‘Business Modeling’ is mocking-up a business.
The goal is to model it, test it, then scale it.
Unit Economics is at the absolute heart of this process.
Here we look at Unit Economics from a strategy perspective, with real data from Farfetch to help illustrate the real-world application of the concepts.
The below post is a very simplistic, illustrative example of how the topic works conceptually. It is meant for educational purposes and is not specific to any business or brand in-particular.
ARPU – CAC- LTV User Economics
When discussing a business and its past performance vs. future expectations, we traditionally hear the majority of the narrative centred around revenue, growth, and profitability.
These ‘Income Statement Metrics’ do create a lot of context around a business, but they do not illustrate marginal performance or give any granularity in relation to how or when a given business could execute on a strategy to dramatically increase its margins, market share and potential to become a dominant player in a given industry.
Farfetch is one of the best examples of that. The eCommerce luxury-fashion platform has vaulted to become the dominant platform in the industry with a market cap that recently reached more than $25 Billion earlier this year, and an industry-leading Take Rate (28.8% in Q4 2020) in one of the most competitive spaces in the world.
Online luxury fashion retail marketplace Farfetch IPO’d today at a valuation of $5.7 billion, selling shares for $20 each. Taken at face value, this is a relatively steep valuation – it would be selling for ~12x revenues despite a -25% EBITDA margin over the past twelve months.
Theta CLV – Farfetch IPO
Theta Equity’s CBCV (Corporate-Based Customer Valuation) formula dove in beneath the headline metrics (forward Revenues and EBITDA multiples) to understand the mechanics of the business model and what that meant for future expectations of growth and market penetration.
Farfetch’s rise in the past several months (which include big, global partnerships) was at least partially predicted by Theta’s analysis, which showed a majority of customers were both loyal to the brand and making repeat purchases on a quarterly basis (in aggregate) as measured by LTV.
These types of underlying trends have not manifested in profitability until this recent quarter, yet the stock has increased in value by 5X and become a leader in the space in the time period since the IPO (~30 months).
Good news for Farfetch: 12 years after launching and more than two years after its IPO, the luxury e-commerce site hit EBITDA profitability for the first time
Vogue Business
This is why we break down ARPU, CAC, and LTV from a business modelling perspective and use Farfetch’s own data to illustrate the application in the context of a business at scale.
Average Revenue Per User (ARPU)
This is not a GAAP Accounting measure, so there is no official standardized way to calculate it.
Average Revenue per User (ARPU) is a computation of the Average Revenue (whether Gross, Net or ‘Adjusted’ is dependent on the business) earned on a customer over a certain period of time. Typically that period of time is at least one year, but it can also extend into a ‘lifetime view’ of 3 – 5 years (on average, per customer).
The net result of this lifetime view helps to create a number where loyalty – and therefore recurring revenue – becomes extremely important.
A good way to help alleviate some of the calculation complexity for ARPU is break down customers into certain ‘segments’ based on either years (as shown below) or certain profiles based on behavioural or geographic characteristics.
In simple terms, let’s say a business earns $10,000 in Revenue in a Quarter from 500 Customers, ARPU is $200 ($10,000 / 500 = $200). Let’s say in the next Quarter ARPU is $225, then $250 in the next Quarter, then $275 in in the final Quarter of that Fiscal Year (FY). In this case, you would say that ARPU is $237.50 (($200 + $225 + $250 + $275)/4 = $237.50) for the year.
Farfetch Example
GMV (Gross Marketplace Value) was graphed by Farfetch on an annual basis and then expanded on by Theta for the period between 2012 and 2017. We can see from the extension of the bands that a % of customers from each Cohort continued purchasing clothes on Farfetch several years into the future.
Note: GMV is a calculation of the total volume purchase on Farfetch, it is not Revenue.
To get Revenue (and thus be able to estimate ARPU)* we look at the “Take Rate,” which was 32.9% in 2017 (and 28.8% in Q4 2020). This means that on a GMV transaction of $1,000 Farfetch would make $329 in 2017 and $288 in Q4 2020 in Platform Revenue (or Adjusted Revenue).
*As we are dealing with non-GAAP metrics where some terminology is different from one company to another, we try to infer what the ARPU is from this data for illustration purposes
Because Farfetch does also offer ‘Platform Fulfilment Revenue’ (or Fulfillment) to a certain % of brands, which they refer to in their F-1 as a “pass-through cost,” they actually net this out of their calculation in order to get ‘Adjusted Revenue.’
They also have 1st Party Revenue which is related to inventory they sell through from their own retail channel, which is not commission based. Using 2017 data, the Take Rate is derived from the following to get 33.1% (Adjusted Platform Revenue $294.4M / $894.4M Platform GMV = 33.1%).
In 2017, Farfetch stated Adjusted Revenue was $311,784,000 ($311.8 Million) for 936,000 Active Consumers. This would give a ballpark ARPU of $333 ($311.8 M/936K =$333) for 2017. To put that in context of AOV (Average Order Value), Farfetch’s AOV in 2017 was $622.10 per order made.
