New research from Wharton is changing the way investors are looking at the valuation of businesses by taking a closer look at the worth of those firms’ customers. Wharton marketing professor Peter Fader and Dan McCarthy, a professor at Emory University who earned his Ph.D. at Wharton, are behind this new method that has recently received a lot of attention for the insights it generated about the value of businesses including meal delivery services Blue Apron and Hello Fresh, and online furniture retailers Wayfair and Overstock. The professors joined Knowledge@Wharton to discuss their research and its potential effect on transparency. They previously talked with K@W about how they developed this method for valuing subscription-based and non-subscription-based businesses. The research was partially funded by Wharton’s Baker Retailing Center.
Knowledge@Wharton: You started out with this model for subscription-based businesses. Tell us how it works.
Dan McCarthy: What a traditional financial analyst would do is project out future revenues and have that drive a traditional model for the overall valuation of the firm. Basically, the analyst would model how those revenues trickle down into profits and use that to estimate the valuation of the firm. We don’t dispute that at all. But the main thing that we would claim is that embedded within that revenue number is a lot of potentially very useful customer behavior decompositions that can be done.
Instead of thinking about revenues as a monolithic unit, we can think of it in terms of revenues coming from new customers and existing customers who are being retained. Companies that do a very good job of retaining their customers and developing the value of everybody over time should be awarded a much higher multiple than another business that does not retain their customers as well, even if the historical revenue patterns look exactly the same. At a high level, I think customer-based corporate valuation is taking a standard valuation model and providing an extra dimension to how those revenue projections are made.
Knowledge@Wharton: The next stop for this research was non-subscription-based businesses. Those are a little more complicated because it’s harder to know when a customer churns in or out. What data becomes particularly important when you turn the model over to this type of business?
Peter Fader: Let’s back up a step. Companies disclose different kinds of information. Some disclose nothing when it comes to their customers. There’s no obligation to do so. But some companies do, for whatever reason. Maybe they just think about them as trophy metrics. Maybe they think there’s some diagnostic value — and there is. The big issue in our mind is whether it’s a subscription business or a non-subscription business. Non-subscription is much more difficult because you don’t know if the customer is canceling a contract. All you know is they just stop purchasing.
This has been an aspiration of mine for many, many years. Can we just take some aggregated, company-disclosed metrics and back out the nature of the unobservable lifetimes? How long are these customers going to stay around? How many transactions are they going to make? How much are they going to spend? This was a challenge that I put in front of Dan that ended up being the heart of his dissertation. Both the statistical results as well as their managerial implications are, well, awesome.
Knowledge@Wharton: You applied the subscription-based method to a meal-kit company called Blue Apron. At the time, the company was about to do an IPO and was getting a lot of press for its impressive growth. Tell us about your findings?
McCarthy: They had put out an IPO prospectus and threw in some customer metrics. It’s almost throwaway statements that further support how the company has been able to rapidly grow. They gave just enough information for us to be able to use the same sort of methodology that we described to you, that was also applied to Dish Network and SiriusXM. I used similar sort of methodology for Blue Apron. It ended up uncovering that even though the company had been able to generate very rapid growth, it was primarily driven by heavy marketing spending. They had been just throwing heaps and heaps of money at customer acquisition. It was bringing in a lot of new customers, which was generating revenues, but existing customers were not staying around for very long.
After six months, 70% of customers that had been acquired had churned out. That did not spell well for the future success of the business because it implies that they’re on something of an acquisition treadmill. To be able to continue to show that very strong revenue growth, they’re going to need to spend more and more and more money on customer acquisition. That’s inherently just much less profitable, which made it much less likely that they were going to be able to achieve profitability at some point in the near future.
Fader: Dan put out a couple of LinkedIn posts. It was almost like, “Hey, look at what we can do with the kinds of metrics that Blue Apron is putting out there! We can find all kinds of insights that they didn’t disclose and tell you the parts of the story that you want to know.” Those LinkedIn posts went viral. A lot of people, more in the investment community than academics or marketers, picked up on it and said, “This is really important information. It completely changes the way that we see Blue Apron.”
