Ben Zweig

Clara (0:00): Welcome to Conversations in Data by Tech Accelerator, the podcast where we talk to the builders, the founders, and the data obsessives for crafting how the world really understands information. I'm Clara and today I'm hosting Ben Zweig, the CEO and co-founder of Revelio Labs, a workforce intelligence company that's building something really ambitious, a universal HR database from public data alone. Ben, great to have you and welcome.

Ben (0:27): Thanks so much. Yeah, thanks for having me. Happy to be here.

Clara (0:30): Absolutely. Thank you. Thank you so much for joining us today. We'll get started with, you know, a little bit about your original story. You have a PhD in economics. You were doing workforce analytics at IBM. You were at a hedge fund and then you went and started a company. What was the insight that made you think this has to exist and I am going to be the one building it?

Ben (0:54): Yeah. Okay. So, it sounds like a weird background, but I think it all kind of comes together. I spent a bunch of years being really obsessed with finance and there's this book called Capital Ideas by Peter Bernstein. Anyway, it's about the birth of finance as a science. So like in the 50s and early 60s, the field of finance was very non-sophisticated. There was no data really, it was just a bunch of ad hoc thrifts doing lending based on nothing. It wasn't really rigorous, it wasn't scientific. He's a financial historian, also an economist, but cataloged this history of what happened which led to the world of financial sophistication and rigor that we see today. And I remember just thinking like, wow, these are larger than life characters. There's Harry Markowitz and Merton Miller and Mike Bloomberg and all these characters that are part of that story and it just was so cool. And then eventually I went to IBM in the corporate world. The first day I started, my manager at the time kind of introduced what we were doing and he said, "Employees are two-thirds of our expenses at IBM and we don't have any clue how to manage them. We are completely in the dark." And coming from finance I was like, you know obviously I know that labor markets are twice as big as capital markets. That's two-thirds of the economy and capital is one-third of the economy. But I was just like, why are they so behind? Why is it so backward? And you know, I was like, someday someone's going to write a book like Capital Ideas, but it's going to be about how labor markets became scientific and rigorous and sophisticated. And I was like, I think I want to be one of the characters in that book. So that was really the birth of the idea for me. And then of course, how that led to unified data and scraping public sources and whatever, like that is kind of downstream of this original spark of wanting to advance the science.

Clara (2:54): That is extremely interesting and very random. It was something related to finance and then you got into the labor part of it and yes it is so much... it is so meaningful. And now talking a little bit more about what Revelio does. Can you explain the core product to me? You're calling it a universal HR database but you know every company's HR is in a locked box pretty much. So you've said it before. How do you actually reconstruct that from public sources?

Ben (3:23): Yeah. Okay. So, I like the way you're framing it as a reconstruction because that is essentially what it is. Like you said, every company has their own HR data, but they would never ever share that data. But a lot of the information that's in an HR database is also mirrored in the public domain. So on someone's resume or LinkedIn profile or whatever, they have their job title, their company name, their start date, their end date, the description of what they do, the skills they have. There's so much information there that is just very rich information. And the nice thing about public sources is that you can analyze a company that you have no affiliation to. So, when I was at IBM, we would never in a million years dream that we could see the HR database at Oracle... like they wouldn't share that. But if we could at least approximate it and at least get hiring rates by roles or attrition rates by roles or skills or seniority, whatever we want to do, then that can get us pretty close. And when we started we didn't really know how close we could get. Like, could we actually measure the attrition rate of some role with precision? And it was really hard. It still is really hard. But I think we've got a lot more validation now that that works actually really nicely. Oh, yeah. It's been a lot of work to get there because I think another challenge that I didn't really address but is kind of the core challenge is actually kind of standardizing classifications of employees between organizations. So they use different titles. They have different levels of seniority. They have different classifications for their business units or their geos or whatever it is. So being able to kind of unify that requires a lot of universal taxonomies.

