This is a free two-part episode of Marketing BS. My guest today is Peter Fader, professor of marketing at the Wharton School at University of Pennsylvania. Peter was one of my early marketing mentors and I loved this interview. This is Part 2 of the interview where we dive in deeper to the ramifications of Peter’s signature research around “Buy until you Die”.
You can subscribe to the podcast in your player of choice here: Apple, Sticher, TuneIn, Overcast , Spotify. Private Feed (for premium episodes).
Transcript
Edward: This is part two of my interview with Professor Peter Fader. Today we're going to dive into signature research Buy Till you Die. Peter, can you start by explaining what this idea is?
Peter: Yeah. That sounds really weird. Buy Till you Die. What's up with that? As we discussed briefly yesterday, it's not a model that I invented. In fact, I was actually against even trying it in the first place.
The idea is that if you look at the way customers behave—it's not just customers—if you look at the way that repeated decision-makers make repeated decisions over time. I'll give you a wide range of examples as we go on. There's this remarkably consistent pattern. I'll stop short of calling it universal, but it's so robust, it's so common, that we should treat it that way. Yes, we should acknowledge exceptions, but they don't happen that often.
The idea is this. Here's the analogy that I could tell, think about it in the case of a customer making repeat purchases of a particular brand or product. They're basically flipping two coins. Every day, you're going to flip coin number one, the buy coin. Will I buy this thing or not today? Simple as that, but they're also flipping the die coin. When that coin comes up heads, all that means is okay, fine, I'm still alive, I could flip the buy coin. It doesn't mean I will buy but it means I can at least contemplate it. But when that bad boy comes up tails, I'm gone, and I'm gone for good.
That's why we call it Buy Till you Die. There's no coming back. There's no resuscitation. You just buy things for a while, and not necessarily at a very steady regular cadence, but you do have an underlying rate and undying propensity to buy things, let's say once a month.
I don't know that you going to buy it once a month, but on average, you going to buy it about 12 times a year, but then something happens. I even no longer have a need for this particular product. I move away. You’re no longer tracking me. Perhaps I really do die. I don't know, and then that rate, boom, drops to zero.
It sounds really artificial. It sounds harsh. It sounds unrealistic, and I don't argue with any of that. I pushed back against it myself. When you put it up against actual data, and you allow these two coins to vary, it's remarkable how well it can capture, explain, and offer useful diagnostics about repeat purchasing behavior or, again, repeat decisions of almost any sort, and we'll dive into some of those almost bizarre examples.
Edward: Yeah. Let's talk a little bit about, you said right at the beginning, there are some exceptions, but they're very rare. What would be an example of an exception where this doesn't apply to you?
Peter: If you have some product or service where early on your customers either don't fully understand it, they can't use it as usefully as they can. Maybe some of the use cases for it don't emerge until later on, there might be some other complementary products or some changing behaviors. You might find people not just buying at the steady rate and dropping off, but there will be some cases—not just a person, but a whole cohort of people—will actually increase their purchasing for a while. That can happen, but it will level off and it will start to go down.
Eventually, the Buy Till you Die will kick in, but sometimes it might take a while. If we started with that theory of going in, we might understate things is. There could be lots of other little twists there. I don't want to get too technical about it.
For instance, it could be other changes in the marketplace itself, whether it's promotions that the company does, changes in competition, changes in the macroeconomy, that could make things a little bit less rigid than pure Buy Till you Die, and commercially, we can account for all of that. We have our basic core model, but then we can bring in some of these other situations and bells and whistles to make it just a bit more flexible, and sometimes it's very important to do that.
Edward: Are there industries where it doesn't apply like church attendance or travel to Florida? Are there things where that radically different than just like purchasing all of Amazon or it doesn't work? What are those crazy things do?
Peter: It's so funny that you mentioned church attendance because that is exactly the domain where this model was first dreamed up. I kid you not, Don Morrison who was a professor at Columbia at the time and then moved to UCLA. He's recently retired but is he just an interesting guy. He dreamed up this model literally while he was sitting in church in the Upper West Side in New York. He was looking at some empty pews and saying Mr. and Mrs. Smith, sit over there.
They missed this week, but you know what, they're often sporadic about their attendance, so that's okay, but Mr. and Mrs. Jones, they usually sit over there and they never miss church. The fact that they're not here this week, gets me worried about it. I wonder if they're ever coming back.
He actually dreamed up this model, and then did the math behind it in church, and then applied it to lots of other settings like that, whether it's nonprofits. Whether it's event attendance, all kinds of things, works really, really well there.
