Wharton's Peter Fader and Dan McCarthy discuss their research on non-contractual firm valuation using customer metrics.

When a business has a steady customer base, it’s easy for it to make estimations and projections. But that task is very difficult for companies that are non-contractual, meaning they have customers with inconsistent buying patterns. Wharton marketing professor Peter Fader and Wharton doctoral student Dan McCarthy are looking to close the data gap in their new research paper titled, “Valuing Non-Contractual Firms Using Common Customer Metrics.” They recently spoke with Knowledge at Wharton about the findings of their study.

An edited transcript of the conversation follows.

Knowledge at Wharton: Could you give us a summary of your research? Is this a follow-up of previous research where you looked at the valuation of companies that have customers who buy subscriptions?

Peter Fader: This idea of customer-based corporate valuation is a concept that’s been floating around, but Dan and I are taking it very, very seriously and want to bring a tremendous amount of rigor and breadth to it. In the previous work, we looked purely on the contractual or the subscription-based side. If a company knows when a customer is leaving, it’s easy to project what the rest of their life is going to be, and the payments tend to be steady. But most businesses are what we call non-contractual, which is you’re just buying things on occasion. This is the kind of deal you have with a retailer or a travel firm, or pharmaceuticals or media consumption. Most businesses are non-contractual, and that makes it much more difficult to understand who’s doing what and to project it into the future.

A big part of what we’re doing here is trying to take this broader concept of customer-based corporate valuation and make it just as palatable in the non-contractual setting as it has been in the contractual, subscription-based one.

Knowledge at Wharton: Dan, do you have anything to add?

Dan McCarthy: I’d say one of the key challenges in this work is [determining] the data that we need to be able to perform this estimation and then project forward how many purchases customers are going to make and how much they’re going to spend on each of those purchases. We want these methodologies to be as broadly applicable as possible. The vantage point of this work is someone who’s on the outside looking in. Maybe it’s a shareholder, a hedge fund, private equity firm, any number of financial institutions that are looking at the public markets and asking, “Do I want to buy this company? Do I want to sell this company?” Those people don’t have access to the internal transaction logs of the company. What they have access to is usually the information that a company would put in its quarterly or its annual filings. We went at this by saying, imagine that a company were to disclose a certain, very small set of common customer metrics. What are the ones that would really enable us to be able to perform this first estimation procedure for the various model components, and use those to project forward what total revenues are going to be?

“Most businesses are non-contractual, and that makes it much more difficult to understand who’s doing what and to project it into the future.”–Peter Fader

Knowledge at Wharton:  How did you figure out what those were? And once you did, what did you do with them?

McCarthy: We figured it out by doing a large-scale simulation analysis. Step one was saying, this is a set of metrics that we think could be usable, and they need to satisfy two main criteria. The first is that they’re actually used by companies, and the second is that they help inform our model. The way that we answered that was to say: Imagine that I will generate data from all these different possible types of worlds and just let these metrics kind of duke it out. We took every possible collection of this set and said: Imagine that I only get to see those metrics. Let me try to predict the future and see which combinations of metrics percolate to the top.

Fader: I think it makes sense to talk about the metrics themselves. We’ve done a lot of scraping of financial statements, listening to CEO conference calls, seeing what third-party firms are saying about the size and nature of customer bases, and we came up with six metrics. The most common one that a lot of companies will report is what we call active users. How many people have made a transaction, used our product or service sometime within the trailing 12 months? No. 2 is what we call heavy active users. How many people have made a repeat purchase, have engaged with us at least twice over that trailing 12 months? No. 3 would be this idea of a forward repeat rate. Of all the people who made a transaction with us back in 2015, how many came back and did it again in 2016?

The fourth and fifth would be the flip-side of that. With all the purchases that we had today, what percent of them are from customers who did something with us in the previous year? Of all the customers who bought with us, what percent were with us previously? Or of all the orders that were placed with us this year, what percent of them are by customers who have bought previously?

Last and least reported is the idea of frequency. Of all the customers who have done anything with us in the past year, how many things did they do? How many purchases did they make or sell on? So, that’s the big six. We went into it without a really strong sense of which one or ones would win out, but it was really interesting to do this grand bake-off across lots of different worlds and see there were some very strong, consistent results.

Knowledge at Wharton: When you did this bake-off, what was the winning metric? What are some of the conclusions that you were able to make?

