Understanding Your True LTV and Company Value | Prof. Daniel McCarthy

Speaker A: For most D2C brands, 40 to 80% of their customers are going to buy one time and they're never going to come back.

Speaker B: Delighted to be joined from across the pond, Daniel McCarthy, founder of Theta, uh, associate professor of marketing at the University of Maryland.

Speaker A: Good customers are born and not made. Trying to do a whole bunch of magic to try to make the bad customers into good customers is not as good of a proposition. What CLV should be able to tell you is does this company have a really good path to profitability or not? And if you're not able to get that from clv, then your definition of CLV is not correct. We'll often see, you know, people talking about ltv and it's wrong on so many levels.

Speaker B: What are some of those key metrics they should be reporting into? Like maybe the next board or leadership meeting?

Speaker A: I think it's a very helpful exercise to ask.

Speaker B: Welcome back to another episode of D2C Diaries. Delighted, um, to be joined today online again, um, from across the pond by Daniel, uh, McCarthy, ah, founder of Theta and associate professor of marketing at the University of Maryland. Definitely the most qualified person we've had on the pod. So very excited to be, um, to be talking today. Um, thanks for coming on, Daniel.

Speaker A: Hey, great to be here.

Speaker B: Yeah. Um, I think I first became aware of yourself and the work that you guys are doing, ah, at Ah, the 40, uh, events that we attended in January in New York City and uh, you presented on that day, uh, around some of the topics we're going to dive into over the next sort of hours, 45 minutes to an hour. Um, I think we've probably recorded nearly 100 episodes of the podcast to date and we've pretty much focused every single one of those on how to acquire customers. We've not probably given enough air time to everything else that's involved in running a successful E comm brand. So really excited to kind of dive into that. I guess it'd be great just to start as with every guest, just with a little bit of a brief introduction about yourself, your work, uh, some of the stuff you're doing at both the university and theatre currently.

Speaker A: Yep, yes, I kind of wear a couple different caps. The first is as an academic. So I've been a professor, uh, for a number of years now, um, since 2017. And um, I've been focusing on this problem of how can we predict what customers will do in the future and how can that inform an understanding of how much businesses are worth, um, probably a lot of people in the audience, you've heard about customer lifetime value and uh, a really big question is um, what do finance people think about that? How can we look at that through the lens of the CFO? Um, my PhD dissertation was actually all about that topic. Uh, I'd spent a handful of years on the buy side before coming back for a PhD. Uh, even though I'm a marketer formerly, my PhD is in statistics. So not your typical marketer. Um, the other thing that I've been doing in addition to um, kind of teaching about customer lifetime value and doing research on clv, uh and how it kind of rolls up into corporate valuation is as an entrepreneur. Yeah, so I've started a few businesses. Pete uh, Fader was uh, my advisor in all but named um back when I was at Wharton, uh getting my PhD. Uh, we co founded a company called Zodiac back in uh, I think it was 2012 and um, yeah predictive analytics software. As a service firm we primarily help marketers do tactical customer acquisition and retention. Um, again doing the same thing that I was doing for my research really which is let's get those hyper accurate models for what the customers are going to do. And um, we grew it, we sold it to Nike in uh, March of 2018, took some of the proceeds and uh, started Theta. We were able to work into our non compete with Nike that we could use very similar or even the same models as long as uh, we primarily just worked with private equity firms pursuing customer based corporate valuation. So um, as long as we're not doing work for firms like Puma and Adidas, um, and just kind of helping PE firms kind of kick the tires on uh, the brands that they're potentially evaluating for acquisition. Um, I think that was kind of far enough from what they were doing that they didn't perceive it to be a threat which it really wasn't. So we've been doing that ever since. Uh, so yeah, now Theta's been around for quite a while. We've probably run the numbers as part of paid engagements on over 450 distinct companies and really kind of runs the gamut from uh, largest telecom firms in the world, some of the biggest quick service restaurant firms in the world like McDonald's, um, pharmaceutical companies and then obviously tons and tons of direct to consumer brands. So it's certainly direct hit for the sort of things that we'll be talking about here. Yeah, so that's uh, I kind of been living it both as an academic and as a practitioner which hopefully should you know, give um, pretty kind of unique differentiated kind of perspective On a lot of these issues.

Speaker B: Yeah, super interesting.

Speaker C: Quick one before we get back into the episode. Every business spends money to make money. Ads, uh, software, flights, travel, the hundred unavoidable little things to keep the lights on. All that spending every month and none of it comes back to you. It just goes out the door. In Card is a financial platform built for digital businesses and it flips, uh, that on its head. They're giving 2% cash back for your first year on ads, SaaS, travel and everyday business spend. And it's all uncapped. The same money you were spending anyway, except now a slice of it lands back in your account. You also get multi currency accounts, corporate cards and Shopify, Meta, Stripe and Xero connected in one place. A true source of truth. But honestly, you're spending regardless, so you may as well earn on it. Sign up for free. Use the code SWU2K26 links in the description. Now back to the episode.

Speaker B: I really want to dive into the sort of service and the value that you're providing to PE firms and businesses exploring acquisitions through Theta and how, um, maybe that can encourage different ways of thinking about customer base and uh, tying that to acquisition and retention strategies. Beyond that, what's like the typical questions that businesses are coming to you to answer, uh, when they're engaging with you guys?