It’s also carried over to a number of other subscription-type businesses. That analysis that Dan did has really become the de facto way to approach these kinds of thing-of-the-month clubs. I think all of them, and their investors, are paying very close attention to the work that Dan has done and the things that he continues to say in social media.
“This really was a story of fundamental financial weakness, and I truly believe that these methods would have helped investors dodge that bullet.”–Daniel McCarthy
Knowledge@Wharton: Dan, give us more details about how the companies and investors are taking notice of this.
McCarthy: Blue Apron had originally priced their IPO at $15 to $17 per share, which put them up at about a $3 billion valuation. That’s a very healthy valuation multiple of their then-current revenues. I put out my analysis about four days later. That was the main analysis that ended up going viral. Five days after I posted that, they slashed their IPO target range down to $10 to $11. They ended up issuing the IPO at $10. Last I checked, it was hovering at around $3 a share. Even from that kind of heavily discounted IPO price, they’ve fallen a further 70%. That’s no coincidence, relative to the 70% of customers that churn out after six months.
The implication on the stock price has been very stark. A lot of people had been blaming the Amazon acquisition of Whole Foods, but it’s just total baloney. Every time Blue Apron had released a new set of earnings, their fundamental performance at that time was very poor. In each of the last couple of earnings conference calls, on the day that they disclosed their earnings, the stock had dropped by 18% or more. This really was a story of fundamental financial weakness, and I truly believe that these methods would have helped investors dodge that bullet.
Fader: You would think that Dan McCarthy would be Public Enemy No. 1 in the eyes of anyone associated with Blue Apron, yet they seem to have a great deal of respect for him. Various people have conversations with him about what’s behind the analysis and what it means for them on an ongoing basis.
McCarthy: To their credit, they’ve been very mature. They’ve been receptive to the analysis, and I think they’ve been taking a lot of steps to improve the underlying retention issues that I’d called attention to. I’ve had a number of ongoing communications with their communications manager. Jared Cluff, their chief marketing officer, has acknowledged the work publicly, saying that a number of the executives are following the work because of its relevance. I think that they’re doing what you would hope that a mature shareholder-value-focused management team would do. Instead of trying to shoot the messenger, they’re really trying to solve the problem at hand.
Knowledge@Wharton: You also did a case study using the nonsubscription-based model on online furniture retailers Overstock and Wayfair. That paper also had a pretty big impact on one of those retailers. Tell us about that.
Fader: Let me set the stage, and I’ll let Dan get into the details. Remember, this is primarily academic work. We’re trying to establish both the credibility of this direction of analysis as well as the methods that we use to implement it. We looked around hard. We needed to find companies that, in their public disclosures, were giving just the right kinds of metrics that would let us back out the lifetimes of customers. Overstock.com and Wayfair just happened to give the right metrics that let us do that. That’s the reason we chose them. We have no particular interest in those companies, and the fact that they share a similar sector doesn’t really matter. In fact, they’re very different kinds of companies. But they gave us the right kind of data.
“The companies that have a very difficult time are the ones that don’t do a good job of retaining the customers.”–Daniel McCarthy
Very objectively, taking into account none of the kind of institutional details of these companies, just looking at the numbers, we projected the number of customers to be acquired, how long are they going to stay, what they’re going to do, how much they’re going to spend and how that varies across the customer base. The results were stunning, both in terms of what the numbers looked like and how they were received.
McCarthy: The numbers for Overstock were a little more like Dish Network and SiriusXM. We ended up with a price estimate that was very similar to the then-current stock price. For Overstock, we basically concluded that it seemed about right. For Wayfair, that was very far from the case. I think the stock had been trading at about $64 at the time that we did the analysis. Our price estimate was down below $10. They definitely took notice, as did a lot of other people within the investment community.
We’d posted it to SSRN (the Social Science Research Network), which is the standard place where people post marketing science work. I believe it was the very next day that a very famous short seller, Andrew Left from Citron Research, began tweeting about the analysis. It created this veritable flood of interest and downloads in the work to the point that it’s on the verge of hitting the top all-time most downloaded list.
