Skip to content
EPISODE 050 : 02/24/2022

Daniel McCarthy


Daniel McCarthy is an Assistant Professor of Marketing at Emory University’s Goizueta School of Business. He specializes in several areas of research pertinent to retail including analyzing consumer behavior and predicting future shopping behavior. Dan’s research has been featured in several major media outlets including Harvard Business Review, Wall Street Journal, the Economist and CNBC.

Host: Ned Hayes and Ashley Coates
Guest: Daniel McCarthy

Listen to every episode

Topics discussed in this episode

  • Importance of customer lifetime valuation
  • Using data to predict customers’ behavior and buying habits  
  • Explaining more about customer-based corporate valuation
  • Understanding customer lifetime value to make a better business decision

Watch Spark Loyalty’s Small Business Success Channel

Play Video

Audio Transcript

Ned Hayes [00:00:01] Welcome to SparkPlug, where we talk to smart people working at the intersection of business and technology. Brought to you by SnowShoe, your smarter loyalty leader. Well, it’s the end of the year and SparkPlug is happy to welcome Dan McCarthy to the podcast. Dan is an assistant professor of marketing at Emory University School of Business. He specializes in several areas of research that are really pertinent to retail. Dan analyzes consumer behavior. He predicts future shopping behavior as well. And Dan’s research has been featured in several major media outlets, including Harvard Business Review, Wall Street Journal, The Economist and CNBC. So welcome to the podcast, Dan.

Dan McCarthy [00:00:41] Hey, thanks so much for having me, Ned and Ashley. 

Ashley Coates [00:00:44] Yeah, so happy to have you here. We’ll start Dan by having you tell us a little bit more about yourself and your career history. Ned mentioned that you’re an assistant professor. You’re also the director and co-founder of Theta and co-founder of Zodiac. So sounds like lots of things going on. 

Dan McCarthy [00:01:00] Yeah, my career is kind of gone all over the place, so I started as a kind of traditional Wall Streeter. I had gone to Wharton for undergrad and went to work at a hedge fund for about six years, and I was kind of somewhere at the end of the fifth year. I’d always wanted to go into academia, but at the end of the fifth year, you can reach this point where it just becomes, practically speaking, harder and harder to be able to make that transition back because you’re just too far in. So it was around that point I started working towards going back for the Ph.D., ended up going back to Wharton for the Ph.D. in statistics. The Zodiac story was that in the third year of the Ph.D., I not only made a pivot into marketing from statistics and primarily working with Wharton Professor, Peter Fader, but we had both made the decision to start a company using some of the models that we’ve been building. So Ned was mentioning predicting customer behavior as basically all that Zodiac did, and that’s all that Theta does now. But essentially, what we would do is we’d work with marketing departments at companies large and small, and we would take in all their transactional and CRM data, run our predictive models on them and use that to come to an understanding of what are these customers going to do in the future or how many purchases will they make, how much will they spend and then rolling it all up into things like customer lifetime value. We were backed by a number of leading VC firms, including First Round Capital and Felicia’s Ventures. We grew it and sold it to Nike in March of 2018. So this was then after I had finished my Ph.D., I had started as a professor of marketing at Emory University, and it was one month after the sale to Nike that we took some of the proceeds to co-found Theta together as a sense by the middle of the Ph.D.. We’ve always had this parallel track where we’re doing a lot of great academic work, publishing a lot of papers, but at the same time doing things in the real world, bringing it to life, interacting with people both on the marketing side. But then also on the investing side, I’d say the big difference between Zodiac and Theta is that whereas Zodiac was primarily predicting customer behavior to help the marketers, Theta is primarily been doing the same thing, but it’s to help the investors. And oftentimes, when we do work with companies directly, it’s to help them when they have their investor cap on. So it could be the Corp Dev M&A division, or it could be the executives as part of some sort of a fundraise. So all these different pieces, it seems somewhat disconnected from each other. But I would say that Theta in some ways kind of ties them all together, that it’s all about using cutting edge statistical models in a somewhat financial application. But what we’re doing is we’re projecting customer behavior, which is of ultimate interest to marketers. 