Speaker A: Yeah, if we're working with the growth private equity firm and they're evaluating some consumer brand, uh, usually they have their own internal way of thinking about valuation. So there are some firms that kind of come to us to go all the way to valuation. Uh, but a number of the firms, they're positive on the company and what they want to do is they want to make sure that there's no skeletons in the closet that could be gleaned from the transaction data. And so typically at the early stages, there's not as much data, there's a lot more other firms. And so, um, the rich data is typically not made available in the data room at that point. But when the circle gets smaller and you're kind of closer to the final stages, then, um, usually firms will spend more, PE firms will spend more on diligence, and the potential targets will put a lot more data into the data room. So we'll come in on behalf of the firms, the PE firms, and just uh, help them understand what is this telling us? Is there anything that we need to worry about? And um, I'd say that that's kind of like the big thing that uh, PE firms would come to us most often to do.

Speaker B: Makes sense. And I know from reading some of the material that you shared previously, you tend to do that through the lens of a growth report card. When looking at the health and quality of a business. For anyone who's listening and isn't privy to some of the frameworks used within that process, would you be able to walk through what's some of those main metrics are and how you approach that?

Speaker A: Yeah, the big thing is really, uh, is the company creating value through growth or not? And, uh, you know, oftentimes it can be a little bit hard to parse it all out if you don't have access to the transaction data, because you can have companies that are growing really fast. And, uh, oftentimes, you know, for series A, series B, series C, um, you know, if you're growing really fast, it's enough of an indication of product market fit that you can kind of make it to the next round. But, you know, a big question is, are those customers who are, you know, signing up for the first time, are they repeating and are they kind of coming back over and over and over again? Um, or not. And if you're throwing enough marketing into customer acquisition, oftentimes you can kind of create or kind of manufacture a lot of growth, but it won't stick. And you're kind of stuck on what I would often call the customer acquisition treadmill, that you just have to keep on throwing more and more and more money into acquisition just to kind of keep growth up. And actually, the tricky part is if you're growing revenue quickly enough, you can give the illusion that you have more product market fit than you do. Um, because oftentimes your overhead is not growing to the same extent. And so even if your repeat's kind of bad, you still might be more profitable or less unprofitable over time because you're flexing the overhead. Um, but actually, if you kind of look at CLV contribution profitability by cohort, what you'd see is that there's not a whole lot of durability there. So, yes, I think done right. If you're a, uh, PE investor and you're kind of looking at these numbers, what CLV should be able to tell you is does this company have a really good path to profitability or not? And if you're not able to get that from clv, then your definition of CLV is not correct.

Speaker B: Makes sense. That is interesting. The point on those businesses that may be playing more of an arbitrage play on first order, but then have, like you said, limited fixed Cost, I'm assuming you've probably seen more of that now with the rise of AI and how much more individual people are able to do and how that allows brands to keep that ah, cost base down relative to maybe a few years ago and achieve these ridiculous sort of growth, growth numbers. I know in the deck you shared with myself that you approach like the value equation for consumers very much like bottom up rather than top down. Um, um, so top down being more homogenous and bottom up being more like individual led. Would you be able to speak to that and how you actually sort of quantify clv?

Speaker A: Yeah, the top down view is you say, you know, I can see this company has been growing well over the past four years. This is the historical growth rate. If I'm going to be conservative, let's cut it in two and assume they can do that for the next few years. But there's nothing where you're kind of like looking at customer acquisition, looking at from the bottoms up in the sense of saying they've acquired this many customers thus far. This is my projection of what future acquisition will be. This is how many orders my existing customers will place. This is how many orders I would expect to get from the new customers. And I'm going to sum it all up across all the customers to get what my estimate of revenue will be. And so that's kind of what we mean by bottoms up. And it can be a really helpful way of kind of like vetting a, uh, revenue forecast. Because I would say that's one of the other things that companies uh, will often come to us to evaluate is you'll look at these pitch decks and inevitably they always have that slide in there and it shows revenue going like this. And then the question is, well, what would it take if we were to kind of invert that, like what sort of customer acquisition, retention and repeat purchase would we need to have to actually be able to get that? And then you can say, well, would I expect that? And so it kind of like inverts the question. Yes. I don't know if you're familiar with Michael Malbison, but he's kind of all about this. He calls it expectations investing. And the idea is to say the company is trading at a certain valuation right now. What would it take to be able to get that, to rationalize that valuation? And so I think that this could be a really useful way of kind of grounding you. Uh, and oftentimes if you find that our projections just come in low relative to what management is saying, it could be because there's this over optimistic view of acquisition or something else. Uh, and that might be kind of what would actually be driving the difference of opinion.

Speaker B: Makes sense, I guess, what you've just explained there, um, and you do see it in every pitch deck, certainly every pitch deck I've seen is that radical optimism because you've uh, got to paint that picture, I guess, when you're trying to go through that process, when you're coming with that, uh, maybe that more realistic m. Maybe not pessimistic, but realistic view, like where are you, what metrics are you usually finding are off, maybe overstated or just off most frequently. What are the assumptions that you're challenging the most? I guess in that process, if they

Speaker A: actually give the cohort data, it becomes hard to get too far off. Usually if they're like way off, it's because they're just not even attempting a bottoms up, you know, a bottoms up way of getting to the number.

Speaker B: Um, so I maybe meant the other way. Like maybe if you're the one doing that bottom up forecast and comparing it to more of the top down, homogenous, maybe a little bit of a rougher, rougher approach or looser approach to forecasting, less scientific, less exact. Where do you find the very, the differences to, to be the greatest when you bring that level of rigor to the process?

Speaker A: Yeah, usually if you've got a bunch of existing customers and you've been operating for a few years, you got a lot of cohorts. And so usually you get a pretty good sense of like this is what customers do after they're acquired. Uh, and usually there's not a whole ton of variation across the cohorts. If anything, usually the cohorts get a little bit worse over time that the customers that you're acquiring today. They don't repeat quite as much as the early adopters, because the early adopters, you know, they were like really into you. You know, they acquired back when you were, uh, back when you were young and probably the product wasn't what it is now.