I’d say that work has also generated a lot more attention, even more so than Blue Apron, from your traditional hedge fund investment analysts. We’ve received a number of calls and a lot of interest from just various XYZ capital hedge funds and sell-side equity research firms. Unlike the Blue Apron work, there were actually two sell-side equity research firms that put out research notes that were entirely devoted to the work that we had done. The reason why was because the day after we posted our analysis, the stock had dropped about 10% without any other news on the day. It really did seem like the reason that the stock had fallen that day was because of this work and perhaps the visibility that was generated by the short seller calling attention to it.
Knowledge@Wharton: All of this reaction that you’re getting seems to point to a desire on behalf of investors to get this kind of information. It indicates that something was missing before about how we value companies.
Fader: It actually goes both ways. Dan said how a number of these analysts picked up on the research, but it doesn’t mean they agreed with it. Take into account the point I made before, which is that we didn’t look at any of the kind of institutional, contextual details of the company. We were going with the numbers. A lot of the analysts who didn’t like our results, some of the big shareholders in Wayfair, said, “Well, you’re ignoring this detail or this speculation or the overall spirit of the company.” And we’re saying, “Yeah, we are. We want to be as objective as possible.”
A lot of people just didn’t like what we were saying and were finding excuses to shoot it down, but it did spark a conversation. It did force them to confront some of the numbers that Wayfair itself is putting out there. And it did force them to try to explain why some of these repeat purposing numbers were so weak. Some of their explanations were also pretty weak.
Knowledge@Wharton: One of the analysts had written that your research didn’t take into account that furniture-buying is cyclical. People often do it seasonally. You said that was not necessarily a factor.
“A lot of people just didn’t like what we were saying and were finding excuses to shoot it down, but it did spark a conversation.”–Peter Fader
McCarthy: That’s definitely one of the big concerns that some people raised — that given how people buy furniture, these methods would not apply. It’s definitely not true. We can speak to that from two angles. There’s the theoretical angle, that our models allow for long purchase cycles. If someone only buys once a year, our models can definitely say that this is a person who has a low base purchase rate but is definitely still with the firm. Our models can certainly infer that.
The other way that we can approach it is from a pragmatic, internal data standpoint. Pete and I are also co-founders in a predictive analytic startupcalled Zodiac. Through Zodiac, we’ve seen many, many firms that have provided us their data that have very long inter-purchase cycles. We know firsthand, from having predicted data for those types of firms, that our models can certainly model and predict future behavior for them very well. So, both theoretically and practically speaking, it’s really not something that I would be concerned about.
Knowledge@Wharton: Do you feel like this research points to an inherent problem with companies over-investing to acquire customers and under-investing to retain them?
Fader: My casual inference from the Wayfair episode is exactly that, that they’re spending like crazy to acquire very, very inefficiently. If you just look at the surface level, you see this kind of hockey stick, exponential growth. But it’s all acquisition. They’re just getting a bunch of people to come in. Those same people aren’t coming back nearly as often as you might expect them to be. In fact, their repeat purchase rates are a lot lower than they were for Amazon, even close to 20 years ago. And this is from their own data.
I’m not saying that Wayfair’s doing anything unethical. Again, if that’s what the shareholders find most appealing, to just pump all that money into acquisition, then they’re following the right instincts. But if you’re being really objective about it, and if you’re trying to find this balance between acquisition and retention, it’s really important to look below the surface of the data.
McCarthy: You can have two companies that both pump a lot of money into customer acquisitions. What’s really going to differentiate the winner from the loser is how well those companies are able to retain those customers after they’ve been acquired. If you were to take a company like Harry’s or Dollar Shave Club or Netflix or Amazon, they are tremendously able to keep their existing customers coming back. Roll forward the clock 10 years and those companies don’t have to spend very much money to acquire new customers. And they’re able to become profitable.