Ashley Coates [00:03:34] Very cool. And in your position as a professor at Emory, what courses are you currently teaching? 

Dan McCarthy [00:03:39] I currently teach one course and it is literally called customer lifetime valuation. So yes, I had taught a course just the traditional marketing research course for a couple of years, and I wanted to teach this course. And thankfully, the university was very flexible and they, yeah, sure, you can go teach that thing. So I love this. The material is amazing. It’s exactly the same sort of stuff that I’ve been doing for Theta and for my academic research. I think the passion can be infectious. People can just tell this is something that you’re really enthused by all the same stuff. 

Ned Hayes [00:04:11] Yeah. Well, speaking of being really enthused by things, what’s on your radar right now in terms of an area of research that you’re exploring for the future? 

Dan McCarthy [00:04:19] It’s a great question. There’s a few things that I’ve been thinking about. I think one is there’s a lot of people who use these predictive models for customer behavior, but they don’t necessarily know how much they should trust them. And it could be that they’re a young company. They don’t have a whole lot of prior transactional behavior for one reason or another. I think a lot of people use these models, but they really don’t know how much they should trust them. So we have this data set from a company called Earnest Research, which has a really large credit card panel data set. You can basically think of it as being your credit and debit cards observed every month for five to six years, for about two or three million people. That’s a lot of people. And so because they have that breadth of perspective, we can. See what’s going on at a lot of companies. Basically, as long as we can observe through the credit card statement detect string to connect that with a certain company, we can draw some interesting conclusions about how those companies are doing. So they basically gave us all of their data. So this is data for about 3000 companies over that five or six year period. And we can just see when you take all these different models, how well do we predict? And how well can we predict customer adoption? And what can we predict customer repurchasing? And how did these different models that have been proposed in the marketing science literature? Do they actually do well when we really rigorously analyze their performance across thousands of companies? Kind of an open question. So really excited to see kind of how that one turns out, as you may be aware as well. One of my core areas of research is this topic of customer base corporate valuation, and with this data set, we can be able to say what we would expect total revenue within the credit card panel to be. We also observe total sales for the companies from the companies. I think another question would be can we accurately predict future revenue for these companies through the panel? And I think that’s something I strongly suspect the answer is going to be yes, we can do it pretty accurately. But exactly how we go about that, it will be kind of an empirical question. So to say that that’s one of the more interesting recent ones, I’ve been doing more policy work as well. So I have a project right now that’s about the economic impacts of the deployment of e-scooters. So it’s kind of a topic that a lot of people have very different opinions about. Some people love them, some people hate them. And I think the idea here is what we found is that when a city allows e-scooters to operate in a city that people tend to spend more. And so there’s been a lot of attention that’s been generated through this work because as you can imagine, at first, it’s something that no one had studied before. Everyone, when they think about e-scooters, they usually gravitate either towards safety or towards environmental impact. But this is to say, by lowering transit costs, people are going to places and spending there that they wouldn’t have otherwise spent had ahead of scooters not been on the road. So variety, different projects, you have to kind of dove into some of the other ones, but those were two of the ones that are kind of top of mind right now. 

Ashley Coates [00:07:07] Absolutely. Well, we were actually hoping that you would dive in a little bit more to customer based corporate valuation. Can you explain that term a little bit more for our listeners and how that works? 

Dan McCarthy [00:07:18] Yeah. So I think the idea is it’s a simple accounting identity that every dollar of revenue that a company generates is coming from a customer who’s making a purchase. So if we have a good ability to predict future customer acquisitions, how long they’ll stay, how many orders they place while they’re alive, and then how much they spend, then if you just sum up all the activity within each quarter. Well, that’s quarterly revenue. So it’s just exploiting that. And what we’ll do then, is we’ll kind of bring in all these Time-Tested marketing science models for customer adoption and customer purchasing and spend and just leverage those in a finance context which traditionally hasn’t looked at the world. This way, they think about revenue is somewhat top down way. And this is to say, if we think about it in a bottoms up way that we can not only get more accurate revenue projections, but we have all these diagnostic insights about the unit economic health of the company. So there’s all these companies right now, some doing very well, some doing not so well where I think had people been focusing on unit economics and customer lifetime value, it would have been a lot easier to see that Company A had a strong path to profitability. Company B, because unit economics weren’t so solid, they didn’t. And so they’ve just been struggling and persistently unprofitable. I think that’s an area where these sort of models can be especially valuable. 