Almost any setting where people are making repeated decisions to do something, whether there's a purchase involved or not. It might be watching a particular media like we've applied these models to Hulu. Whether it's visiting a website. Whether it is making a purchase. Whether it's posting social media content. It's just remarkable how well this simple model can characterize forecast behavior.
Edward: Where is the resistance of the idea? You've been working on this stuff for decades, and yet I don't feel that it's like inundated the popular consciousness of business, even among experts in business and people who are the gurus of this stuff. Where's the resistance coming from?
Peter: From lots of different sides, especially when we talk about marketing. Yesterday, we were talking about how I have this heightened respect for the finance people. Even though I'm a marketer, there's a lot of BS that goes on in marketing.
When I bring these models forward, a lot of people will say, well, that might work well for company A, but our company is different, our practices are different, our customers are different, and besides, we're constantly being disrupted. We're constantly changing.
Marketers will come up with all kinds of excuses not to have some formal, regular, predictable characterization of customer behavior. I can go to them and say, give me some data. I'll show you how well it works. I don't even, you know what, you only give me half the data. We'll hold out the other half, and we'll show you how well the forecasts work, and this is what I've been doing for decades, and they'll still push back.
They'll say, okay, you know what, you can go talk to the nerds and analytics, but I have a business to run here. I need to focus on the brand. I need to focus on which celebrity we want for our Super Bowl ad, and they just don't want to be bothered with this technical stuff, but it wasn't till we commercialize it through Zodiac, which we spoke a little bit about yesterday.
Especially now that we're starting to win over CFOs and other finance people who can see how well these models will help them do their job, and they are willing to trust models. They are willing to look at forecasts and not only accept, but look for regularities in the marketplace. That's been very, very, very helpful.
Again, once the CFO accepts something, it makes it much easier to get the CMO on board as well, but sometimes they’re still will be resistance. That's one of the reasons why I've been writing a lot besides founding the companies.
One of the things that we haven't touched on is all of these kinds of books that I've been writing. All these books on customers' centricity, that are just basically a façade, a motivation, a Trojan horse, to get people to accept the models, to get people to care, to get people of want to run them, to get people to trust the outputs from them.
Writing these books on customer centricity has also been very helpful, but again, sometimes companies will say, okay, okay, okay, I'm with you, how do we do this? Then we'll start to bring in the models, and then their eyes glaze over once again. It's hard. It's getting easier, but it's no guarantee.
Edward: Can you give some examples of why it matters? Now I have these models, the models predict my future customer purchases far more accurately than anything before. My lifetime value of my customers could be different now. What does it actually change? Now I'm a CMO running my business, and I'm trying to figure out my next Super bowl ad. What is it going to change and what I'm actually doing on a day to day basis?
Peter: Yeah. There are some enormous implications that pop right out of the models. One of them is summarized pretty well in the subtitle of my first book. The book is called Customer Centricity, which doesn't really mean anything, but the subtitle, Focus On The Right Customers For Strategic Advantage.
There’s really three messages there. Message number one is that not all customers are created equal, you better not talk about the customer, and you better not focus on the average customer because they're wildly different from each other.
Thing number two is that the customers on the right tail of the distribution, they're not only more valuable than most of your customers, they’re orders of magnitude more valuable. I mean, there's, wow, are they good? Wow, are they going to continue to be good?
Thing number three is, there are ways that we can build our business around them. Let's really focus on those very, very valuable customers. Again, I'm talking about projected value. Not just historical value, although the two might line up with each other. Let's say what makes them different. How do they use and talk about our products differently from the average soso customers? What other services can we surround them with? How do we acquire more customers like them and what are we willing to pay to do so?
If we can build our business around those really good customers, we can make more money in a sustainable, defendable, ethical way than just trying to play it right down the middle, saying, will our average customer find this product or message appealing? It's wildly different than the usual way that people go to market, but the models strongly support it. That's why I spent a lot of time racing, okay, you got the models fine, but let's really talk about these implications, and they really matter.
It's been very gratifying to see a number of companies—I wish there were more—but a number of companies waking up smelling the lifetime value and starting to make decisions accordingly.
Edward: Is that the opposite of what Byron Sharp would say? Because Byron Sharp says, I think that your loyalty is effectively a function of your market share, and the way to get more loyal customers is just to get more customers and some percentage will be loyal. As you get more market share, your loyalty increases, and your double jeopardy law applies. Do you argue against that or is it a supplement to that?
Peter: It's a supplement. I'm glad you phrase it that way because pretty much everything that Byron Sharp, and of course, his original role model Andrew Ehrenberg said, 90% of that stuff is correct. Even there, it's going against the grain of conventional wisdom.