Speaker B: Um, golden cohort, almost is.

Speaker A: Yeah, yeah. So, so usually, you know, if you say, all right, I'm going to assume, imagine we just assume retention, repeat purchase, basket size. It's all going to be the same as the recent cohorts. If anything, that might be over optimistic. And so then the question becomes, if we kind of made that assumption, what is this saying about acquisition and if acquisition is going to jump like this, uh, oftentimes that's what ends up being quite Unrealistic. Um, usually that's where you kind of hope. And they'll often have this in the pitch text as well. They got that obligatory slide, and it's got the TAM and it's got the sam and it's got the, you know, the current penetration. And the penetration is like a speck of sand. It's like, all right, you know, here we go. But, you know, I think if you were to kind of really be rigorous about that and kind of break down the market and say, like, this is the actual achievable market, given what they're doing right now, um, what can we really expect? You know, there are some times where, um, the implied acquisition forecast, like, blows out, you know, any sort of notion of sam. Um, so, yeah, so that could be a tell that uh, something is not quite right. But, um, typically customer acquisition, when you look at the trajectory over time, usually you see some regularity to it. And so even if we were to be fairly optimistic, there's not this humongous range that, uh, that you'd kind of expect above that or below that. Um, usually you can't deviate too much from what you would have expected given the historical data.

Speaker B: I guess when you've got that data, that predictability makes sense, um, within that equation of cltv. I guess just give a quick, um, run through, I guess, of your approach to defining and measuring that. Just for the listeners who maybe aren't familiar with that.

Speaker A: Uh, yeah, it's a big question. Uh, we'll often see people talking about ltv and you actually ask them for the formula that they're using. And oftentimes, um, it's as simple as realized cumulative revenue per customer. They were born some time ago, revenue up until today. That's it. Uh, and it's wrong on so many levels. Um, first is CLV should be a profit measure. And it should be. Specifically, it should be a contribution profit measure. You've got how much you spent to acquire the customer. You know, we call customer acquisition cost. Uh, you've got that stream of kind of contribution profits over time. Obviously the contribution profit should be all the revenue that you get. But then you got to subtract out all of the variable costs. And so if you're selling a product, all the labor, the materials, the shipping, the payment, processing, all of the expected returns, all that stuff, because all that's going to scale directly with revenue. And, um, got to be able to project that out suitably far into the future, but not too far. And then you really want to discount that to account for the fact that a dollar today is worth more than $1 five years from now. So yes, accumulated revenue is just not going to cut it. Um, but I would say people will make every mistake. They'll either not include cac, they'll use revenue instead of profit, they won't discount. Uh, sometimes people will use a finite horizon. They'll say the first three months or the first six months or something like that. Um, oftentimes they'll get the costs wrong. So they'll know that it's supposed to be contribution profit, but instead they'll use gross profit and they're actually potentially quite different from each other. Um, or they'll know that they should do contribution profit, but they won't include all of the variable costs. So, yeah, so there's, you know, there's a lot of nuance to it. And obviously the tricky part is, uh, especially for the customers that you've acquired relatively recently, if you want to, you know, to get like a two year ltv, you need to have an accurate prediction model. And that's really where, you know, theta and where Zodiac had come in was, um, you know, that prediction problem's tough, especially because, you know, for most C2C brands, 40 to 80% of their customers are going to buy one time and they're never going to come back. And then you have this really dedicated 5% of the cohort. That's amazing. They just love it. They keep coming back, they buy over and over and over again. And uh, when you have such extreme, what we call heterogeneity, which is just kind of variation across the customers, um, it becomes really hard actually to predict what those customers will do. So I would say maybe that's kind of the final thing that I'll see people get wrong is, uh, they'll kind of get the LTV to cac, but you're kind of averaging all of the bad customers and that tiny sliver of the good customers. And I think what could be the most actionable and the most diagnostic and the most useful to the investors is to get individual level estimates of value. And what you might often find is that there are these systematic differences and what makes those best customers the best. You can actually figure out what they are and why it is that they're so different from everyone else. And that can lead to a whole bunch of, um, profit enhancing things that the firm can do to acquire more like the best and acquire fewer of the worst.

Speaker B: Yeah, turning that into actionable tactics, whether that's creative product offer, um, Even retention tactics makes a lot of sense. How segmented would you say, uh, is optimal for that level of analysis? Obviously going to an individual customer level is extreme, but when we're looking at like cohorting out the different relative values of various customer cohorts. I think I saw an example in your, again, the material you shared with me previously, that of a brand where, uh, 2% of their customers made up, was it north of 80% of their total value or total revenue or something to that ilk? Um, you can just by hearing those numbers, you can see the value in doing that analysis and then turning that into action. But how, how in depth would you see brands going into that process? What do you, what are you, what's your advice there?

Speaker A: Yeah, you don't want to slice the baloney too thin, so to speak. So I would say if, yeah, if you've got the individual level estimates, very good models can get you that, and then you can do whatever, ah, cohorting you want. And the tricky part is if you knew what the very best segments should be, then you can just kind of pre segment the data that way. But typically the problem is you don't know. What you typically have is you got the transaction log data on the one side and you got the CRM data on the other side. And then the question is, all right, so what is this going to tell me? Um, if you can get those estimates of what each and every customer is worth, you can effectively run this big regression where you try to explain variation in those values as a function of all of the other stuff. And, uh, the best segments are the ones that really have the most signal? Well, it's kind of a combination. I'd say the best way to segment, it's a function of the strength of the signal. Does it really do a good job of discriminating between the best and the worst? And then the second thing is obviously can you actually do something differently because of it? So acquisition channels, maybe there might be a little bit less signal, but you know, you can allocate more to Facebook and less from Google, you know, so that's a very actionable channel. Uh, the same could be said for, um, you know, like product of first purchase or potentially like, um, the customers who make their first purchase online versus in a store, you know, so, yeah, yeah, that's stuff that you can really work with.

