Well, entirely relevant on, you know, the previous job that you had. If you had 4% increases for your entire career, 3% looks like junk, right? If you had 2%, 3% looks really good. When I'm doing compensation, I don't know what your experience was at previous companies, right? So there's all this type of stuff that like, we sort of assume that like, oh, well, at this company, 3.5% is the average. So if you're at 4%, then you should be really happy with it and 3%, really unhappy. But we never tell the employees that, right? And it's the same thing with analytics, right? We sort of—

You should be happy with your 3%. That's all we can afford.

Right?

But that's less than the depreciation of the dollar. Well, it's still less for us too.

I used to run that report every year just for my friends. I'm like, hey everybody, you got your increases, right? Here's inflation, here's depreciation.

Fuck you.

Yeah, right? Yeah. So, but yeah, and this is the thing that I really hope that people analytics is going towards, right? Of sort of not just sort of being in the corner and just like running this report or dashboard or like whatever, but being really involved in the business. Because like, you know, I've been a manager too. There's times that if I just had this one more piece of information, I could do a better job being a manager, right? I could manage my team better. I could sort of make sure that I'm making them happier or I'm sort of keeping them for longer.

There's lots of different ways to do that.

But to do that, you need to be really involved in it, right? And that's where, you know, people analytics is going. And I would love to say if we were there, but And that's a little bit of why we're trying to do some stuff like this, right? Talk to larger communities to this, try to encourage people to come into People Analytics, try to make some videos, try to show people different aspects of it. There's a lot of different things that I think if we can do it well with people like you and me and get more people involved and asking the right questions, it could be a really different world a few years from now. And at the end of the day, your job has a huge impact on your life, so maybe people's lives could be a lot better.

Yeah, it's definitely a big impact. So I'd like to interrupt this broadcast for a special thank you to our sponsor who makes it possible for me to keep the lights on. This is a nonprofit broadcast here. And as you know, John, it costs money to do things like hosting and hiring the Restream platform and all that. So Gem very graciously has sponsored the first half of the year and made it possible for us to actually keep the lights on. So thank you very much to them. And also wanted to point everybody to John's LinkedIn and YouTube channel. LinkedIn group and YouTube channel where what he just mentioned about some of the videos that he does. I've learned some from his YouTube channel there. So, check out srcn.co/jk4tube or jk4hr and visit those two. But yeah, to wrap up, I think what you were saying in my summarization is I get asked a lot, very frequently, I get asked for executive dashboards. And to me, that seems kind of like a counter, like it doesn't belong. So, I wanted to clarify that around dashboards and scorecards. I see it this way. This is my— I'm very partial to this definition of it. To me, the dashboard is like the dashboard in your car. It tells you how fast you're going, Tells you how much gas you have in the tank, what your engine temperature is, how inflated your tires are, you know, all these system diagnostics. But the executive isn't the driver of the car. The executive is perhaps the passenger, and they don't need to see any of that. They just want to know if we're getting closer or are we there yet or how long we're gonna How long till we get there? So I like to take that dashboard information, and which already in my opinion should have metrics in it. It should have metrics. If you're going to show turnover, you need to show turnover versus last year for the same group or compared to, you know, other, et cetera. But I would like to take that and say, you know, for example, on the turnover question, if our normal turnover is 17% and we're tracking at 15%, then the the scorecard would literally just show a dial where in the middle is 15% because that's our target. And the needle either is above that, higher than average, or below that, lower. And then that's literally at a glance, we know if we're doing better or worse than we planned. Are we there yet or not? Right? And similarly with, like you said, we're trying to increase sales. I would have the dial that says 20%. As the max and then how far we are to get in there. And if they need to know more, they can click through to the dashboard. And the dashboard can say by quarter, by department, and all that other stuff. But what I found is that if you give people too much, if the dashboard has too many things in it, they'll either focus on the one thing that hurts them the most, that is most painful to them, or they'll either laser focus in on, oh my God, we have 17% turnover. This is terrible. How do we fix it? And ignore everything else. Or exactly. They get overwhelmed by all of it and they go, I don't understand this. Can you explain it to me? And it's like, okay, either one of those are not very beneficial. So just my 2 cents, you know?

