Many businesses will say that their customers are their most valuable assets, but few understand how true that is — or how accurately a measurement of customer-based value can price their whole enterprise.

In their new paper, “Valuing Subscription-based Businesses Using Publicly Disclosed Customer Data,” Wharton marketing professor Peter Fader and Wharton doctoral candidate Daniel McCarthy, along with professor Bruce Hardie of the London Business School, successfully built a model that allowed them to do just that — with formulas that concretely link the value of a business’s customers to the overall value of the firm. Fader and McCarthy recently spoke with Knowledge at Wharton about how their work can help corporate finance folks and retail investors alike, why they are making their models public, and why all customers are most decidedly not created equal.

An edited transcript of the conversation appears below.

The Value of the Customer

Peter Fader: This paper brings together two topics — one very old and well established, and one that is new and emerging. The old topic is corporate valuation. Everybody is talking about how you look at a corporation and value it. The new topic is customer valuation. Can we look at individual customers or group of customers and say what they are going to be worth in the future? This paper is all about bringing the two together, in a really rigorous and practical, real-world way. Can we do corporate valuation from the bottom up by looking at the value of current and future customers, adding all that up and saying, “That’s the value of the corporation”?

The basic idea has been around for awhile. It has been done by a few marketers in the past. But it has never had the real rigor to win over the respect of, say, financial people and accountants. That is what we are trying to do, the right way. So it’s customer-based corporate valuation, but done with all of the rigor, with all the really high standards and careful use of data, that accountants and financial professionals would respect.

Daniel McCarthy: There really have been two silos of work here: the one in the marketing domain, and the one in the financial domain. The financial domain has really hammered home how you value a business by projecting forward cash flows, discounting those back. Doing all of those nitty-gritty little financial details, all in a very precise and theoretically correct way. So we want to make sure to really draw upon that — take this problem that really has been solved in finance, and apply it to the marketing domain, where perhaps some of the financial details have been a little bit looser.

Experimenting with Your Satellite Provider

Fader: One of the really interesting aspects of this research is that we did this valuation exercise for Dish Network — a big, publicly traded company. The most interesting part about it is that we have had no contact with anyone at Dish Network. I have never exchanged e-mails, gotten phone calls or gotten data from anyone at Dish. This is using purely publicly available data. So the data we use, anybody could get access to, and, in fact, the methods that we are using are fairly common and transparent as well. There is really no secret sauce black magic here. It’s just taking publicly available data [and] reasonably well-established methods. We’re just combining them in what one might call a clever but at least thoughtful way, in order to come up with this kind of valuation. And because we’ve done it in this case, there’s no reason why we cannot repeat this exercise for other companies that make similar kinds of data available.

McCarthy: The other thing that we’re going to do to capitalize on what we think is a very fundamental methodology here, is make all of this available. We’re going to be collecting data not only on Dish Network, but also on many other companies that disclose the sort of metrics we need to perform customer-based corporate valuation. We’re going to release the data to everyone, so they can perform the same exercises themselves, and also release the methodology so people can actually implement these models, too. We really want to de-marketize this, and make the core concept of valuing the company by valuing the customers widespread.

We decided to perform work on Dish Network really by happenstance. There was no cherry-picking involved. We didn’t say, “Let’s find all of the companies this works for, then whittle it down to the one that a model fits.” No, we basically picked Dish just because it was one of the first companies that we happened to find a decently long time series for. We actually didn’t even know how long the time series was when we first began working on the company.

We definitely have a pipeline procedure for whittling down and identifying companies you could perform this analysis on. I think that process, in and of itself, has been quite interesting. Basically, we identify a universe of companies, all of the companies in the stock market, and some percentage of them use words in their SEC filings — specifically the 10-K and the 10-Q — that indicate they may release the sort of customer metrics that we would need. Then we whittle those companies down to a list of those that actually disclose the metrics. And then we have a team of great Wharton undergrads, as well as a few colleagues in India who have been working with us, to actually turn that into raw data sets for each of those companies. So Dish is just one example, but there are many others. We found at least 35 firms that also disclose the sort of data that we would need to fit these models.

“There is really no secret sauce black magic here. It’s just taking publicly available data, reasonably well-established methods.”–Peter Fader

Key Data – and Whether or Not You Can Get It

Fader: The real key here would be what kind of statements companies put out about the number or the nature of their customers. It could be how many customers they have acquired, or how many customers they have at the end of the period. Some companies would go further and say something about their retention rate or their churn rate, some kind of derived measure that takes some of the raw count data and turns it into some kind of more diagnostic metric. We would rather work with the raw data; we just want to know how many customers came in, how long they stayed around, how many were there at any given point in time.