Ned Hayes [00:08:37] Right. Well our sponsoring company, SnowShoe, works a lot with small retailers. So I’m curious if you could ground your research in the lived reality of a small, independent retailer. What does customer lifetime value mean to someone who’s opened a coffee shop or somebody who is open to specialty gift shop and is just running this as a small operation? How do we translate that to their experience? 

Dan McCarthy [00:09:00] I think there’s two ways in kind of, broadly speaking, I think about customer lifetime value through measurement framework and through a management framework. And the measurement framework is to say, let’s use this to understand how healthy the company is. Think of it like a doctor. You go in, you get your checkup, you get your blood tests, and you want to make sure that you’ve got a good, clean bill of health. I’ll think about CLV in a measurement framework in that same way that we just want to know are you healthy or not? And then there’s the management framework, which is to say, Well, what do we do? Where do we invest our money? And are there some places where we’re earning a lower rate of return than other places which may motivate reallocation budget? And so it’s more about the tactical use cases of how we can do things differently to improve customer value. In terms of measurement, if you’re a small retailer, in some ways, the problem is simpler because you don’t have a whole bunch of different business units that can complicate how you think about the customer, especially if they are working with you across different channels. Just have one product line, and they’re buying through their product line, so you just want to know. All right. How many purchases are all my customers tending to make over the course of their lifecycle with me? How much did I spend to bring them in the door? And when I take into account direct materials, direct labor, shipping, fulfillment, payment, processing and all my other variable costs, how does all that revenue translate into incremental profits? That’s the overarching framework. Thankfully, if you are a small retailer, I assume you have very good understanding of how much revenue you’re bringing in all of the direct expenses associated with that revenue. Hopefully, you should have a pretty good sense for what your customer acquisition cost is. Obviously, for many companies, they don’t have just one acquisition cost. They have costs associated with each of the different acquisition channels that they’re bringing customers in through. So certainly keeping track of that is very important. But I say at that point, the one thing that can differentiate the ability to properly measure is just how good the data recording systems are. So if you don’t have a very good ability to track individual customer behavior over time, say you are that coffee shop and you’re not using Clover or Toast or someone else that allows you to really say, Oh, this is Bob, they’re just some credit card. And because of consumer privacy reasons, you can’t actually tie that to a specific customer account. If you can’t do that, then you can’t really get started. But thankfully, because of improvements in data recording and companies like Toast and Clover, and a variety of other companies that do software within the restaurant or other small retailers, setting the ability to track those customers over time is just getting better and better and better. It’s how I think about the measurement problem. Happy to dive into management as well. 

Ashley Coates [00:11:43] Yeah. Well so I’m curious Dan, this methodology for measuring customer lifetime value. How does this lead to better business decision making? And maybe you give some examples. Do you have any changes you’ve seen in both small and large companies based on this methodology? 

Dan McCarthy [00:12:00] Yeah. The standard first thing that will almost always do is we’ll compute customer lifetime value by acquisition channel across acquisition cohorts. So there’s all the different ways that are bringing the customers in. They each have an associated customer acquisition cost, some bank 30 bucks to bring in customers to Facebook, 20 bucks to bring in customers to Google. I’ve got my Tik-Tok, my Instagram, et cetera, and the 30 and the 20. Those numbers are evolving over time. It could be that customers are getting cheaper to acquire or more expensive to acquire. And so we want to know what is that evolution over time. But then ultimately, if you have the ability to predict what customers can do in the future and convert that into a net profitability figure that allows you to know for each of those channel acquisition cohorts what is my return on investment? So I’m earning one hundred percent rate of return through Facebook. I’m earning a 200 percent rate of return to Google. If I really believe those figures to be true, well, then maybe I should be putting some more money in Google. So it’s that sort of thing that you kind of like. Duh, just makes sense. But again, I think the beauty of customer lifetime value is allows us to get expected return on investments for things that before we weren’t able to. So that’s what a reallocation can be a real direct hit. There’s a variety of others, though. If you were to have a whole bunch of service touchpoint data like you knew for every dish going back to Ned’s example of the coffee shop. If for every customer you knew the barista had prepared the coffee, it could be that some baristas are associated with higher value customers than others. And again, maybe that could be important from a performance management standpoint that someone’s just not preparing good coffee and after people come in, they’re like, Eh, and they just don’t come back again. Well, if you have the data, you could in theory do that, especially if you ran some sort of an experiment where you randomly shuffled your baristas across times a day or days a week. So there’s a whole bunch of use cases that are like that where, again, the North Star is, how can I improve customer value, improve my rate of return? You get the ability to take action based on those use cases is how good is my data recording that I actually have the ability to observe that activity from within my dataset? 