No, no. I think it's the perfect analogy, right? Because if you look at cars from the past, right, there was a ton of stuff on the dashboard, right? And like, it looked cool, honestly, right? Like you have like all these gauges and stuff like that. But which ones did you use, right? If I'm a hybrid car, right? And I'm looking at it and the gas is almost, you know, I almost have no more gas. I take an action, I go to the gas station, right? Because they're very clear. See this thing, take this action, right? If we don't have that in our dashboards for everything, right? If an executive sees that turnover is high, what's the action that they should take? If you don't have that written out before you put that dashboard out, right, it's useful. You didn't do a good job, right? And if you sort of look at some more modern cars, right, so a lot of electric cars now, like, you need to still charge them, but you don't tell it— like, it doesn't have a little thing that like tells you like, you know, oh no, you don't have enough charge, right? It just brings you to a charging station. right? It's solving the answer for you, right? And I think that's where we need to sort of look at this, right? Is there a thing that we can sort of do in the computer that's gonna help solve this, right? And really get the fix to it? Or is there a thing we're gonna do in the computer that's gonna have a human that's gonna solve this and get the fix to it? Right. And just write all of that out beforehand. Right. And anybody can do this. You don't have to be, you know, in a data science team sort of thinking about it. Right. And put together a report, ask the person, well, what do you do with this report? Oh, I send it to Harry.

Oh man.

What does Harry do with the report? I don't know. Right. Right. Like, you know, everybody should be doing this, right? No matter like, you know, where you're doing it.

I can't tell you where this happen, but let me tell you a funny story. So I was, as a consultant, working for a large company, a worldwide company, and they brought me in to answer some questions about talent. So I had to do some audits. And one of the things that they took me to meet this analyst that had this report that was literally like an inch thick of paper. Yeah, this was a while back, right? It was a PDF, but most of the time he had it on his desk as a paper report. And I looked through this and I'm like, Wow, this is, this is a really deep, like there's some very deep data here that, that makes a lot of, um, you know, predictions and it's, it's, it's prescriptive and it's analytic and really in-depth and everything. And I, um, asked if I could have a copy. I studied it. I went back to my, my sponsor who was the, the person that hired me for the consulting. And I said, Who's getting this report? And he goes, I don't know. So I went back to the report writer and I said, who's, who's, who are you sending this to? And he's like, nobody. Like, who's seeing it? Is it being, you know, is it being published anywhere? No. So you just make this report every year and it just sits there and nobody asks you for it. And he's like, nope. Does it, does it go, is it, does somebody take data from here and do something? Like it was literally this guy's job just to produce this thick report that had a lot of really important information in it from multiple sources and nobody was even seeing the report.

Right.

I was just, I was shocked. Like, what's the point of this? Well, somebody somewhere, somewhere before someone at some point had asked the question. And had, they obviously were an officer and they had mandated that this report be produced. They have since moved on. And so the report no longer has a customer, but the report is still being produced because it's always been done.

And it's like, it's unfortunately so common. That happens all the time. And like, you know, I think what you're describing is like a pretty big scale to it of like making a, you know, a few inches of binders. But the reports that we run, right, somebody has Some like, oh, I think it would be interesting to know this data. I don't really care if it's interesting or not. Are you going to make a different business decision on it? Okay, we'll put together the data, build the dashboard, go through all this type of stuff, serve it to them. They look at it, they're like, oh, that was interesting, or wasn't.

Do we really need to maintain that for 6 more years after they leave?

It's not adding value to what we really want to do. And also it's like, it's not helping me. I'm not going to learn more stuff by building this sort of dumb report.