Right now, there are absolutely no standards for disclosing any of this data or how it would be disclosed. One of the things that we want to do is to start to create some of those standards. That’s not the objective of this particular paper, but if we can shine light on the value of these customer metrics, and how they can be useful to marketers, to finance people, to people throughout organizations, then maybe we will start to have a conversation about which metrics would be reported, how they would be defined, how often they would be reported, and standards around the derived metrics that come off of them as well. It’s very, very early, so none of that has happened yet in any kind of formal way, but we’re hoping that this is the first step in that direction.

Key Takeaways

McCarthy: I would say the big one is, the procedure really does work. You can perform this, you can project out the cash flows, but now we’re augmenting the core financial model that you would use if you were working at an investment bank, and incorporate this customer metric data and use that to get an even more precise estimate of what these companies are worth. We arrived at a price that was on the order of within 5% of what the company had been trading at when it disclosed the filing that we used to train our model on. So we were very happy with the fact that, without any sort of jimmying around with the numbers, we were able to come up with a very sensible valuation for the firm.

There were a number of other interesting findings. For example, you really do need to allow for the fact that different customers are different from each other. You have some who are valuable, and some who are not so valuable; some who want to leave immediately, and some who are going to stick around for a very, very long period of time. So even though these companies don’t disclose this sort of data, we were able to infer that Dish, for example, had a hard-core, loyal segment of about 15% of their customers who had been around for over 10 years — it was very, very long term. By allowing customers to have different values, we arrived at a much higher valuation than we would have had we done the standard thing within marketing, which is to assume all customers are like the average customer. That would leave you with a price that is much, much lower than what the company had been selling for at the time.

Fader: I celebrate heterogeneity. I love to find and exploit and explain and talk about the importance of differences across customers.

This project is quite different because we have no direct indications of heterogeneity. We have no ability to track individual customers to say, this one left after a month, this one left after 10 years. We have no data like that. We just have aggregated counts of how many customers came in at a given time and how many were present at the end of a given period. We need to infer the heterogeneity, we need to fit models to this very poor, very sparse, overly aggregated data, and infer our best guess about how the customers differ from each other. Not only is that essential for us to come up with accurate valuations, but it also leads to very, very useful diagnostics that managers — whether of Dish Network or another firm — could really make great decisions from.

For instance, we think about the 80-20 rule — which states that a small fraction of our customers are going to provide a disproportionate value for the corporation. It’s a really valid concept. I spend a lot of my time trying to take a company’s internal data, to try to uncover the 80-20 rule. Is it really as concentrated as 80-20? Or maybe it’s less concentrated than that. In this case, we don’t have the data to do that, but we do it anyway, because we’re building really good models on the sparse data. We can still make inferences about how long different kinds of customers have stayed around.

We find out in this particular case, our best guess is that it is far less concentrated than 80-20, but still there’s still a good deal of heterogeneity, and if we ignore it, our valuations and our ability to make action-oriented statements about the company would be way off.

Do you think it would be kosher to at least drop a hint about the Wall Street analyst war? Why not.

“You really do need to allow for the fact that different customers are different from each other. You have some who are valuable, and some who are not so valuable.”–Daniel McCarthy

So let me start with it. One of the most interesting conclusions from the work doesn’t show up in the paper. In a lot of the work that I do, I like to hold out forecasts — look at the models that we’ve built and project what sales or whatever is going to be in the future, and see how well we do. And we tend to do pretty well. But in this case, besides just coming up with forecasts on our own, there already exists a universe of forecasts from Wall Street analysts. We have dozens of Wall Street analysts who, on a regular basis, are making statements about what the earnings of Dish Network — or any publicly traded company — are going to be one quarter, two quarters, sometimes two years from now. We can compare our models against their models or judgment or forecasts. Keep in mind, these are people whose jobs depend on coming up with reasonably accurate forecasts. So one of the things we did — and again, because the paper is so rich, there’s so much going on, we actually didn’t include it in there, but we’re always happy to talk about it — is we compared our own forecasts to those of the Wall Street analysts, and on average, [our forecasts] were actually considerably more accurate. Not to say that the Wall Street analysts do a bad job, but some of these very simple marketing models applied to relatively simple, sparse data can do at least as good of a job as folks who have access to the CEO and data that we couldn’t even imagine having.

McCarthy: I wanted to build on a point. It’s pretty simple. The crux of this work really is coming up with customer lifetime values. We’re going to take our crystal ball out and project forward what every customer is going to do into the infinite future. Really, that is what is driving this model for what revenues are going to be, and thus, what the stock prices are going to be. What’s most surprising is that even as we’re making such short-term forecasts, we are able to do so well. Essentially, to come up with a stock price is more of a long-term game. But we are actually able to do very well doing short-term predictions as well. So that was definitely very striking to me.

