The ultimate guide to VC’s LTV/CAC analysis — Pt 2
Customer Value and Lifespan: The Lifetime Value (LTV)
“How did you calculate your Customer Lifespan?”
In the previous article I introduced the differences between a “normal” LTV/CAC (e.g. for an e-commerce or SAAS business) and the same analysis for a marketplace business. What makes things complicated is that a marketplace (e.g. Uber) needs to invest to acquire demand AND supply, running into double Customer Acquisition Costs. But I will cover the CAC’s in my last piece.
In this article I will focus on the Lifetime Value (LTV — or referred to Customer Lifetime Value, CLV). The calculations, for as simple as they might seem, are quite tricky and require a good level of abstract thinking. My goal is to make such process more transparent and, if possible, leave out common mistakes and/or misconception that will inevitably poison your analysis (and give you bad surprises when dealing with more experienced VCs). No approach is flawless: assumptions are necessary. The key here is to rely on the most defensible aka logical set of assumptions.
LTV — Demand or the Supply? The dilemma is served
The first element you need for the LTV is the:
a) AVG revenues per customer (during a timeframe — net of variable costs)
Here’s the first dilemma: how would you identify your “customer” in a marketplace model?
In other terms, who is your main stakeholder? Is the buyer of the service/goods or the provider? Keep in mind: focusing on either of the two implies making an assumption on the other.
For instance, if you were to focus on the demand side, you’d have to account for the supply side by assuming an AVG required number of suppliers to fulfil the purchase. Imagine you have 100 customers (let’s ignore their recurring purchases for a second) and that each of them buys 3 products in a year. My suggestion would be to estimate how many suppliers were required to fulfil such demand. Say the 300 (3 X 100) products sold were fulfilled by 50 different suppliers: there you go the magic number required to estimate your Supply CAC. We will look into this in the next article. Now let’s focus on where to start from.
My general rule of thumbs is to focus on the side which has the least predictable lifetime pattern. Usually, this is represented by the demand. The main factors to look for are a) the switching costs or b) the business model.
Let’s look at some examples.
Uber: I can fairly assume that drivers (the supply) tend to have a high retention aka lifespan because of the high costs associated with becoming a driver. It will take years before recouping from the investment. In other terms, the switching costs to, for instance, another career, are quite high. It’s reasonable to assume a low churn rate.
AirBnB: same considerations. The platform doesn’t require any high investments for the supply, true, but it doesn’t imply fixed costs either. The costs on the suppliers are connected to the volume of sales in the form of commissions... so nobody gets hurt. Namely, why would a supplier churn if just being on the platform at least creates the chance of making more revenues? In addition, the liquidity on the platform and the revenues potential is so high that it doesn’t make any economical sense to NOT be on there. Long story short: high lifespan.
Amazon: same considerations, although for different reasons. Amazon represents an hybrid marketplace because it’s marketplace + SAAS. Amazon does have fixed costs (subscription) next to purely variable ones, which might lead suppliers to churn. Nonetheless, Amazon is not just a marketplace connecting a supplier with customers: it’s a full suite of SAAS services, which enable suppliers to even serve their own online demand. This justifies the subscription price. All combined, this also leads to very high (expected) retention aka lifespan.
(note: in my view the marketplace + SAAS represents the natural evolution of the marketplace model. Offering matchmaking is hardly a sustainable competitive advantage, and requires constant investments. In addition, the SAAS model represents a way for an established player to further monetize).
There might be cases where the demand is more predictable then the supply: some services/goods are by nature non-recurring (or unlikely to be recurring). For instance:
- Wedding Planning (...at least it’s meant to be “forever”…)
- Real-estate investments (e.g. buying a house)
- Relocation services
- Plumbing, house restructuring etc etc.
- Cars and other vehicles
- Collectibles and high-value items
The list could go on an on. You might think these are bad businesses, but you’d be very wrong (at least, on this sole premise). Truth is, one-off purchases usually imply high basket values (i.e. value of the purchase). This should be pretty clear from the example above, although I didn’t mention any companies in these fields (it wouldn’t add much to the discussion as only few readers might know the companies). As long as the take-rate is high enough, the model can be highly profitable. It’s also true that for these businesses usually require a higher diversification level before they can reach scale. E.g. a a collectible marketplace will most likely offer all kind of collectibles, from watches to paintings.