Ned Hayes [00:14:15] Right? Well, the ultimate form of predicting behavior and extrapolating that into the future. I think the person who first posited that as a field was Isaac Asimov with his foundation series. Are you familiar with the storyline at all? 

Dan McCarthy [00:14:29] It’s been a long time. I read it and it was like 20 years ago. 

Ned Hayes [00:14:33] Well, the basic idea is that they have enough data in order to predict human civilization. So you see your data mining and data analysis leading to that kind of predictive outcomes. I mean, maybe not on that scale, but you think you can predict the future to some extent? 

Dan McCarthy [00:14:49] I know we can. Yeah, I’d say that’s one of the beauties of having that parallel track in industry is if I was purely an academic, I don’t know that I could. I mean, it depends on the dataset. But in Zodiac, we had run the numbers on, call it, 250 companies on Theta. We have run the numbers on an additional over 100 companies, so call it 350 in total. So we just been able to observe historically across a really large number of companies. How well can we predict and certainly whenever we do any diligence, we always do it. Would you call like a hold out validation when we leave aside six months or a year, maybe even two years and say, I’m going to train my model on everything before 2019, I’m going to predict the next year or two how well do I predict? And the great thing is these models, they predict very well. So I’ve got confidence, both through the theory and through just bring it to life a bunch of times that we can predict the future very well. I think there is a question when you talk about Isaac Asimov of how far into the future can you predict? And I know at least two years, and it feels like when you look at the patterns of purchasing over time at the cohort level, they tend to be fairly well-behaved, which gives us confidence that there’s no signs that we’re going to go totally off the rails after a year or two. But I would say an indefinite horizon CLV estimate can be more uncertain. So cutting, you know, say at the five year point can be prudent. 

Ashley Coates [00:16:14] Fascinating. Also, if we could go back at just a few years, Dan, back to Zodiac, we’re really fascinated by the story of Zodiac. Can you share more about how you came to found this company and then how it came to be acquired by Nike? 

Dan McCarthy [00:16:28] We were building these predictive models for customer behavior, for the academic research. There was kind of this point where we said, you know, it’d be great to just bring this to life, you know, be great to try this out on a whole bunch of companies. And so Pete myself, I was a Ph.D. student in Pete’s marketing faculty, so it’d be very hard for us to to successfully found the company just given our other commitments. So we were very committed to it. But we also had two other full time co-founders that helped make sure that the operation was running. Obviously, things like recruiting, but then scaling up the data science side of the organization. So there were four of us, they co-founded it, and thankfully because he had been cultivating a lot of great relationships with companies over the 30 year history that he’d been in Wharton, we were able to land a few great customer contracts relatively early on in our lifecycle. And so I think that that allowed us to both get revenue in the door but also start to work towards product market fit a lot more quickly than would otherwise have been possible if we didn’t have a lot of those early customers to iterate. Thankfully, and we were able to get the early backing from those VCs, which allowed us to build out the engineering and data science teams and then the product and customer success teams. And one of the clients was Nike. And basically, they had reached the point where, as you know, they’ve been making this big transition towards establishing direct relationships with their customers instead of selling through the foot lockers of the world. And they wanted it all. Yeah. So essentially they said, well, what actually happened was we had completed our A round negotiations. We had a term sheet on our desk to kick off the next stage of our growth. But Nike had come in with an offer that we couldn’t refuse not to bring in The Godfather analogy. But yeah, it was something that was attractive enough that from a fiduciary duty standpoint, we had to move forward with that. So we ended up getting involved within the Nike family. And the rest is history. 