I am just adding an extra layer on top of it. I agree with the notions that you just described, the double jeopardy law, the duplication of purchase law. If your listeners aren't familiar with it, and that means that they're not listening to you enough, because I know you do a good job of talking about it.
Byron and company don't go quite far enough. I mean that in two ways. Number one, they assume that the models that they build, the fancy word for them would be the Dirichlet Multinomial Model. They assume that it's stationary. They assume that it's static. They assume that yes, there's the heterogeneity, but people don't change over time, and they do. They do in the way that we've been describing, Buy Till you Die, that there will be some non-stationary.
There will be some worsening of customers, and it's important to capture that. When we add that extra layer in, it does not take away from double jeopardy, it just adds another light to it. Here's the other part is that, again, Byron and company acknowledge that customers are heterogeneous, but they refuse to acknowledge that some of those customers in the right tail are so, so, so, so, so good that if we put a little bit of extra attention on them, that we can do better than just trying to be everybody's best friend.
All these things fit together, and I could get into lots more technical detail with it. Again, I believe everything that Byron says but he's leaving money on the table, by not allowing behavior change over time, and by not fully exploiting heterogeneity as much as I do. His points about, you still need to focus on mass marketing, and you still need to come up with products that are broadly appealing, I actually do believe all of that stuff. It's just that we want to put a disproportionate amount of attention for the care feeding and acquisition of those extra special customers.
Edward: I think I've totally bought in on to the acquisition side of things. I think more acquisition is always great. My concern a little bit is about you have these customers who are really, really, really good customers for you already. To go and give them additional incentives to go and buy more, at a certain point, the really good customers are almost spending all their wallets within the category with you already. They're already super loyal. How do you shift them to become even more loyal? Am I missing the point?
Peter: I got two words for you. First of all, and I never said the word incentives. That's your word, not mine. We got to find other ways to be crass about it, to squeeze more money out of them. Here are my two words, premium services. It's as simple as that.
You think about something like a LinkedIn premium. At first, there was a lot of pushback about a lot of the features and functions of LinkedIn premium offered. Folks at LinkedIn were saying, well, man, most of our customers don't want that stuff. Why should we offer it?
Well, the fact is, there are those right tail customers who are so good and use you so much, and use you so differently than everybody else. If we can come up with products and services that meet their fairly idiosyncratic needs and get them to pay for them, then we can make more money than just trying to sell them the same stuff over and over and over.
I look at something like Twitter. I'm a big power user of Twitter. I know you are, too. There's no question that I would pay $10 a month for all kinds of features and functions that most people couldn't care less about, to edit my tweets, to have more control over my timeline, to have more visibility, and whatever. There's a whole bunch of things that power users would want to use, but companies like Twitter, Facebook, and so many others are just too chickenshit to go out there and make these premium services a priority.
Jack Dorsey has made some noise about it recently but gets to it. That your heavy users want to pay more money, as long as they're getting good value for it. I think that's the key. It's not just giving them incentives. It's not giving them freebies, because you're right. They're going to buy from you anyway. It's getting to pay for more stuff that most customers wouldn't want.
Edward: Is the opportunity more in a product than it is in marketing? It should be helping to product team more than the marketing team?
Peter: It's a little bit of both. There's no doubt that we need to come up with products and services that are uniquely appealing to those customers will help us acquire more like them, but it is also in the messaging.
Instead of just going to an ad agency and saying, hey, ad agency, come up with a fun ad. I look at what some companies doing in my favorite company on these lines would be EA, the game company, Electronic Arts. They will look at their most valuable customers every day, by the way. They're updating lifetime value for every single one of their multi-billion customers around the world.
They'll look at the most valuable ones and say let's look at how they're playing a certain game. Whether it's Battlefield, Madden Football, or SimCity, and let's find out how our power users are using the game, talking about the game, what things they're doing in the game, and let's feature those kinds of aspects in our next set of ads.
Let's change our messaging as well, to make other customers aware of some of these features and some of these uses because maybe they'll find that appealing, or maybe it will help us acquire new customers, who will then become power users themselves.
There are ways to take some of these forward-looking metrics and models and use them in messaging as well, but you're right. It is more about either developing products and services or partnering with other firms. Maybe we won't even make any money on it, but if we can go to our best customers and say, we're going to surround you with all of these different sources of value, we're going to build a whole ecosystem for you. That's the way to lock them in and acquire more like them.
Edward: Can we do most of that without your models? I imagine most marketers know who their best customers are, or they can find that out fairly simply without a great deal of math. And then once they know who their best customers are, they can then go and build products and services for them. They can go try to acquire more of those customers. At what point do they need to have a Buy Till you Die model to do that?