Clara (5:14): Yeah, that was going to be my follow-up question to you about the primary data sources, but now that you said about LinkedIn and job posts... I do this every day and I realize there's just so much information out there in those, such as salary or location like, "Oh, can you commute to Liverpool Street in London three times a week?" And of course online you can find so much more information about businesses than you could before. Like right now if you want to look up a company in Europe I think that you can even see their revenue and that's just public information. And in the United States and everywhere, but like exact numbers, to the cents you know. So that's very very different from what HR really is because behind companies' closed doors there's still the exact salaries, the exact titles and how people change, but then people go and put that on LinkedIn. So it's very very interesting.

Ben (6:12): It's really wild. Yeah. I mean, even now, and we've been doing this for like seven and a half years or so, and even now, we're still discovering new data sets that are out there that we just didn't know about.

Clara (6:24): Yeah, that's really fun.

Ben (6:26): Yeah. And just trying to think of an example... we found out that in the US there's these published layoff notices. They're called WARN notices and that we've been collecting for a while. But we found out that in Europe you can actually get notices of expansions, of planned expansions. We were like, "Oh, that's cool. Like, who knew?" You know, and added it to the feed. And we just found out that in China when municipalities are hiring public sector employees, before they hire them, they have to list out all their characteristics, their demographics, their experience. They publish it on their website so that the local population has a chance to contest it or whatever. And we're just like, that's out there? Like, this is crazy.

Clara (7:12): That is insane. Yeah, that is insane. You know, putting out the demographics, that is so problematic. Oh my goodness.

Ben (7:17): Yeah. Yeah. I mean, you know, sure, but from as data people, I'm like, "All right, if it's out there, it's out there. Like, I'm going to collect it. I'll take it."

Clara (7:26): I'll take it. But I just love random information and random data. Maybe that's why I'm here at Tech Accelerator and Data Sharp. But yeah, makes sense. But at Accelerator, you know, we talk to a lot of data buyers and one thing that we hear constantly is that raw data is actually the easy part, right? It's the cleaning, monetizing, normalizing, enriching. That's a lot of work. What do you think about the data pipeline at Revelio?

Ben (7:50): Yeah, you know, it's funny... like when you start selling data to hedge funds, you get a lot of advice that's sometimes good advice and sometimes bad advice. I think one thing we heard early on that turned out to be not good advice that thankfully we disregarded was that the sophisticated hedge funds just want the raw data and don't do any of the enrichment. They're going to want their own pipelines. They're going to want to fiddle with it and construct it any way they want. And we didn't think that was right. First of all, because in this vision we were trying to build toward where we wanted to unify and centralize and standardize employment data for the world, for HR, for anyone analyzing employees... we knew that hedge funds weren't the final end buyer. But even among hedge funds, I think it's just not true that they want to standardize themselves. Return information or finding correlations between securities or constructing factors, like that's their thing, they can do that. But if it's something like constructing a taxonomy of occupations, that is our thing. Like there's no way that Two Sigma is going to be better at that than we are. You know, we are in the business of curating employment related data. So for employment specific metrics, like even inferring salary or creating seniority levels or creating skills or work activities, that is stuff that we need to do. And the data itself, it's just raw free text. So there's so much that we have to do to get that in a nice neat tabular data set that can be embedded into a traditional analytics workflow. So we have a pretty intensive data pipeline, but it's really at its core about making it nice and neat for the end user. It's just about solving problems that exist in the raw data, and we have solved more of those problems that are probably useful to everybody. So, I think it does kind of make sense to think of... I don't know, for whatever reason, I'm thinking of like an oil pipeline. You know, you could have crude oil, but it's not actually useful for anything. And then you need this enrichment layer before you actually do anything with it.

Clara (10:13): You've talked about the vision behind Bloomberg Terminal for labor market data. Bloomberg became Bloomberg because it had data nobody else had. Delivered in a way that people depended on. What's Revelio's thing?