Speaker B: I actually got an example of some analysis I remember I did a couple of years ago for a brand that I was working closely on, where we layered CLV across postcode Data in the UK and found that we've really over indexed into ethnic minority um, postcodes. But then when we, when we analyze the creative we were putting into market the relative diversity it was. So it was, we had, we weren't indexing towards that at all. So it was a, maybe not quite as granular but an example of how that then turned into a tactic of intentionally recruiting um, individuals that would better resonate with those demographics.

Speaker A: Um, yeah, exactly right. Same would go for product. If you find that the highest value customers, they all like a certain product, um, put it in the creative, put it in the advertising, it just kind of makes sense. Put it on the website 100%.

Speaker B: Um, you mentioned first products there. I wanted to touch on that side of the equation. Like aov. I always say this clients. I think the ability to create meaningful changes in LTV post first purchase um, is significantly less impactful often than focusing more on AOV or what people buy at the point of first transaction. How true do you think that is and how do you see that play out generally across the businesses that you, you see and work across?

Speaker A: I would say yeah the general point that um, good customers are born and not made, I think um, that kind of rings true to me. We often find that if you had a model that allowed different customers to just have ah, different inherent levels of goodness, that doesn't really change very much. Um, but the different people will have different levels of love for the brand. That model can work really really well. Um, so to put it differently, if you spent a lot of your attention on acquiring the right customers and doing it in the right way, that can often be a lot better than trying to get a whole bunch of customers in the door who potentially could be pretty crappy and somehow try to do a whole bunch of magic to try to make the bad customers into good customers. You know that just um, often is not, not as good of a proposition.

Speaker B: Yeah it makes sense. I think that's like where's the point of leverage there and where it's better to spend your time. And I think the data often points towards it being on that um, better tactics to bring in better customers on a higher value product because they like they, they turn into high value customers long term. On the other, interestingly, like the force that we often see certainly over the last two years, it's almost like working against that uh, is discounting versus full price. Um, I feel like when we, I was pulling some analysis across all of our clients and discount rate at the start of this year and even if you just drag it back over two years, it's like a constant like force of, like a runaway train of just discount rate increasing over time. Um, at an aggregate level. How have you seen that play out into some of these models? Like do you, do you find that it decays? Usually just continues to decay value of customers over time post first purchase?

Speaker A: Absolutely, yeah. So that's uh, kind of one of the other. So there's what I call the taxonomy of clv. I've got a whole lecture in my class. It's just about like, I call it the taxonomy because you know, you've got like the golden definition of CLV and you got like all the other stuff that's like relevant, useful, but it's just not quite the same. And one of them is what we call NetCac and then net CLV or you know, kind of repeat CLV. Um, and the reason that's important is because if you acquire a customer on a discount, mechanically it's going to lower their CLV because the first purchase is less profitable, like right there. Even if they were identical after acquisition, just the fact that they came in on the discount means you're going to make less money from them, all else being equal. And so a question could be, um, what if we kind of strip that out, you know, so we kind of just look at everything that happens after that point and say, is there profitability after that first purchase Different. And what you often find is that um, those customers that came in on discount, they're bad for two reasons. The first is you're getting less profit on the first purchase. And oftentimes we find is their net clv, you know, that the value after uh, that first purchase is also lower. So uh, it just brings in a lower quality customer. So I'd say that that is something that um, we found to be true more often than not.

Speaker B: Makes sense. Such a tough drug to get off as. Well, once you've kind of as a brand have wedded yourself to discounting, the pursuit of greater volume of sales over time and not as much focus on quality. Um, certainly from what we see across our, across clients who come to us and we run that analysis.

Speaker A: Yeah, play devil's advocate. Like if one were to only offer that discount to new customers, then one could consider that to be customer acquisition cost. And in truth then it would be because it's only going to the new customers, it's not going to anyone else. And so one could imagine, you know, if you didn't spend up uh, a whole bunch of money on some big glitzy advertising marketing campaign, but instead just offers like a new customer promo. Maybe it could end up being that that's more effective, but the proof's in the pudding. Question is, does that actually play out that way? Um, and usually when these companies are offering the promos, they're offering it to everybody. And so oftentimes then you've got all the stuff that's happening with the new customers, but then you're kind of taking all those existing customers who might have purchased at full price anyways. But if you kind of dangle free money in their face, they're going to take it. Uh, and then they'll get used to it. Um, so it can kind of also have some negative repercussions for the value of uh, those existing customers too.

Speaker B: I, um, hadn't thought of the way you explained that around. Weighting that uh, discount into the CAC equation is really interesting. It makes a lot of sense. Um, I definitely agree on the repeat point is just like pulling full value future demand forward at a discounted rate is. I feel like people see that a lot.

Speaker A: If a customer's been with you for two years, they're pretty set in their ways. It can be kind of hard to like change what they want, you know, if they've been kind of getting the same stuff. But you know, if you give them like free money, anyone will take free money. So, so that can be the one thing, uh, that kind of gets them to do something to kind of take you up on the offer. But the problem is that's kind of what you don't want them to take you up on, you know, because they're going to probably not buy any more than they did. Um, but you're going to make less profit when they, when they do buy.