Fader: Of course, it’s always easy in hindsight. It turns out that the No. 1 metric is the least commonly reported one that I mentioned, which is frequency. If you can tell us among the people who did anything, how many things did they do on average? By itself, any one metric isn’t going to get you far. But if you take that frequency metric and you combine it with any of the other five, it doesn’t even matter. Frequency plus anything gives you really good predictive capability. Just winning by a nose among the five was the first one I mentioned — active users. If you can say, here’s the number of people who have done anything with us and the average number of things that they’ve done, you take that one-two punch. And it’s remarkable how well you can uncover future purchases.

McCarthy: One thing that was a bit surprising was the fact that when we went from two metrics to three, there was virtually no improvement in our ability to predict future purchases. The bulk of the benefit that we got was by moving from the best single metric to the best pair. Almost equally interesting was, you can’t have too much of a good thing. We kind of kept that process going: What’s the best quartet? What’s the best combination of five? When we went from the best combination of five to the best combination of six, performance got slightly worse and a bit more variable from scenario to scenario. So, sometimes less is better.

Knowledge at Wharton: What’s the best way to apply your findings?

McCarthy: I would just demand that companies disclose these metrics — I think emphasizing [that these metrics are very important] both to the company itself and perhaps even to regulators, whose responsibility it is to ensure that investors have the information they need to make investment decisions…. It would be wonderful to speak up and get people to disclose the metrics.

On the back end, we have this fairly complex estimation technique that’s needed to take that scattered bunch of customer metrics and map it down to the underlying parameters for how customers are acquired over time, how many purchases they make and the spend associated with each of the purchases. There is some math involved and some heavier duty computation than we saw in the subscription-based paper, but it is certainly quite doable. I think as these methodologies take hold, we’ll see them be made more readily available.

“One thing that was a bit surprising was the fact that when we went from two metrics to three, there was virtually no improvement in our ability to predict future purchases.”–Dan McCarthy

Knowledge at Wharton: Do companies have not to release this data? Is there a reason that they wouldn’t want to?

Fader: First, let’s talk about some of the other benefits, then we can talk about some of those strategic aspects as well. Obviously, the motivation here is customer-based corporate valuation, so the investor can make more informed decisions about the current and future health of a company’s customer base and, therefore, the value of the enterprise. But it can be useful in other ways. This can be a great source of competitive intelligence. Given that these metrics are not that hard to get from a third-party firm or a company’s first-party disclosures, you can start to see rival companies playing this game about other competitors in the same sector to understand where they stand, not just in terms of overall sales but in terms of the nature of the customer base. We think there will be quite an imperative to be reporting these things, not only out of some kind of fiduciary responsibility but for some of these competitive activities as well.

That goes directly to the other part of your question — this idea of should there be some sense of when you disclose which kinds of metrics. Right now, it seems kind of haphazard. There’s some that disclose some metrics all the time and others that never do it. There seems to be no logic as to why certain companies do or don’t and which ones they disclose. There are some companies that just boast about disclosing metrics, maybe because it makes them look more rigorous or science-y. But no one ever looks at these things and can make heads or tails out of them or map them back to actual or projected future revenues.

We want to see much more smarts about who discloses what, when. Maybe companies will then start to say, “Are they revealing these metrics now because they are trying to signal something? Or maybe they’re going to stop revealing them because they’re trying to hide something?” There’s going to be this whole chess match about who discloses what, when. That will be kind of interesting when we get to that point. Right now, it’s just so early on that if companies are doing it at all, it seems like more of the CEO’s ego as much as any kind of real information value.

Knowledge at Wharton: Do you feel like people are going to become more interested in doing this, and there will be more of this realization that maybe you’re not saying everything there is to say about a company by just disclosing financials?

Fader: That is right. Many companies go to Wall Street on a quarterly basis, and Wall Street’s looking at them, saying, “Your earnings aren’t up to snuff.” But they say, “You know what? We’re investing in the customers.” Well, so far it’s just been a “trust me.” But now we can provide definitive proof. Now investors can say, “If that’s really true, then show us these two metrics so that we can make that assessment for ourselves.” I want the companies to be more forthright about it. I want investors to demand it. I want competitors to be kind of curious – [and] regulators, as Dan said. I just want a more active conversation about this concept of customer-based corporate valuation. I think it’s in everyone’s best interest. Of course, I’m a marketing professor. I think that if we can establish that there is something here, then it will also trickle down through the rest of the organization and start to impact other kinds of decisions that the company is going to be making.

McCarthy: It’s a very legitimate concern that some companies would have that as soon as they start disclosing these metrics, there is some sort of negative repercussion from that. There has been some very interesting recent work that was just put out in one of the top marketing journals by a colleague of ours who came to the conclusion that when companies disclose forward-looking metrics, it really helps investors make more accurate projections of revenues, thus lowering their perception of the uncertainty about what’s to come in the future. And that can help increase the valuation of firms. All else being equal, disclosing forward-looking metrics can be beneficial to the stock price, so it can help overcome some of those concerns the companies might have.