Ben (10:27): Yeah, it's funny. I don't know that I necessarily agree that Bloomberg had data that no one else had. I mean, in terms of the raw data, it's out there. It's in EDGAR filings, like FactSet has it, Refinitiv has it. It's out there and it's publicly filed data. Now, I'm sure they've moved into some proprietary stuff, like now they get trade flows and stuff, but when they started, I think it was really just that they were digitizing data that wasn't easy to access. And I feel like we're kind of doing the same thing. Like there's data out there that anyone can buy raw data. It's just so messy. It's so complicated. So I think there are so many problems to solve. Like we are on version god knows what of company mapping and I think there are others who are trying to enrich it and maybe it won't always stay this way, but for now we're ahead. We've cleaned and curated it in ways that others have not and we are able to uncover insights that are unique. So I think we feel like there's a technical advantage. And you know if others do that great, then we'll solve the next problem. And I don't really see a world where we kind of run out of problems to solve.

Clara (11:49): That makes sense. I wanted to chat a little bit more about the workforce data space. It's getting a little crowded. You know LinkedIn like we were saying has talent insights. There's Lightcast, Glassdoor. Someone could argue that you're all fishing from the same pond. What's the honest answer to why Revelio over others?

Ben (12:13): Yeah, it's funny like when I first started this I thought, "All right, there's two problems. One is: can this be done? And another is: is it defensible?" And I thought, "I'm very concerned on how defensible it is, but I'm sure it can be done." Now that's sort of flipped and now I'm really not concerned about how defensible it is, and I'm much more concerned about whether it's even possible. So when we first started we had a different set of competitors. I mean you mentioned Burning Glass, that's not around anymore. There was Thinknum, there was Cognism, like those were our competitors. None of them are around anymore. LinkedIn Talent Insights... LinkedIn shut down Talent Insights. They let go of their team, like that's shutting down.

Clara (12:53): How are you staying? If everyone's leaving, can you tell me how you stay?

Ben (13:00): It's a war of attrition, you know? It's who's the last one standing, you know? I think we've been lucky enough to kind of stay standing and keep growing. And I think we made a choice that was kind of unique in the space to focus on the company as a unit of analysis. Whereas companies like Lightcast, TalentNeuron, they have focused more on the market as a unit of analysis, you know, labor market insights. And us, we don't really use the term labor market intelligence. We use workforce intelligence because we feel like you could have a workforce of a company, but you can't really have the labor market of a company, like that's not a coherent idea. So I think we kind of started with a little bit of the harder problem and I don't know... again, I'm biased so take with a grain of salt, but I actually don't think it's a very competitive space. I actually feel like there's more legacy software out there that clients aren't terribly thrilled with. And I feel like it's a little bit wide open. I mean, I honestly hope there will be more entrants into this market because I do feel like there's a lot to do.

Clara (14:17): You want competitors.

Ben (14:17): Yeah. No, I think so. I think it's actually easier. I mean, as an example, we sell into the US and Europe. We also sell into Latin America. And you would know better, but in Latin America, there's nothing like this. And it's exciting because we're introducing this idea to a set of markets that really don't have anything even remotely similar. But it's harder to do that. Like I wish that we weren't the first ones. I wish we were the second, third, fourth and people kind of got it. I think it's not always good to be the early adopter, but whatever. There's pros and cons to that.

Clara (14:55): Yeah. No, a very interesting intake on, you know, when you have competitors, it's easier to explain the product. It's easier to sell the product. You can have a different pitch, a better pitch, and then you sell it. So, it's like really teaching people how to use it. But now jumping into, you know, who's really buying the products and why. So, you started it with hedge funds, which is a fascinating go-to-market move. Tell me about that. Why did investors want workforce data before HR teams did?

Ben (15:27): Yeah, I mean it's interesting... right now our biggest market is corporate HR even though our earliest market was investment management. I think the advantage to starting in investment management is that the sales process is very easy and the engineering process is really really hard. Like here's kind of what I mean by that. Hedge funds are really very particular on data quality. They hold data quality to an extremely high standard, as they should.

Clara (16:04): As they should. Yeah.