Speaker C: Quick one. Before we get back into the episode, every business spends money to make money. Ads, uh, software, flights, travel, the hundred unavoidable little things to keep the lights on. All that spending every month and none of it comes back to you. It just goes out the door. In Card is a financial platform built for digital businesses and it flips that on its head. They're giving 2% cash back for your first year on ads, SaaS, travel and everyday business spend. And it's all uncapped. The same money you were spending anyway, except now a slice of it lands back in your account. You also get multi currency accounts, corporate cards and Shopify, Meta Stripe and Xero connected in one place. A true source of truth. But honestly, you're spending Regardless, so you may as well earn on it. Sign up for free. Use the code SWU2K26 links in the description. Now back to the episode.

Speaker B: I wanted to just quickly touch on, um, I guess subscription, specifically subscription businesses and churn, um, and your approach to measurement of churn, because I know that's something that you also see has often been flawed, um, in people's approaches. We feel like everybody we come across is trying to launch a supplement subscription business in consumer at the moment. So it's definitely question of when will subscription fatigue really hit. I feel like they're about to roll out a law in the UK around making cancellation far easier and more mandatory. So interested to see how that plays out into the US as well. But how are you approaching, uh, measuring true churn, whether it's subscription or not? Um, and where do you see people getting that wrong as well?

Speaker A: Yeah, I mean that's kind of. You alluded to the biggest distinction, which is subscription versus non subscription. Um, non subscription customers still churn, but you can't see it. They don't tear up a contract or something like that. And so it's actually ironically at uh, theta subscription, it can be more like shooting fish in a barrel. It's just easier to do. You can specify a richer model for churn, um, but we'll often find that we add more value for non subscription firms because it's just trickier because you need to kind of deal with this non observability. We'll say that you kind of need to sort out, um, how often customers repeat and whether they're churning from the same purchase data. Whereas for a subscription firm you can sort out retention just off of the retention data and then you can sort out how much they tend to buy, how well they monetize while they're alive from what people do while they're still with you. And so it just becomes a much more separable, easy to model process. Um, so usually if you're within one of these firms, there's a whole bunch of best practices. Subscription firms got a model by cohort, um, you got to separate out, um, the seasonal cohorts versus the cohorts that are happening when you're not in a heavy promotional period, um, you really want to do the modeling again at the individual level. And there it really matters because you'll have a lot of customers that drop out early and then you'll have some that stay for a very long time. And so if you assume that they all kind of share the same retention rate, uh, you're going to horribly underestimate the value of the cohort. So, so that's incredibly important to do. But you know I would say to their credit if you're at a um, at a reasonably large subscription firm typically they'll have at least a decent, half decent model for customer retention. Um, you get to the smaller firms and they often just don't. So you really want to make sure you're kind of doing that right. Um, now non subscription, it's like the wild wild west. We'll work with the biggest companies in the world. Uh and it's amazing. Some of them, it's like they have nothing. So um, yeah. So there it really becomes helpful to have uh, basically what we would call a latent attrition model. It's just a model that allows for repeat purchase and then at some point the customer goes kaput. But you have to infer it from the data. It's just not something that's observable. And uh, thankfully there are some time tested models to do that. Um, they do better than models that don't allow for that. Uh, but at the same time we found again over the course of the engagements that we've done that there are a handful of uh, enhancements that you really need to have to be able to do the modeling well in that setting. So we've got this model that we call CLV Ultra. Um, and that's what we think to be the very best model for both kind of repeat purchase and retention modeling uh, in a, in both a non subscription setting and in hybrid settings. Because I'd say that's the other setting that often has been getting more and more popular these days is it's not quite full subscription but it's not full non subscription either. It's some mix of the two. Um, and this model tends to be really good for those kind of mixed subscription, non subscription type settings.

Speaker B: What's some of the complexity? Um, just speak to a higher level. What brings so much complexity to modeling that for non subscription businesses? Like why is it such a gap that you see, uh, is it just difficult to generate the data or general um, lack of understanding of how to build and run that model? What would you say the common issues are? Bottlenecks?

Speaker A: Uh yeah, the first issue is you can't observe churn and so turns happening. Uh, but we need to infer it. And so that's kind of level one. Um, level two is there are a lot of these kind of systematic dynamics that just we tend to see over and over and over Again, there's kind of dynamics as a function of the customer life cycle. So oftentimes what we find is that there's like this, ah, honeymoon phase. You acquire a customer and over the first, it could be between the first month or the first, you know, three months that they tend to, um, to purchase more frequently than you would think. And then they kind of settle into this baseline. And um, you know, we just find that time and time again. Before we actually had a model that didn't allow for that. And uh, then we started doing work for one of the largest gaming companies and it just did not work. And you can kind of see because I'd say the one thing that you always want to do is be empirical about your data. And uh, you may have your beliefs about how customers behave, but the way you can test it is just hold out the last six months of your data, train your model on everything else, predict the last six months and then see how good your predictions were along a handful of different dimensions. And that will tell you, is my model good or not? And we just found there were systematic issues when we did not allow for a honeymoon period. Um, so you want the data to tell you the length of that period. Basically you want to account for all those dynamics that happen as a function of the life cycle. Um, the third thing is calendar time effects. There's really two of them. There's the ones that are seasonal. People buy more Christmas time, they buy less during some other time of the year being spring season. You want the data to tell you that, and that will kind of hit all of the cohorts within those certain calendar periods and they will recur year after year. But then the other calendar time effect is what we call the non seasonal calendar time effect. And that's the stuff that kind of hits the one time, but does not tend to recur year after year. And so probably biggest example of that Covid hopefully hit one time, hopefully they come back. Uh, but boy, it hit all those cohorts. You need to account for all that. And there's kind of that baseline point that I made before that all your customers are really different. So you want to allow for those differences to exist too. So it's kind of like there's just a lot of stuff there that you

Speaker B: have to very complex.