The companies that have a very difficult time are the ones that don’t do a good job of retaining the customers. They are never able to dial back on customer acquisition, which makes it very hard for them to get out of that loss-making situation. I think it really puts a very strong emphasis on the importance of retention.
Knowledge@Wharton: What’s the lesson here? How can a company dealing with this particular issue turn things around?
Fader: First and foremost, just use your own internal data better. Find that right balance between acquisition and retention, just as Dan said. Just given the external incentives from Wall Street, companies haven’t been doing that very well. I’m not saying it’s easy, but they’re not even looking to do so. Also, what metrics should they be disclosing?
McCarthy: I think for subscription-based firms, it’s pretty clear. If companies simply disclosed the number of customers they acquired by quarter and the total number of customers they have every quarter, we can do a reasonable job of being able to model and forecast. And we would be able to back out the underlying retention of customers at those firms.
Where it becomes trickier is with these nonsubscription firms like Wayfair. For them, I’d say we’re still doing some ongoing research on that. But I think the main conclusion that we’ve reached so far is that if they were to simply disclose one additional metric, we can play the same sort of game that we do for subscription businesses. The number of customers that they acquired period by period, the total number of customers who are still active every period, and the number of orders that are placed each period. Just those three. But now, we can back out not only how many customers are acquired over time, but how many purchases those customers will make before they end their relationships with the firm.
Knowledge@Wharton: If companies are doing a good job of retention, if they’re disclosing this, that’s good news for them. But if they’re not, that’s bad news. Then why release that data?
McCarthy: I think the fact that Overstock and Wayfair have released that data is more just to be helpful. I don’t think they actually realized that people could read between the lines and fill in the blanks about what that implied for their customer retention. If they did, that’s wonderful. But yeah, I think it really would speak to, “Why would you be disclosing this if the numbers didn’t look good?”
I think there are a lot of companies that have taken the opposite tack from Overstock and Wayfair and just not disclosed any meaningful set of metrics at all. In fact, most nonsubscription firms do not disclose anything like what Overstock and Wayfair did. That’s part of the reason why I think Pete and I, going into the valuation exercise for these two companies, were almost sad that that was the conclusion for Wayfair. Because at the end of the day, we would like a lot more companies to do what Wayfair is doing and disclose exactly the same sort of data. We really have no vested interest at all. If there is any vested interest, it’s that the price is more normalized.
“If you’re trying to find this balance between acquisition and retention, it’s really important to look below the surface of the data.”–Peter Fader
Fader: We hope it’s going to come down to investors demanding that these companies disclose the kinds of metrics that are going to let them do their job properly. At this point, there’s nothing about accounting standards. We’re decades away from anything like that happening. But if investors are going to these companies and saying, “I want to come up with as accurate a forecast as possible for you, so you have got to give me these rolled up, quarterly metrics about total orders and so on,” I think it’s gonna happen more informally like that.
Companies don’t want to start and stop. They’re either going to do it and stay with it, or not. We hope that some of them will have both the discipline and transparency to put this stuff out there and hope that people can get a clearer picture on what they’re all about.
Knowledge@Wharton: What additional work needs to be done on this method?
McCarthy: I’d say the paper that we wrote on non-subscription businesses is pretty much a wrap by this point. We feel pretty good about that work. In terms of expanding the scope and applicability of the methods through cutting-edge research, one of the big next steps would be to apply this method to a lot more companies like Overstock, Wayfair, Dish and SiriusXM.
One of the things that would be very interesting would be to explore alternative data sets and see if perhaps those could be leveraged to expand the universe of companies that would be applicable to these methods. I think that’s possible, but I think there would be a lot of very serious methodological challenges that would have to be tackled along the way. That could be the other potential way that we could get a lot more companies to disclose these metrics on their own.
As Pete was saying, one of them would just be that the good companies start disclosing, and it makes the bad companies feel like they need to disclose or the investors are going to call them out. But the other one would be that the investors are getting this information through a proxy that’s good enough, so you might as well disclose the actual data yourself because they have the answer already. I think this sort of work could really help move us in that direction.