Ben (16:05): Well, I mean I think everyone should, but they do it the most. But the actual sales process... you meet someone at a conference, you tell them the data, they ask some questions like, "Oh, does it have a lot of history? Does it have broad coverage? You know, what's the frequency?" and they check some boxes, and if it checks their boxes they'll say great, we'll trial it. The deal kind of closes itself, that's all you need to do. So you don't really need a sales team in the space.

Clara (16:32): I wanted to ask you, so how many sales people do you have and how many data engineers do you have?

Ben (16:37): Yeah I would say like two years ago we had one salesperson and like 50 people who were writing code. Now we've got 12 salespeople.

Clara (16:47): Mm-hmm.

Ben (16:49): And like maybe 65 people who write code. So, we're getting more balanced, I guess. I mean, but really because in the corporate HR market, it's very different. They're not as particular on data quality, but you really need to know what they're working on. You need to know how they want to buy it. Like there's a lot of solutions development, I would say. So, that is more intensive on the sales account management side. And on the data quality side, I feel like we don't really have that much to do because we already did it.

Clara (17:30): That's a better ratio, right? 12 to 60 than 1 to 40 or 50. So, I'm sure that they're more in synergy now that there are these better ratios. But thanks for sharing with me now.

Ben (17:42): Yeah. No, that's what we do. Yeah. Yeah.

Clara (17:47): Okay, cool. You mentioned that Revelio has produced its own version of jobs data and that it came out before BLS and moved markets when BLS itself was in turmoil. That's pretty extraordinary. Can you walk us through that?

Ben (18:04): Yeah, it was a little wild. Basically, yeah, I forget the month... I want to say October maybe, but there was one jobs Friday. There was a very unflattering jobs report. There were big downward revisions and the president fired the commissioner of the Bureau of Labor Statistics, Erica Groshen. So as soon as she got fired, there was a lot of concern that the BLS would become politicized and we wouldn't be able to trust the jobs numbers and they would be cast into doubt or maybe provide some incentive like, "Hey, if the numbers are bad you're fired." So it got the whole economics community very scared, and that was on a Friday. And then on that Sunday I called our chief economist Lisa Simon and I was like, "You know, I think we have to do our own jobs Friday release," and she was like, "I was thinking the same thing." And then we called our chief data scientist and we're like, "We gotta do this," and he was like, "That's what I was thinking." Like we were all thinking the same thing, it just had to happen. It really felt like we had no choice because we knew we could do it. We've been providing micro-level data for so long and aggregating to macro statistics shouldn't be that hard. Turns out it was a lot harder than we thought, but we pulled it off. We had to work that month to get data that would represent the macro statistics.

Clara (19:34): Yeah, I really loved that. Sorry to interrupt you, but I really love that you said that, you know, on Friday this happened and on Sunday you called your colleagues, team members, and then everyone was thinking the same thing. I feel like in an organization that has to be the goal. Everyone being on the same page and constructing and building things together. It's fascinating to me.

Ben (19:57): Yeah, it really felt like a nice moment. I mean it was a very intense month. Like it only cost us our sanity to some degree, but it felt like we were all driving toward the same thing. And actually the timing really couldn't have been better for us because the next month the government shut down and it was the longest government shutdown in history and there was no jobs data from the BLS. So we were kind of the only game in town and so you know, we didn't sell this. This was something we just put out as a public service, but we got a lot of attention from it. And actually, I mean, this isn't something we formally announced yet, but you know, hear it here first, that we're planning on expanding that to more geographies. The next big goal is we want to globalize Jobs Friday. We want to expand to more countries so you can make comparisons between countries and then if there's some policy change then you can analyze the effect of that policy change because you have a comparison group. So it just seems like a much more useful way to get jobs data out there and we have enough validation that it works really well.

Clara (21:03): Congrats on the upcoming release of, well, not yet but you know thank you for the advance. Yes, but you know we hire a lot of offshore talent at all the groups. The group is Data Sharp and then we have companies under the group. We hire offshore talent for pretty much every single company.

