The proliferation of DTC, SaaS, and subscription-based companies has had many venture capitalists and investors to call for a better valuation strategy. Traditional DCF models have often required investors to take leaps of faith over blindspots that, according to the Retina team, can be avoided.
We spoke with Retina.AI Founders Michael Greenberg (CEO) and Brad Ito (CTO) about how customer lifetime value (CLV) can be used to provide a much more accurate representation of a company’s valuation than historical metrics such as a DCF.
Retina urges its users to “Focus on the right customer, not the “right now” customer” by developing an incredibly intimate understanding of total predictive customer lifetime value, or the amount of money a customer will spend over the course of their lifetime.
However, it’s worth noting that this isn’t your simple one-size-fits-all CLV calculation. Retina’s machine learning model combines recency, frequency, and magnitude (RFM), as well as churn rates to map out accurate customer journeys.
What inspired the mission to use CLV for company valuations?
Michael Greenberg: “Our idea for CLV as the best metric for a company valuation was built on the findings of Peter Fader (Wharton) and Dan McCarthy (Emory). They came up with these “Buy Till You Die” (BTYD) models that were a good proxy for customer lifetime volume.
In the current state of DTC companies, people have a tough time going through classic transaction metrics such as RoAS or first-order value and extrapolating some sort of company value. They use these metrics for projects such as budget planning and marketing spend planning, but not really for valuations.
The other thing that came out of these BTYD models was the concept of churn. Churn looked at product performance and points of customer drop off, where people would cancel subscriptions. It helps show whether a product is performing at a high level, but only at the cohort or average level. It’s not good for the predictive level and didn’t work for the individual customer level.
Peter Fader and his team put these models together that were a reflection of Recency, Frequency, and Magnitude (RFM) and made a few breakthroughs.
A company was actually built out of this called Zodiac (acquired by Nike before its A-round). What Zodiac did is go through a company’s database and effectively score their customers with a CLV score and churn prediction with 90-95% accuracy 1 year out.
This was the first indication that CLV score could actually be a really good signal for a company’s valuation. It was acquired by Nike internally so they could power their programmatic ad spend and personalization.
We created an automated tool called Retina Go, and an associate at a VC firm started uploading their transaction lists. They said we didn’t have a way to sum the values of all of a company’s customers’ CLVs, but we found that if we did, we could use the sum as a replacement to value a company instead of using a DCF.
If we really hone in on customer lifetime value, which is now possible with exhaustive machine learning models, we can build a much more mature understanding of what a company is really worth.”
What makes CLV better than the incumbent valuation metric, DCF?
Michael Greenberg: “I’ll preface by saying that DCF models work perfectly fine for businesses where you don’t have data prerequisites. Both DCF and CLV have their roles.
For businesses that have customer level transactions with a cadence that’s more than once every 3-5 years, then having things built in like accurate forecasting on what purchases will be starts to become a good replacement for classic DCF.
DCF is built on some error-prone assumptions made on aggregate levels. The first is the future cash flow assumption. By definition, you’re at the problem of averages at a cohort level. If you can go into the customer itself, you can sometimes see 20% to 40% more accuracy from even sophisticated DCF models. That’s just because of the swings in error rates when you go from cohort averages to a customer level.
You can apply a DCF on a customer lifetime value.”
Brad Ito: “If we know what the future cash contribution for each customer, we can in effect do a summation on that and apply a discount rate accordingly. We think that’s the technique we’ll see in an automated fashion. We’d be summing CLV values across customers, to go from an individual level up to more robust KPI metrics.”
Michael Greenberg: “Retina does the entire CLV scoring automatically so they receive what their CLV will be in the form of revenue. These can be sequenced out in yearly periods for the purpose of valuations, predicted churn rates to figure out which one can be applied, and finally, can get readouts on the overall revenue distribution for that customer base.
They have to apply their own summation, discount rate, and multiple– each firm can be as aggressive or conservative on that top line multiple.”
Brad Ito: “You do need to combine the value of the customers you have with an estimation of how many customers you can acquire.”
Where does CLV work best?
Michael Greenberg: “At a high level, consumer, service-based such as insurance and financial at the retail level (if there’s a consumer on the other side), and long-tail B2B.
If there’s some frequency of purchases and you know the customer, that’s where this model works best.”
Brad Ito: “We need to look at what data you need in order to predict per-customer CLV. For each customer, which purchases did they make, when, how many dollars. For some companies, identifying their customers can be a challenge. If they’re a wholesaler selling in bulk, and you get a purchase for 100 units and that’s the only information you have to work with, that’s not good enough.
How do you take into account the overall size of the market?
Michael Greenberg: “From a benchmarking standpoint, you can look at comps within market size.
For example, if its a fashion business, but non-subscription, you can look at the growth rate and churn rate YoY, and payback window.
A second thing to look at is who’s in their competitive set and benchmarks for that set.
A third one many VCs miss is the age and stage of the company. Are they still in their early adoption window? CLVs of early adoption-phase are usually much higher than middle-adoption phase customers. It isn’t until they hit some critical mass of 10-20k customers until you see the true market adoption CLV.
This should be a deep conversation in the VC space, especially if you’re investing in the seed and A stage company. What’s the right discount you apply at the current growth phase of the startup?”
How about public valuations?
Michael Greenberg: “I think we’re going to move into a world where a lot of hedge funds start using CLV. One of the main metrics for valuing and setting a call price, for example, is a DCF and a few other methodologies. Now, we’re starting to see CLVs in action.
Dan McCarthy, for example, has had some great calls for companies that are underperforming based on their CLV. By studying the data in Blue Apron’s actual public releases (churn rates, customer CLV rates), McCarthy saw that Blue Apron was overpriced.
As VCs have shepard more mature companies into public markets, they can more accurately predict the company’s true valuation over the longer-term before getting hit by a rude awakening.”
***Editor’s Note: Blue Apron snagged $135M in funding at a $2B valuation in June 2015. In 2019, its market cap sank below $100M. ***
Ultimately, when used correctly, CLV provides anyone from startup founders to hedge fund operators an extremely precise way of understanding how much a given company or product line is worth.
Machine learning also enables us to understand our customers at a much more granular level and create better predictive models that highlight the true value of a company, preventing the significant overvaluation that only tends to be revealed in the long-term and within the dramatics of a public market.
CLV can be applied to find a company’s best customers and find more of them, and increase value along that journey– but the buck doesn’t stop there. As we’ll uncover in part two, CLV can also be used to identify new product launch opportunities as well as to increase confidence in M&A transactions.