Speaker A: Yeah, you got to deal with all of it, uh, or your predictions will be bad.

Speaker B: Makes sense. That honeymoon phase is definitely. See that. It always blows my mind how many people will repeat on month zero from uh, when looking at brand data, especially somewhere they're up into the double digits percentage wise for repeat rate in the first month, uh, after acquisition, um, even if they're selling loads of products where they just can't physically need more. Um, so you can see that kind of abnormal behavior in some of that data.

Speaker A: Uh, yeah, exactly. Yeah. And to your point. So let's say that in the first month, people buy a lot more than what ends up happening. If you're doing a cohort model and you don't allow for a honeymoon, as you see, you got these customers that were born two months ago and their first month was amazing. And you're like, wow, these cohorts are just getting better and better over time. And it's like, actually, no, they're not getting better, they're staying the same. They're just in the honeymoon phase and they're gonna, you know, they're gonna kind of settle down to baseline just like all the other cohorts did too. So. Yeah, that's just kind of model mispecification.

Speaker B: Yeah, makes sense. Um, we've talked through a lot of financial metrics. Um, a lot of this language maybe, maybe new for some of the listeners, may, may not. For those that uh, are more, uh, senior stakeholders are in those rooms more frequently. Um, I know you speak to that disconnect between like marketing and finance language. Um, if we're looking at, if we give it. For those listening today, like what, what do you believe that if they're in a marketing role and they're wanting to improve their, like, rigor of reporting and their ability to, to drive, uh, tactics off the back of some of these metrics, what do you think they should be looking at from a dashboard perspective every week? What are some of those, key, um, metrics they should be reporting into, maybe the next board or leadership meeting to really tell that growth story through the lens of customer metrics rather than aggregate top level metrics?

Speaker A: Yeah, I think it's a very, um, helpful exercise to ask what would the CFO care about and would the CFO agree with this? And so the discount rate thing, um, use the weighted average cost of capital. That's what the CFO does. But I can't count how many times I'll just see undiscounted figures. And immediately you lose credibility with the CFO because they're like, well, my investors are demanding a certain rate of return of my firm and you're not accounting for that in your equation. So that'd be one. But yeah, I think that the perceptual metrics, um, CFOs they struggle with perceptual metrics like Net promoter score or the open rate on the emails or we had this much engagement with our last social media post. It's like, okay, it's nice but like, where's the money? You know, I want to see the money. And um, and if it's not trans, if you can't translate it credibly to revenue, then it's like the CFO can't get all that excited about it. Um, so I think it's, it's when it's done right, everyone feels a little bit uncomfortable that the, the marketer needs to kind of like elevate up to kind of cohort monetization and cohort, you know, cohort costs and cohort revenue. Um, and you know, the cfo, they're not going to be quite used to that. But at least we're saying this is how much we spent on customer acquisition, this is how our cohorts are monetizing, this is how much revenue we got off these cohorts and that suddenly it's something that, you know, they think in that, in those terms. And Net promoter score, it could be very helpful if you can then establish a data driven relationship between NPS and monetization. Then it's like, okay, now I got it. You've got the low NPS segment, the high NPS segment and they monetize like this. I can get behind that. And oftentimes if you look to, I've spent a ton of time, if you go to my LinkedIn, um, I'm always talking about kind of company disclosure and we'll have all these examples of public companies that are reporting on this stuff and the CFO will have in the Investor Day presentation, cohort curves. And I love it every time I see it. But the fact that they're disclosing that and communicating that to external stakeholders is such a good sign. I mean, it's like an invitation to the marketing department to be a part of the conversation because they're really the ones who can own that and kind of like manage that better than most of the other people in the C suite could. So stick to that. That's where you're going to get the retention. And I'd say the other thing I would say about um, getting the CFO on board, why that can be valuable. Again, they're the ones that control the purse strings. So if you can kind of get them excited and seeing that you're a demand generator and not in a cost center, you're going to be more likely to get budget. I think you can make a more credible case to the CFO that some investment that you want to make is actually going to pay off over the long term.

Speaker B: I really like that approach as well. Taking some of the. I really like the idea of that approach and the MPS store's super interesting. Maybe quite easy to connect together as well with um connecting MPS to customer profile and therefore LTV versus some of the more like front end metrics. Um so probably one that some of the listeners can maybe take away one follow up questions that just obviously it's somewhat easy to articulate how I guess more direct response like meta. Uh some of these lower funnel channels I guess if you want to use that term translate into customer metrics compared to um maybe some of those like bigger swing um campaigns. Those things that are harder to predict the value of um say we're test. Say it says marketing teams wanting to go to explore out of home. They've never done it before. Um is there a way that you'd recommend people handling some of those conversations or thinking about or lens to think about that through for my own benefit as well. It's just interested to hear your thoughts on that.

Speaker A: Yeah, it becomes tougher. Um. I have a whole lecture on customer acquisition costs where we go into this in my class and uh it's really. I'd say it's tougher because the trackability is often not as good and the potential effect might happen over a longer span of time. So if you're right at the bottom of the funnel you either click through on the ad and buy or you don't. Oftentimes that's just a much quicker cycle. Um and the tracking can be really good because you can track the clicks so the attribution is better. But if you have had a home, you got people going by the billboard, they look up and they see it. That's an interesting uh I might check them out some point down the road. Um but you know the company won't know like you ali saw that billboard. You know they just won't be able to attribute exposure in the same way. And if it's more upper funnel which you know like a digital brand building campaign would also be then the effect might take longer and that kind of degrades your ability to do attribution. So um, those make it trickier and I'm not sure there's like a silver bullet answer to it but I'd say you know one thing that can be potentially better than nothing is some Sort of model that explains variation in purchasing or conversions as a function of your spending on out of home. Yes, it's something more like a marketing mix model, like a good one, uh, that's properly calibrated and saturated, you know, so that way you can say, I've got these other things I'm doing. I've got this thing too. I can allow for, you know, carryover effects that my spend today might affect. You know, this outcome measure of interest X number of months from now. And let me let the data tell me the strength of that relationship. Um, you know, that oftentimes might be the best that one can do.