Ned Hayes [00:18:23] Well, congratulations on that. I am really fascinated by the way that major brands like Nike have gone direct with their customers, as you said. Do you think that’s one of the future changes for retail that larger brands will circumvent the middlemen in the process? 

Dan McCarthy [00:18:37] Certainly, it’s all the rage right now, and it’s not just the the bigger brands. Obviously, there’s a lot of disruptors that are coming in selling direct to nip at the heels, those big companies. I don’t know that it’s for everyone, but I would say that for certain companies, it’s a wonderful way to potentially improve corporate valuation. So in the case of Nike, I think the conditions were particularly good that they had very strong market power. And so they have leverage over the distribution partners that other retailers may not. So they have the ability to sell direct but still have enough influence over the middlemen that that wouldn’t blow up in their face. But as we’ve kind of been doing this more, I think one of the things I’ve come to appreciate a bit more than before is just how hard it is to pull off a transition like this, that there is a lot of people who have become very powerful who are very vested in the old way of the world continuing on. So that’s why I say, you know, it’s not necessarily the easiest thing to do. It could be very painful and hard, but Nike’s a wonderful success case that they not only had the great market position to be able to make this work, but they also successfully were able to kind of change the culture to the point that John Donahoe, their new CEO, he’s a software executive. He’s not a shoe person. And in some ways, it seems like they’re increasingly operating as a gaming company. I mean, in some ways, so begins like do you know how many people need to get fired to have a transition like that actually work. But they’ve been able to make it happen. 

Ashley Coates [00:20:05] Well, so I want to switch over to customer loyalty. This past year, we interviewed Paula Thomas, host of Let’s Talk Loyalty and you were a guest on her podcast recently. In your episode, you touched on the retention smile and wondering if you can break that down for our listeners. 

Dan McCarthy [00:20:21] Yeah, the retention smile. It’s based on this phenomenon that if you were to look at customer activity as a function of tenure, that would you inevitably see early on in a cohorts life is a whole bunch of people drop out early because they tried the product, and for a lot of people, it’s just not going to be for them. It’s not necessarily a knock on the company. It’s just heterogeneity. Different people have different preferences. And so that creates this downward sloping curve early on in the life cycle. But then what happens is oftentimes you see it start to go very flat that the people who continue to stay, they stay for a lot longer than you would have thought. It’s because there is this pocket of people who love the brand and they purchase with great regularity. And so that creates half of the smile. But then I think for leading brands, what can happen is either it’s the network business and they’ve continued to grow. And as a result of that, they get more and more attractive even to people who had previously decided that it wasn’t for them, that they reconsider and or that the people who stayed. They love the company so much that actually for them, they’re purchasing more and more over time. So, there’s just a variety of different forces that can actually push that curve back upwards again. It’s oftentimes what can happen is when you are running your models first. A lot of the standard models don’t allow for that smile. It’s more of a marketing science statistical thing, and if you don’t, that can obviously be a big problem. But even for those that do, if you haven’t observed enough transactional data and you haven’t seen anyone smile yet, well, it’s kind of hard to know, are we going to see that happen or not? Because again, for a lot of companies, they don’t see the smile. Just kind of goes flat. It’s kind of the end of the story. So it’s a really important thing to look for, but it’s also very hard. It’s something to be mindful of. When you’re looking at really young companies, it is an open question mark. 

Ned Hayes [00:22:10] Right? Well, you also talk about marketing science models to measure repeat customer value. And I’m curious how that works more that you’re actually marketing science models. 