No, but if you're doing that report and something screams at you, you need to That's my advice is you see something in there, you need to report it to somebody. Don't just sit on the report. But yeah, so that definitely brings us pretty much close to the end here. There are two things that every guest is punished with when they attend Shelly's Alley and they are brave enough to brave the live episode. Because folks, you guys know this is not scripted. There's— John didn't know what I was going to ask. I didn't know what I was going to ask. I'm kidding. So the two questions are, first of all, who should be our next guest?

Yeah. So it's really hard because I was going to say Adam, but Adam is the way that I found your show. Adam does a lot of really cool analytical stuff. I have a good friend, Mandy, who I think might be interested if she wants to sort of talk a little more publicly. And then also—

It's up to them to accept, right? Just because you nominate them doesn't mean— So Mandy, what?

Yeah, I'm going to— I'll send you a link to it afterwards. I'll send you her link. I don't want to say her name in case she doesn't want to do this. She's not really in front of the camera often, but she's just such an amazing professional. And also our friend Maria, who runs an organization that does improving employees' performance, but in a way that I think is just entirely different than anything else I've seen. So I think they're both really cool. I don't want to sort of out them without asking first.

No problem. I'll definitely ask them and say, hey, just because you nominate doesn't mean that, you know, I don't push anybody, but I do like to discover new people, especially as we get into new topics. And, you know, I've been reading a lot about how retention is the new recruiting. So employee improvement, performance, employee improvement, that was not, that was a boring thing to say. Employee performance improvement or performance improvement and things like that definitely pay into that. Okay, cool. So I'll be waiting for those two. And the last question is for folks to watch this, a lot of them watch the recording or the playback during the weekend because not everybody's available at 1:00 on a Friday. And not everybody is in a place in the world where 1:00 on a Friday is their 1:00 on a Friday. So I want folks to have something to really sort of noodle on. What would you like people to think about? Not an answer, not some sort of, you know, pulpit-based epiphany, but what would you want them to really chew on and think about? Over the next few days.

So, you know, since I know a lot of your audience are from the recruiting side and the sourcing side and other parts of HR, I would sort of challenge the audience to think about what are the problems in society that are not related to HR. And I think that if you sort of try to pull it apart a little bit, a lot of the issues that we're sort of hitting in life are related to HR. And in how we're having employees sort of work with each other. In how we're sort of moving to remote environments and what that means. There's a lot of sort of impacts to society based on that and how we pay people, right? There's a lot of problems with money in the world and especially, it moving to a few people. So, I think if you really sort of lean into that, there's so much the power that we have where I think HR feels like we don't have a lot of power, right? The times that I've been sort of like doing people analytics, being in the back of a back office, right? Feeling so out there. But I think we really do have a lot, and I think we can make some changes in very small ways that a lot of executives wouldn't be against, right? So if we went out and sort of said like, hey, we're gonna pay everybody at the company the same amount of money, right? Executives probably aren't gonna be cool with that. There's some hurdles to it, but really trying to sort of run some better analytics, look at stuff a little bit better. Remember that there are humans out there and that this isn't just affecting them at work, but it's affecting their entire lives, right? One of the worst things you could do to somebody is to terminate them. The number of times that people commit suicide after that, right? Or just have a really terrible experience through that. It's more traumatic than a car crash, right? So there's a lot that we have, both really good and really bad. And I would really challenge HR people to sort of say, what are the things that we can change? And I don't think there's a hell of a lot of them. I think we have a lot more power than we really know.

All right. So the question is, think about what are the problems in society? What are the societal problems that are not related to HR? Uh, deep thought. All right, we'll let— we'll leave people with that. And, uh, with that, thanks again to Jem. Thank you, John, for spending your time with us today. Check out the YouTube channel, and we will see you guys all back here next Friday, April 19th, with Chris Rosenau as our guest. So until then, be kind to yourselves. Bye, everybody.