Speaker B: Makes sense. I guess that's the, that's the conversation to have, I guess is like the rigor and bringing. Bringing it to the CFO with that level of like intention and from both an execution and. But also like a best approach to measurement. Um, it's probably going to get their sign off rather than a Let me chuck 50 grand at this and see what happens.

Speaker A: At least then you'll know. Yeah, it's like after it's done, did it pay off or not? And that can help also. Then it could be a learning experience that if you're not betting the farm, you run a few tests, but you have a very clear measurement framework for success versus not. You can kind of pre declare it to the CFO and then based on how that goes, then that can help inform whether you scale it more or pull it back, you know, but at least then it's like all laid out in advance in principle. So I think they would appreciate that.

Speaker B: I just wanted to flip into. We were chatting just before we started actually about Codex very briefly and your use and enjoyment and value that you're getting out of that as a tool. And I just wanted to spend the last 10 minutes before a quick couple of closing questions on AI. Um, can't really escape AI on a podcast nowadays. It has to be in there somewhere. I feel. Um, I'd be interested to hear your thoughts on how investors um, and general. Maybe public markets as well is looking at AI's impact on valuation impact on um, company structure as well. If you're exposed to that. Um, just. Yeah. Your general thoughts on how that should be viewed within today's landscape.

Speaker A: I mean there's a lot of different ways one can slice that, but certainly, um, a couple of them could be. Are you a part of the. I'll call it like the data or action layer or not. And um, just to kind of give one example of that. I don't know if you're, you know, if you're like a competitive athlete, but I'm um, kind of amateur. But decent.

Speaker B: Yeah, I'm definitely not. I'd love to say yes, but m. But yeah, definitely not.

Speaker A: Yeah, so I'm like, you know, decent with running. Um, and I use Strava and Strava. Um, one of the big things that they've launched recently was this, uh, MCP connection to Claude. Yeah. ChatGPT. They have like an indirect connection through the Apple Health app. So that way you can see the workouts. You can't get all the detailed granular data, but at least you can see, you can see some level of detail. Um, and now I find that I'm constantly saying, so how was my workout? It will tell me. Well, the evolution of your heart rate, you know, from the first half to the second half and here's like the, the seven year trend, you know, the story arc of your running career, you know, it tends to take you like 40 weeks to get to peak, you know, and I'm like, wow, there's no way I could have ever known that stuff before. Um, but suddenly it's like enriched all that data in this really powerful way and I'm like, there's no way I'm going to turn from Strava, you know, so.

Speaker B: Locked in.

Speaker A: Yeah. So it's basically part of the data layer. And um, honestly I think of Theta as part of the data layer too, that it's giving you all these good metrics and suddenly it's like AI has allowed us to understand what it all means for us and what we can do with it in a way that didn't used to be possible before. That just makes it all that data. It's not like I didn't have the data. I could do an export into a CSV file and I could start trying to run it through and you know, software. I'm like, you know, I have a day job and I got a baby and you know, I just have a lot of other demands of my time. I can't do that, you know, so now uh, I'm just like talking at my phone and I can get all that. And so, so I think there is like a generalizable lesson there that if you're either able to take advantage of it, people look to you as like, I want access to this data and now I can um, you know, glean a lot more insight from it. That's a winning position to be in or, you know, for action as well. Yeah. So I use Superhuman. I know they've got this MCP connection as well. So you can kind of very easily, you know, interact, you know, through Interact with the LLM through Superhuman. You know, that that makes it that much more useful to continue using Superhuman. Um, so, yeah, I think that's really helpful. I think there's a whole separate other set of companies where it's just like a crappy version of Anthropic or OpenAI. You go in there and you can kind of summarize my email or help me draft this, but it's using some crappy model. It's just not good. And that's just what I call the wrapper companies. Um, I don't see that as being nearly, uh, as useful. Um, I think figuring out where on that, uh, spectrum are you, I think that that can be a useful exercise. I say the other big thing is, um, when prospects are deciding who they want to buy from, I think that whole process is changing a lot right now. Most of my major purchases, I do a lot of research. Like, you know, I feel like maybe it's a terrible way to go about it because I'm like really different from the typical consumer. But most of my major purchases, I'm doing the research through ChatGPT and um, you know, whether it's my coffee machine, the heavy bag treadmill, anything that's above a certain dollar amount, and I'm always doing that research. And what you might find is that the answers that surface within the LLMs can be potentially quite different from the answers that you would have surfaced had you done a Google search.

Speaker B: I've seen that, sure.

Speaker A: Yeah. Most brands, they don't even know, they don't even know where they are to share a voice within LLMs. And so step number one, figure out where you are. You know, um, hopefully there's something you can do about it. But step number one, at least know where you are.

Speaker B: Yeah, the inputs and what drives rankings is very interesting. Um, you can use things like triple, well, have a good tool for measurement of LLM visibility for those watching who maybe want to do that for a consumer brand quite quickly. Yeah, things, reviews sites and um, Reddit, Quora, uh, things like that are having much greater impact than obviously on Google where they aren't factored in, um, for rankings. Do you feel like what you've explained is super interesting because you. And I also think it's like yourself and maybe that um, higher income, more educated consumer is making more considered purchases for AI? Um, I'm not sure the average customer is doing is on that journey currently. Um, will that happen? It's hard to say, but I think definitely seeing that, like, um, fragmentation of paths to purchase across more and more channels. Like when I'm looking, when you're zooming out, you've got that as an example. You've got things like app loving that have become more meaningful. You've got obviously meta, Instagram, TikTok. There's just, I feel like there's just more and more routes to attention than there was maybe like three to four years ago for consumer brands, which is interesting.