Dan McCarthy [00:22:22] There’s been a lot of literature on how we can take historical transaction behavior and use it to predict future transaction behavior. Yes, so definitely happy to talk to those models. But this canonical problem within the marketing science literature that we’ve been able to make a lot of progress on over the past 20 years. I’d say the workhorse model actually was conceived in 1987, so quite a long time ago. But then it really didn’t get any traction, and it wasn’t until somewhat recently that Pete Fader resurrected the model. He found a way to estimate it a lot more efficiently. And since then, there’s this family of so-called Buy Til You Die models that have grown and flourished and now are being used by many, many, many companies. And in some ways, they’re the foundation of what we used at Zodiac and what we currently use at data. So our models are greatly enhanced versions of the plain vanilla models that are there in the literature. But at the end of the day, the story that the models will tell is, different companies have customers that have different levels of loyalty, and while customers were alive, they have different purchase frequencies. Some buy a lot and some don’t buy very much. And so we’re going to do is we’re going to take all of the historical purchase behavior, and we’re going to use that to infer how loyal or not or how frequent people buy when they’re with the company. We’re going to infer all of that from that behavior and use that to make predictions of what they’ll do in the future. 

Ned Hayes [00:23:45] Yeah. Well, speaking of predicting the future, I’m curious if you could look out five to 10 years. What does the future look like? 

Dan McCarthy [00:23:52] It’s a hard question. 

Ned Hayes [00:23:54] Well especially we’re interested in the future of retail, do you have any specific areas that you think retail will keep growing in or will change? 

Dan McCarthy [00:24:03] Well, for one, I think within retail again, I mentioned those companies like Toast and other companies that provide a lot of end customer visibility in areas that had previously not had that sort of customer visibility. I think that’s going to make it a lot easier for those companies to be customer centric and actually run the CLV models and be able to be better versions of themselves over time by leveraging those insights. So especially in areas where the historical so-called capture rate or the proportion of transactions that they can actually trace back to customer identifiers in areas where the capture rates have been low, I think we are going to see software make that a lot better. I’m kind of interested to see how privacy is changing the landscape. So, yeah, everyone’s been talking about the Apple changes, and we are hearing more and more companies talking about how it’s hurting them. So companies like Poshmark, but a variety of others they’ve been saying that they just haven’t been able to understand exactly what the customer journey is to the same degree that they did, because it’s as if someone hit the delete button on some of the customer ID coming in, especially through mobile. That, to me, is an open question. I think it may get harder to be quite as efficient in terms of allocating marketing budget. So I think that that’s going to motivate other solutions, but then other methodologies to be able to take incomplete data and be able to try and make the most with what we have. So I think those are going to be two themes that I would expect to become more relevant than they are right now. 

Ashley Coates [00:25:32] We’re very curious to see what happens. Well, we have one more question for you, Dan, which is what is your legacy and what would you like to be remembered for? 

Dan McCarthy [00:25:40] Well, I feel like I’m too young to think about things like that yet. I’m just focused, honestly. The main things that I’ve been focusing on for one is, through things like my CLV course and just the other flag-waving that I’ve been doing, getting people excited about this way of looking at the world and teaching the next generation of leaders about these concepts and doing well to the students that I’ve been working with. There’s a lot of people doing great work right now, so to be able to help and support the work that they’ve been doing. I think that’s a big part of what I hope to do. And then obviously, when it comes to the flag bearing Theta you, I think that that’s the other big vector that I’m doing it through. Yes, so through Theta, it’s really companies they want to be able to know more about this and they want to do it. And so to actually be able to help them do it is something I think it’s both personally rewarding, but also just really helps with that broader mandate. Yeah, I’d say that there’s a lot more to me than just CLV, but I do feel like when it comes to really getting more people using it, the combination of classroom, teacher, Theta and then probably social media. Yeah, I hope that those are the big ways that I’ll be able to continue to get the word out there, get more people excited about and using this sort of stuff. 

Ned Hayes [00:26:51] Great, well, thank you for all your insights today, Dan. Appreciate your time. 

Dan McCarthy [00:26:55] Well thanks for thinking of me and great answering all of your great questions. And here’s to a great ’22 ahead. I’m very excited for what the new year will bring. 

Ned Hayes [00:27:03] SparkPlug is a wholly owned property of SnowShoe All Content and copyright 2021 SparkPlug Media.