Ben (21:30): That's interesting. I think what you're describing is very similar to what talent intelligence organizations do within companies, and that's such a big booming subset of HR because they're so forward-thinking on using external data for their talent acquisition strategy.

Clara (21:44): I wanted to ask you about people analytics. You know, it has a reputation in HR circles as a "nice to have" that pay lip service to, doesn't really fund. Do you think that's changing or are most HR departments still not ready to treat their workforce like a data problem?

Ben (22:08): There are things that I really love about the people analytics space and the thing that I love about it is the people. Like the people who work in people analytics are so ambitious and forward thinking and really interesting. The thing that I don't like about it is that... I don't think you could really blame them, but what ends up happening with people analytics teams is that they are essentially the owners of all talent management data. So if you think of HR—forget compliance HR, like analytical HR—you could split that into talent management and talent acquisition. People analytics is really focused on the talent management side. It's like once employees are already there, you know, how's engagement and attrition and promotion and all that. So they kind of own the talent management data and because they are the keepers of that data, they get asked for a lot of ad hoc reports. And they're trying to move into more self-reported things. They're really trying to escape that trap of just being bombarded with ad hoc queries. But I think what that has led to is that they are reticent to take on new data sets because then they think, "If I take on this other data set, then I'm going to be this expert in this other data set and I'm going to get even more random questions." So I think they're in a tough spot. I'm sympathetic to where they are. Now in contrast I think talent intelligence is... you could think of it as the analog. Like people analytics is to talent management as talent intelligence is to talent acquisition. They are the analytical engine that drives TA for the most part. I mean that's maybe a little bit of a simplistic way to think about it but I think it's true enough. But I also think that the borders between these groups are becoming less important. So, I don't know. I think every company faces different identity crises about who does what, but I'm hopeful that they're going to produce more impactful work, which will make them be seen as more valuable and eventually get more resources, etc., and then there's going to be this virtuous cycle.

Clara (24:16): Yeah. And I think that most companies and people analytics teams, HR teams, they rely a lot on internal data. Let's suppose a company that's been around for 30 years. They have, you know, they're using their ATS. It has integrated the data from all past employees, roles, and things like that. But that's internal data. What about the external data? What about the comparison? What's going on out there? Are people being paid what they should be paid? What's the salary benchmarking being done? What about these roles? Is this like a relevant role still? Internally you can look at that and not make such an informed decision as you would if you had the external part. So that makes sense and that's another interesting point here.

Ben (24:58): Yeah, I think that's exactly right. I mean Michael Porter, who's considered the father of modern strategy, has said that strategy... well, strategy at its core is about differentiation. Full stop. Like, what is strategy standing on one foot? It's differentiation, and that requires benchmarking, which requires standardization, like all these kind of hard things. So I think without considering what's happening with competitors, how we think about differentiation... you sort of are condemned to be non-strategic almost by definition. So that's a big issue. I think there's also an old school mindset that if you are owning a certain data set, you need to have exact numbers. And so even getting at estimates, anything like crowdsourced, anything sampled is seen with some skepticism. I think that is being seen increasingly as too limiting. And it's also even more dangerous because what you had mentioned, what people in HR are analyzing from their ATS, their HRIS, they see that as the internal data. But it's worse than that. That's actually only a portion of their internal data because usually that only includes their W2 employees and it doesn't include their contingent workers, their contractors. And as you know someone who hires internationally, sometimes people need to be classified as contractors basically just as a legal loophole.

Clara (26:28): Yes.

Ben (26:29): But for all intents and purposes, they're part of the team. They're part of the workforce and they're not on the ATS and they're not in the HRIS. They are in some procurement database or something. And that is almost always ignored. So, if you want to analyze the total workforce, which is almost always the most relevant thing to analyze, people on all these teams can't really do it or don't really do it, which is a problem.

Clara (26:59): But, um, I wanted to ask you a little bit more about, you know, bets and things that you're doing in the future. So, what is your big bet for the next three years in terms of where workforce intelligence goes and what Revelio's role in that is?