To determine whether you should focus on the demand or the supply side you should think through which of the two is less predictable and requires more investments to boost retention. That’s why industry domain expertise is key here.
Once you have identified your key stakeholder, then set a timeframe (e.g. 12 months) and determine the AVG revenues per unique customers/provider (it removes the effect of the recurring purchases), net of the variable costs associated with the transaction (e.g. payment processing costs are a very good example).
Lifespan
I hate to repeat myself, and I won’t start now. For more info on the two broadest approaches to calculate the lifespan, refer to the previous article. Here I will look at the so-called forward-looking lifespan (my definition — all rights reserved :) ).
The general approach is to use 1/churn rate, where the churn rate is considered to be the average among the customers (or stakeholders, as identified before).
Here things get analytical. For the sake of this article, I will only briefly mention the two possible approaches to calculate the churn rate, and focus on the most “mainstream” one. The most common one focuses on calculating the churn rate in, e.g. this month, compared to the month before. It’s like saying, if last month we had 50 customers, how many are still customers this month? The equation will look something like:
Churn Rate = Active(T1) / Active (T0)
Of course, for a complete analysis this needs to be repeated over multiple periods of time.
The other approach, more complicated, would imply calculating the CR on a cohort-basis. A cohort can be defined on the period in which certain customers were first acquired. Say 100 customers are acquired in January, the question is how many churned in February, March, April… The denominator of the previous equation is always 100, regardless of the month of analysis, whereas in the previous case the denominator is always the active in the previous period.
Lifespan: simple average vs. weighted average
If we follow the common approach where CR = active(T1) / active (T0), the most important thing is to set the timespan of the observation period. That really depends on the behavioural aspects of your industry. It can range from weekly to yearly one. Generally speaking, I think the monthly and the quarterly ones are the most common. What is important is that there are sufficient observations to make an average: 1 or 2 periods won’t cut it.
Once there are sufficient time-series CR’s, the best way to represent their evolution is by means of an average and, specifically, weighted average. A weighted average better represents the relative importance of each time period and avoids data biases. The weights should be defined by the number of customers in each period (which is, in turn, the sum of “new” + “retained” customers). In a startup company, the latter periods will (hopefully) have a larger customer base than the early ones, and this should be reflected in the overall churn rate.
This table below should exemplify that.
In the fictional example above, you can see how customers are divided by the month of first purchase (y-axis) and by the month in where their purchase(s) took place.
Based on that, the CR is calculated each month based on the equation highlighted before. In M2, we have 8 retained customers and 5 churned ones. The CR is therefore: 5/13 = 38.46%. And so on and so forth.
At the bottom of the image, you can see the different average CR’s. If we were to take only the last-period CR, we might run in the trap of artificially selecting a particularly good month, forgetting about all the previous data. If we use a simple average, the effect of the initial CR’s are too high, and it doesn’t correctly reflect the growth in customers the company experienced. This is why the weighted average falls right in between, and it’s the most sound/defensible of the three.
Put it all together
Once the average CR has been calculated, the implied lifespan is determined as 1/CR. In the case of our fictional company, the lifespan is
1/.26 = 3.84 months
Note: the lifespan has the same time value as the period of analysis. Being fictional data, there is no point in discussing the results. I will only say that, if this was a real case, then I would conclude the monthly analysis doesn’t really suit the type of business, and I would extend the range from monthly to quarterly analysis.
Finally, the LTV can be calculated as AVG revenues per customer * Lifespan. Les jeux sont faits!
(note: the first stress test a VC will run on your LTV is by changing your lifespan to 1 and 3, which should yield a LTV/CAC ratio of >1 and >3 respectively)
Stay tuned for the final part of this mini-series: the Customer Acquisition Costs!
If you found this article useful, don’t be selfish and share it with other entrepreneurs out there!
Questions? Comment here below.