“It’s remarkable how well you can uncover future purchases.”–Peter Fader

Knowledge at Wharton: Do you feel like that there are other misperceptions that this research might dispel?

McCarthy: For one, just the competitive concerns. Yes, we’re showing that these metrics can be very helpful for projecting what will happen in the future. I’ve mentioned the one about how you can have too much of a good thing, but I would say another spin on that would be just the overwhelming influence of quality over quantity. It’s extremely important to have the right combination of metrics, and oftentimes, having a very small set of very good metrics can be much more beneficial than having kind of the wrong set of very numerous metrics. In that large-scale simulation analysis, if you gave us the right set of two metrics, it could do much better than another set of five metrics. That kind of surprised me.

In terms of misconceptions, I would say the final one that comes to mind is, up until this point, people have been using customer metrics as kind of a dashboard. Every month or week, the company will have a spreadsheet that is sent out to the organization, and it’s got this long list of all of these different customer metrics and how each of them has gotten better or worse over time. Usually, people are looking at each of these different metrics as kind of in its own world. If it went up, it must be good. If it went down, it must be bad. Let’s figure out why. Here, we’re saying we don’t just need to look at these metrics on a stand-alone basis. They’re not ends unto themselves. We can tie them up and summarize the effect of what happened over the past week or month on the overall evaluation of the firm by putting them all together and weaving them into this integrated model.

Knowledge at Wharton: Pete, do you feel there was the perception out there that maybe companies that don’t operate on a subscription model couldn’t do something like this? Do you think this dispels that?

Fader:  Well, it does dispel that, but actually it was a worse problem — that there were a lot of these non-contractual, non-subscription firms out there that would make metrics up to make it seem like they could use some of these subscription-oriented approaches. For instance, in the subscription world, you have a formal retention rate. You know of all the people who had a subscription with us last period, how many of them renewed. You know whether they did it or not. There is no equivalent metric in the non-contractual world. If I didn’t buy anything from Amazon in the past 12 months, it doesn’t mean that I’m gone as a customer. It just means that I’m a light buyer. What a lot of non-contractual firms were doing is they would take some of these metrics, like the forward retention from the forward repeat rate, and they said, “You know what? That’s kind of like retention rate. Let’s just treat it that way and then pretend that we’re a subscription company.”

Well, that’s just wrong. Not only is it conceptually incorrect, but if you were to follow it all the way through, and some of the earlier published research showed this, your ability to assess to overall value of the customer base is going to be way, way off. The diagnostics that you get around it would be off as well, which takes me back to the other part of your question.

One of the other benefits of doing all this is to create credibility for the marketing organization. We can say, “Here are the really meaningful metrics, and here’s the way that we can tie them together to make statements about valuation,” instead of having these things being little trophies on a shelf, where the rest of the organization, say the people in the CFO’s office, would say, “That’s nice, let the marketers play with their toys.” If we can show the CFO and other folks through the organization that these metrics can help with day-to-day operational decisions, as well as bona fide, long-run strategy, it’s going to cast the entire marketing organization under a very different light.

“Oftentimes, having a very small set of very good metrics can be much more beneficial than having the wrong set of very numerous metrics.”–Dan McCarthy

To me personally, getting that kind of respect is more important than taking some of these valuation models and trying to find a little bargain here or there, as a hedge fund might do. They’re both interesting. They’re both important. But for me, it’s that common language and respect for marketing that’s number one.

Knowledge at Wharton: What’s next for this research?

McCarthy: In terms of the boxes that have already been checked, we think of it along a few different dimensions. What’s the sort of data that we have, who is performing the exercise, what is the goal and what is the type of firm? Right now, we’ve primarily focused on external stakeholders, like the shareholders, the private equity firms and the hedge funds whose goal is really to make an accurate estimate of how much the value of the firm should be. We’ve done that for both contractual and for non-contractual firms. In terms of where we go from here, I’d say one of the big questions is, can we move beyond measurement? Imagine that I’m inside the company. Is there anything else that I might want to do? Maybe we can manage value, not just measure it. Part and parcel with that will be using internal data, as opposed to external data.

In both the subscription-based valuation paper and a non-subscription-based valuation paper, we were only looking at one company at a time. We didn’t really focus on any competitive effects. I think an interesting area of future work would be to imagine that we had multiple companies in the same industry. Could we learn something that we wouldn’t have learned about if we were looking at each of the companies individually, make better predictions, or just learn about competitive effects? We’re just really scratching the surface and very excited to see where we can go from here.