Ben (27:18): I don't know that there's like a singular big bet, but here are a few things that I'd say we're thinking about and working on. One is, like I mentioned, globalizing Jobs Friday. That's a big thing for us. Another is doing more scenario planning. So, doing a little bit more causal analysis of like if you do a certain thing, if you make a certain investment, what are the downstream effects of that? So that you can analyze, is this a good decision? What are the results from a certain decision? And that can be done for public sectors, like if I make a $10 million investment in some industry, what does that lead to? Or even for a company, if I mandate that everyone returns to office, what's going to happen? So I think scenario planning is really interesting and I think we are always interested in getting more internal data. So unifying internal and external data is always very exciting for us. We're starting to do that but I think that probably means we are going to need to have an analytics platform that unifies internal and external data so that you can see everything you'd want for an organization. There's another thing that we're thinking about which is actually getting into the job board/career planning space and providing this information to individuals. Because we have so much data on like if you take this job what does it lead to and how much do people get paid and there's so much information that I think would be useful for individual job seekers or employees thinking about their next move. It just seems like there's so much value that we're just sitting on. So we're starting to experiment with how to make more consumer applications.

Clara (29:08): It just made me think about how role progression is so rigid, you know, when people don't think of the other ways that a role can lead to another or what could the other roles be. So that turned a light bulb in my head right away. But thank you for sharing with me a little bit more about what you're planning to do in the next few months and years.

Ben (29:35): Yeah. Don't hold me to it. Yeah. I'll keep you posted.

Clara (29:40): And last question for you. If you could fix one thing about how companies use data today, whether that's workforce data, market data, any kind, what would it be?

Ben (29:51): One way that I think about jobs data in particular is that if we are thinking about the transformation of a workforce, sometimes that happens between jobs... sometimes you can move jobs, expand a number of jobs, whatever. But more often that happens within jobs. So very often the job itself transforms. So you're a perfect example. You got hired to do HR stuff and now you're hosting a podcast. You are doing something that was not in the job description when you started, but you had this... it's like what you said, sometimes a business needs something and everyone knows that that is the direction to move in and you adapt. And I think that kind of leads me to a couple conclusions. One is that organizations really need to have an understanding of the tasks and activities of employees rather than just the occupation. Because an occupation is really a shorthand for a bundle of activities.

Clara (30:52): Oh yeah.

Ben (30:53): And so once they do that, then I think we need to think about management. So here's a take that I'll put out there which is also another "standing on one foot" thing. I think management at its core, at its essence, is about job reconfiguration. So here's what I mean by that. If you are a manager, let's say a middle manager, there's some organizational objectives that are fluid that change all the time.

Clara (31:22): Mm-hmm.

Ben (31:24): And there's some set of employees that are also fluid. So there's business things that change the requirements of what your team needs to be doing. And then there's changes on the team. Let's say someone leaves, someone's got to pick up their work, or you realize someone's actually really good at something you didn't think they were good at and you want them to do more of something else. You kind of want to manage that team.

Clara (31:43): That is a very interesting point and I really love that you said that. Like sometimes a company will come up with a project and then you know, who's going to take over, and roles change all the time, new tasks get implemented all the time. And yeah there's a lot to be documented and put in process that sometimes isn't, right? And then it gets to the end of the year and then you're like, "Oh that wasn't documented, there was no data behind that and then now I'm going to have to pick up from where I left." Do you have anything you want to say to complement our talk today?

Ben (32:20): Yeah. No, thank you so much. I mean this was so fun. Thanks for having me. I mean, yeah, I would say to any listeners like stay in touch. You know, I post a lot on LinkedIn and stuff like that. All this stuff, it's changing quickly and it's always interesting. So, you know, I think this space is interesting. So, watch this space. That's my advice.

Clara (32:39): So do I. Yeah. Know, I agree. Revelio Labs is something important and I'm excited to see where it goes. So, yeah, we'll definitely keep you there on our socials and everyone should do the same. And thank you so much again. It was a total pleasure and thank you for everyone listening. You'll find much more about this episode on our LinkedIn, YouTube, Spotify.