Speaker A: Yeah, that is kind of like the open question two years from now, what is like the typical Joe consumer doing? You know, um, so am I the canary in the coal mine or not? I think adoption is going to go up a lot. It's just going to get smarter and smarter. So if I had to make a bet, I bet that a lot more people will be doing it that way. But yeah, that's kind of an open question.

Speaker B: I agree for sure. Um, that segues nicely onto my final question I wanted to ask if you were to zoom m out and look at, say, the consumer brands that you've worked on over the last 12 to 18 months. And this isn't in the notes I sent over, so sorry for putting you on the spot. Slightly mentioned discount rate as something that I've seen as like an aggregate thing that's just increasing and maybe decaying value over time. Is there any, like, real macro trends when you look at consumer, uh, metrics across consumer brands that you think businesses need to be cognizant of or aware of? Or is that just not something you can, that may not be something you can pass out, but I'd be just interested to hear your thoughts.

Speaker A: Yeah, the macro trends. Well, certainly, um, cross cohort dynamics are big and you always want to be mindful of them. And that's kind of the whole question. Are the cohorts we're acquiring today, are they the same, better or worse than the cohorts that we acquired a year ago? Um, and they can get better for some businesses, they can get worse for other businesses. So you want to make sure, you know, for your specific business, um, in terms of the macro trends, I think, um, certainly there's the AI factor that, uh, can be a source of cross cohort variation. Um, and I think the gravitational pull that it exerts is going to be a function of where you stand relative to your competitors within your category. Uh, and then as you were saying a moment ago, the degree to which we See adoption of AI as being a way to do market research. More adoption, it's going to pull more. Less adoption, it'll still be a factor, but it won't pull quite as hard. Um, I definitely see that as being one factor, um, for direct to consumer businesses. I do feel everyone always loves to talk about the hot thing de jour. And um, you know, if you're like a shoe brand, you're a shoe brand. You know, like, people still could be buying shoes. They might change the way they do the research. Um, you know, are there kind of other macro factors that might influence the success or failure of a brand like that? Yeah, I would again say that, um, potentially the big B brands have more to lose through AI share of voice. Uh, because oftentimes we found that people who are asking an LLM and they say, hey, I've got this issue with my ankle. I've got weird arch support. What are the best shoe brands for me, um, it's going to surface brands that might not have otherwise shown up as much. Um, incumbent brands could stand to gain a little bit more. Um, yeah. Trend towards privacy. I think that that's something that we had seen with Apple's att. Um, that would make for less targeted marketing. Um,

Speaker B: I've been dealing with that in Europe for years. Americans don't like privacy.

Speaker A: It's a macro trend that's, um, still playing out. It's still playing out right now. Um, so, yeah, so to the extent that continues to manifest, you know, I think that, um, less targeted marketing, you know, it's gonna be harder for them to have precision. So that's all else equal, gonna be, you know, something of a negative for, for DTC brands. Uh, so, you know, that would just be something to expect. Um, beyond that, you know, I guess there's the whole thing of the supply chain and uh, the fragility of it. And yeah, I think that that's kind of an open question. But, um, you know, in general, the closer you are to your, you know, to where your products are being manufactured, you know, I think that that lowers the risk that there's going to be some major disruption that suddenly leaves you, you know, high and dry with your supply chain. Um, but, you know, I'm not sure I would call that a macro trend. You know, it's just like a potential black swan type of risk that seems, uh, more possible now than it did five years ago.

Speaker B: Definitely. I feel like we're in, um, that expecting volatility just as like that is normal, rather than it being like A freak event. Like, we are in an era of just like macro shock after macro shock on a short time scale. How do you plan for that? But, yeah, well, yeah, thank you very much for. I found that super, super interesting. I definitely learned a lot. I'm sure many, um, of the viewers did who have stuck around to the end. I really appreciate you walking through that today. I'd love just to end with a view of like, where can people find yourself? Where can people learn more about, uh, the work you've done? And then if there's any, uh, resources or things you want me to link specifically in the description that people can go to to learn a bit more, we'll definitely get those added as well.

Speaker A: Yeah, my big social media platform is LinkedIn. So, um, yeah, so if you search for Daniel McCarthy, UMD, um, you'll find me there. Um, I'm regularly posting about kind of all the sort of stuff that we talked about. So, um, if this is interesting, I think, um, definitely, let's connect. Um, and in terms of other resources, I'll follow up on that. But, uh, I would say the whole idea of kind of linking customer behavior to corporate valuation, you know, we've got this Harvard Business Review article that's a really nice introduction to that. So, yeah, certainly I'll be, uh, posting that too. But, um, yeah, Theta is my firm, thetaclv, uh, dot com. Uh, check us out and uh, I think if there's anything else, I think, um, certainly I teach this class on customer lifetime valuation. And um, more and more I'm doing executive education at University, uh, of Maryland as well. So, yeah, so if you find this stuff to be interesting, um, certainly. Professor Dan can, uh, I have a

Speaker B: spike in UK cohort next year. We're very UK heavy, so people might be popping across the pond.

Speaker A: Let me know.

Speaker B: Perfect. Well, yeah, I'll be sure to drop all the links in the description. Again, thank you very much for your time. Really appreciate it. Um, thanks everyone for stuck around like. And subscribe if you. If you're here till the end and we'll catch you on the next episode.