Spreading the Word

Fader: I’ve been spending a lot of time lately talking about customer centricity, trying to get companies to understand that there’s gold in them there hills when it comes to customer data, and it can not only improve their marketing practices, but pretty much every part of the organization. We can make better decisions about how our salespeople are doing, what kind of products we develop, how well our production people are doing. And we can make an impact on the finance side as well.

So for me, this idea of customer-based corporate valuation has been kind of a holy grail I’ve been going after. If I can win over the CFO and the other people that are running the finance and accounting and control parts of the business, then I can get these strategic ideas of customer centricity to spread much more broadly, much more impactfully throughout the organization. So for me, it really is reaching out and saying, “Hey finance guys! A lot of the stuff that I’ve been talking about to the marketing people, it’s relevant for you, too, and it will help you make better decisions about the company and maybe your competitors.” And so there, I have been really happy to see the results, but of course, it is much more than that.

A Takeaway for Investors:

McCarthy: One of the other practical implications is this: Just speaking from a financial standpoint, I had been at a hedge fund for a number of years before coming back, seeing the light and going to academia. So you can take the Wharton kid out of finance, but you can’t take the finance out of the Wharton kid. But you can definitely construct some sort of a trading strategy to be able to exploit this data systematically across all of the companies that disclose it. So again, there are on the order  50 to 100 companies, let’s say, in the statistical universe that disclosed enough data that you could use it to construct some sort of customer-based valuation model. The idea would be that you could find some companies that seemed overvalued, and some that seemed undervalued, just by the results that the models entail for each of these companies. Buy the ones that are cheap, sell the ones that are expensive, and you’re not taking on any real risk. I think such a strategy could potentially be very interesting. And in turn, that makes it very interesting for the executives of those firms, knowing that people are buying and selling their stocks based on this information.

The whole idea of standardizing these metrics, making sure that you are improving the health of your business in a bottom-up way — making sure your customer metrics are strong — I think it just reinforces something that they really probably know they should be doing, in any case.

“We compared our own forecasts to those of the Wall Street analysts, and on average, they were actually considerably more accurate.”–Peter Fader

Fader: As Dan mentioned, there are all kinds of interesting hedge fund possibilities, and a lot of folks who are managing hedge funds or working with corporate valuation in other ways have been looking very carefully at this work. I want to say right now that I am a marketing guy: I have no intention of developing a hedge fund. If other folks want to take this research and do it, then that’s great. I will find that really rewarding even if I don’t have a piece of it. For me, it’s more about corporate strategy more generally. It’s more about making better use of the data assets that we have. For me, the big payoff would be if folks with hedge funds and other investors go pounding on the doors of public corporations, saying, “You must reveal some of these other customer-oriented metrics. We can’t make the right decisions until you do so.” And maybe for accounting standards boards to start to make some noise about this as well. If we can change the kind of data that is being put out there, and the kind of conversations that we are having, beyond just the “making money” thing, I want to change business practices and the kinds of data that we look at, and the way that we use it.

The Customer as a Real, Long-term Asset:

McCarthy: One other very practical implication is, there is a whole budding stream of research about this basic question: If I’m going to spend marketing dollars to acquire customers, should I take that hit on my income statement immediately, or should I capitalize that as an asset because I’m acquiring these customers who are worth a lot of money? Part of our research is showing just how valuable these customers are, and one of the implications is they are going to stay around for a very long period of time. We came up with the estimate that the average customer at Dish is going to remain a customer for about five and a half years. That is very, very long. And you have some customers who are going to be around for over 10 years.

So if I am going to spend subscriber acquisition expense dollars, I am investing in an asset. The customers are assets, but they’re not acknowledged to such. I think over 80% of Dish’s subscriber acquisition costs are immediately expensed, and as a result, they take the hit if profits don’t look good in the short term…. If I’m going to spend money and basically acquire an asset, I should be able to recognize the asset and have the expense associated with the asset bleed away as the asset’s value bleeds away….”

Fader: Or to put it another way, … what’s the chair metaphor?

McCarthy: Yeah, it’s like if I buy a chair, I’m going to capitalize that immediately. If I buy a building, that’s an asset on my balance sheet. But if I buy a customer, it doesn’t show up anywhere.

Buying Customers:

Fader: We see these ginormous acquisitions taking place, in many cases were one company is buying another largely because of their customer base. So you can think about when Facebook bought WhatsApp, or just endless rumors about Yahoo! or AOL. [The companies say,] “We’re just buying the customer base.” But what are those customers worth? Well, that’s what this research is all about, and it’s interesting to see that sometimes in the press, you will see back-of-the-envelope calculations about what one company’s customer base is worth, and they’re appalling. The analyses that they are doing are terrible, the numbers that they are coming up with are way, way off. So we like the idea that people are having that kind of conversation, but we want to make it a little bit more educated. We want the answers to those questions to be as rigorous as the kinds of things that we would expect to see in a conversation about finance and accounting, not kind of lightweight approximations that might be associated more with marketing.