Peter: It's an excellent point and the answer is yes. Let me elaborate. I'm so obsessed with these models, not only because of their practical value but even just because of their mathy elegance. That maybe I get into the model too much, and I used to really believe as I was writing the books and founding Zodiac that I can just give you the CLV magic wand, that money will just come raining down from the sky.
You're right that the models are just a means to an end, and you can actually come up with some decent proxies for lifetime value. It might be based on historical value, it might be based on something like Net Promoter Score.
There could be all kinds of proxies that aren't quite as accurate, aren't as predictive, aren't as precise as the models themselves, but they still do a pretty good job of sorting out who the good customers are from the not so good ones.
The harder part is, first of all, just to look for that. It's just to say that's what we got to do is to sort our customers out. To develop the insights, the capabilities, the organization, the corporate culture, to allow us to do all the things that I was just talking about a minute ago. That's the hard part, and absolutely, you could get away with some imperfect proxies of lifetime value, as long as you have the capability to do all the other stuff that I mentioned before.
You're right. You don't necessarily start with the models. You start with the mindset, you start with the tactics. You start with the organization and the messaging, and then once you're comfortable that you can do that, okay. Now let's bring the models in.
Edward: To refine it and make it better. If you're a CMO, and you're looking to make initial steps to move in this direction, because, again, at any large organization, we know that trying to change radically is very difficult. What's the Trojan horse to get this thing started?
Peter: Yes. I come in lots of different ways. I mentioned the books before, so let's start the sea level, and say the sky is falling, you're doomed to fail. It's going to be an utter catastrophe unless you repent and follow me. I'm overstating a little bit there, but this basically says there are fundamentally different strategies that you haven't thought about before. They're going to really celebrate the heterogeneity of your customers that can help you make more money. Let's start trying a big picture, like what are the limitations of traditional growth strategies? What are the windows to some of these new ones?
There's all that and then there's the data. Again, I've glorified the models maybe too much. I'm in the process of writing my brand new book, with my partner in crime, Bruce Hardy, and yet a new partner in crime Michael Ross, interesting guys. This new book is going to be called The Customer Base Audit: The first step on the journey to customer's centricity. Before we have any models, before we have any forecasts, before we look forward at all, let's just look at our historical data, stuff that's right there at our fingertips. To understand a lot of these ideas that I've been talking about, about how customers differ from each other, about how they differ over time and about how they differ from each other and how they differ over time.
Let's take a look back and just understand the basic patterns, but do so in a way that's both simple, but also very sea level motivating. Let’s just get you to appreciate the goal that's in them their hills and to really motivate the strategies, the models, and all that thing.
I'm coming at it every which way. Whether it's looking at historical data, whether it's writing books, whether it's focusing on finance, whether it's looking at other bizarre use cases of the models, I'm coming at it from every angle, eventually hoping that the message gets through and that the company says, you know what, let's try it out. Again, it's a long, long road ahead, but it's been working reasonably well over the last few years.
Edward: When is that book going to be up here?
Peter: Well, we're about halfway done with it. Actually, I just sent a revised proposal to my publisher, Wharton School Press. Sometimes, I'm going to guess, the middle of 2021, but then if any of your listeners are interested, I could probably send at least a sneak preview, a quick overview or even a sample table of contents, because we're really interested in these ideas, and the way that it really helped us build a bridge, from the big broad, almost qualitative strategies, to the technical forward-looking models to really complete the whole picture.
I think is going to really make a difference, and this is, by the way, is the first place I've spoken about it. You’re getting an exclusive, and I hope people find it appealing.
Edward: I hope so too. Thank you so much for being on the show today, Peter. Before you go, can you talk to me about your quake book? What book really changed the way you thought about the world, and you can't use one of your own.
Peter: I wish it was some mathy kind of thing, and there's no doubt that some of the books, papers, or journals that I've read as a professor have helped me out. But one book that makes me say whoa, and then I go back and read again and again and again, it's going to sound really strange, is Breakfast of Champions by Kurt Vonnegut.
I'm sure that a lot of your listeners might have seen that book years ago. Go back and read that book again. It's astonishing, just the creativity, just the mind-blowing alternative worlds that Vonnegut creates. I found that so inspirational, just in how I tried to think that there are no limits, and now I think that I can be just a wild creative guy and get away with it. Besides the literal story there, there are so many lifelong metaphors that are taken from that book. I'm going to sound really strange, but I can't recommend that one enough.
Edward: Thank you, Peter. This has been fantastic. I would love to have you on again.
Peter: It's always a pleasure talking to you and I look forward to the next opportunity.
Podcast: Peter Fader, Wharton Professor, Part 2