McCarthy: One example that’s been in the news lately is on topic: Dish Network. They are in the news about potentially being acquired or merging with another company. They’ve been in the news themselves, trying to find a partner in a way to move forward, so I think this is a perfect example of how customer-based valuation can be useful. Our research is all about valuing the customers, so all of these other assets that the company may have that really aren’t coming from the value of the customers, we’ll leave that to the finance guys. But specifically for the value of the customers, we think there is real gold in basically looking at them in kind of the way that we are. And all the better if we had internal company data. We could just make those forecasts even more precise. But we think, even with Dish Network, there is really more potential to think about it in this way.

Correcting Common Misunderstandings:

McCarthy: I would say perhaps one of the biggest misperceptions we have seen is not acknowledging the paramount importance of the value of the customer to the value of the firm. And really establishing that link and then focusing on the value of the customers. I would say perhaps one of the most striking examples of this is just in the accounting rules itself. If a company was to buy a chair, they would be able to capitalize that as an asset on their balance sheet. And they don’t recognize an expense immediately. But if I was to spend money to acquire a customer, that customer doesn’t show up anywhere. They’re called an intangible asset. One of the interesting byproducts of our research is that, for Dish, for example, these customers are expected to basically stay with the firm for about five-and-a-half years. So they have very long, useful lives.

And, in fact, with Dish in particular, their example of the chair is the actual dish itself. The money they spend on these dishes that you see, they are able to capitalize and they have a useful life of only four years. So you actually have this kind of striking example. It’s like a tale of two worlds. You have this inert object that has a short, useful life. I am able to capitalize that, but here are these valuable customers who are generating cash flow over very long periods of time, and I need to expense that immediately. So we think that is really something that should be looked into more carefully.

Fader: Another important point that the paper addresses would be the gap, or I would rather say, the synergy between marketing and finance. Marketers will often talk about this idea of customers as assets, and will often say, almost in a purely conceptual way, if we add up the value of the customers, boom we get the value of the corporation. And while it is a nice idea, we haven’t seen work so far that has really won over the finance and accounting community. But you know what? There really is something there.

“You could find some companies that seemed overvalued, and some that seemed undervalued, just by the results that the models entail for each of these companies.”–Daniel McCarthy

Now, I am happy to say there have been a few mavericks in finance and accounting that have been saying some of these words, but it has not really become mainstream yet. We’re hoping this paper, not because it’s coming out of marketing, but just because of the rigor involved in the data side, on the analysis, on the substance of implications, is going to win over broad interest from folks in all different business disciplines, and say, “You know what? The marketing folks really do have something to add to the finance and accounting conversation.” And of course, finance and accounting have a lot to add to marketing. If we can break down stereotypes, and break down silos, and make everybody across the organization not only smarter, but looking at the same kinds of metrics, then that’s got to be a great thing for shareholders.

What’s Next

McCarthy: All of the work thus far has basically been to value a single company. So we lay out a framework and then we apply it to a company. We want to show, in the next paper, that essentially you can apply this across many, many, many different companies, along two dimensions. It’s not just rehashing the same analysis that we did for one company 100 times over. It really is that there is a whole broad spectrum of data availability with these companies — some that have disclosed a lot, and some that have disclosed very little. And there actually is a whole heck of a lot of information you can learn about those companies that may not have disclosed very much, when you look at the patterns across companies. So we want to bring all of those companies together in what they call a Bayesian hierarchical model, and be able to fill in a lot more of the blanks for companies where the data is not so good, and apply this in a systematic way across all of them simultaneously.

Fader: Another big question is, what metrics should be disclosed? Right now, we are limited by what companies choose to disclose, but are those the kind of metrics that investors should be demanding? And would it be the same kind of metrics for every kind of business? So an important caveat of this work is that we have only looked at companies that operate in a contractual setting where we sign customers up, they pay on some regular basis, and then they go. And the company knows that they have left; it is observable churn.

So this model applies to pretty much any contractual setting, and that’s great, but there are many companies out there, such as Amazon or a grocery store, where there is no formal contract, where people just buy things on an occasional basis. Then there’s this long gap, and the company is left wondering, “Is that customer gone, or are they just taking a snooze between purchases?” If you were in a non-contractual setting, what would be the right kind of data to disclose? If you are an investor in a non-contractual company, and you are pounding on the door of the company, what data are you going to be demanding for them to disclose? It’s going to be different kinds of metrics. So we want to have a broader understanding of what are the right kinds of metrics that should be disclosed in different kinds of business settings, and then of course what kind of analysis would we lay on top of it in order to do the corporate valuation.