This would equate to roughly $206 ($622.10 x 33.1% = $206.16) in revenue per order. That would mean that the average customer would be ordering 1.6X ($333 / $206) per year.
In reality, we know that a certain % of customers will make only 1 purchase and never come back, while another % will come back and make multiple purchases every quarter for many years into the future.
Customer Acquisition Cost (CAC)
This is not a GAAP Accounting measure, so there is no official standardized way to calculate it.
Total ‘Demand Generation Expense‘ including the cost of Ads (Digital or otherwise), the salaries to pay the Marketing/Growth Team, and any payments to external agencies for PR, etc.
It is generally associated with New Customer Acquisition, not marketing/branding costs that are related to current customers.
In simple terms, let’s imagine that a business spends $6,000 on New Customer Acquisition in a Quarter and acquires 500 New Customers. CAC would be ($6,000/500 =$100) $120 for that Quarter. Maybe the next Quarter it is $125, then $130 in the next quarter, then $135 in the final Quarter of that Fiscal Year (FY). In this case, you would say that CAC is $127.50 (($120 + $125 + $130 + $135)/4 = $127.50) for the year.
Farfetch Example
Between 2015 and H1 2018, Farfetch’s CAC actually trended downwards. This was a heartening (albeit unusual) sign because normally CAC trends upwards as markets become more competitive.
Furthermore, many companies will ramp-up marketing/demand-generation spend ahead of big fundraising events in order to improve headlines metrics for investors such as Revenues, Growth Rates, etc. even if they are losing money on those customers.
We can see that Farfetch’s CAC was $100 in the years 2015 – 2017. At the time of the IPO in Q3 2018, Farfetch had acquired 2.3M customers in total.
As of June 30, 2018, the Farfetch Marketplace connected over 2.3 million Marketplace consumers in 190 countries to over 980 luxury sellers
Farfetch F1
Simple back-of-the-napkin math shows that they would have spent about ($100 x 2,300,000 = $230 Million) on Customer Acquisition over this time period. Note: this does not mean that is how many active customers they have, CAC is a function of total money spent on demand generation / total customers acquired over that same time period.
If you look at their Q4 2020 numbers for additional context, they acquired 500,000 Customers and spent $67 Million on ‘Demand-Generation Expense’ over that same time period. That gives us an inferred CAC of $134 per customer.
It’s been — 2020 has been really interesting and incredible year in terms of the new customer acquisition. We’ve acquired nearly 2 million new customers, quarter 4 was 500,000 new customers.
Farfetch Q4 ’20 Earnings Transcript
This shows us that we can expect CAC to trend-up over the long-term as industries become more competitive. This can have significant impacts on strategy depending on a company’s view on the respective S-Curve for their given market relative to their respective Unit Economics.
Lifetime Value (LTV)
This is not a GAAP Accounting measure, so there is no official standardized way to calculate it.
Lifetime Value (LTV) is a function of the ARPU and CAC. If you spend X on CAC and net back Y on ARPU, the resulting Z gives you a ratio on LTV as a rough ROI on CAC investment over a period of time after you factor in Churn (% of people who ‘unsubscribed’ or made only 1 purchase).
The tricky part of estimating CAC for non-SAAS based businesses is estimating Churn. In the case of eCommerce, Churn would be customers that only make 1 purchase, which will end up being a very high % of customers overall. If we use the 2017 data from Farfetch/Theta’s model above as an example, 55.6% of Customers were Existing Customers and 44.4% were New Customers.
In simple terms, let’s imagine that the business we have referenced above has a ARPU of $237.50 for a given year. We need to be careful in each specific context whether we are speaking of Gross, Net or Adjusted Revenue as we saw with Farfetch, as the specific definition can dramatically change the LTV calculation as we will see below. From that same business, we said that CAC is $127.50 for that same year.
If you simply took ($237.50 – $127.50 =$110) you get a number of $110 that does factor in any expected repeat purchasing behaviours of the customer base beyond that year.
At the most basic level, if 40% of the customers for this business leave in that year post-acquisition (ie. one purchase only), you would expect this formula to yield (1/40% = 2.5), meaning that the LTV would be 4.65X* ($237.50 x 2.5) / $127.5 = 4.65X) which is very good.
*The nature of this calculation is much more complex due to the issue of decay beyond the first year from a statistical perspective.
This result implies, however, that all of the 60% of ‘Existing Customers’ in future years remain with the business and continue purchasing frequently over time, which in reality is not the case. A % of them will Churn as well, but not necessarily 20%, and their ARPU will also likely increase or decrease over time as well depending on the business. But to illustrate how to deal with this complexity is too much for this post.
If you want to calculate churn by cohorts instead, then you need a model of how the cohort decays.
StackExchange
The main lesson to pull away from the addition of Churn to the formula is that the more repeat purchases from existing customers over time, the better the marginal profitability for the business on a per customer basis.
Farfetch Unit Economics
Farfetch’s calculation of LTV (via both Theta and the company themselves) will be much different than a SAAS-based subscription model, but we will break down the usage of CLV (Customer Lifetime Value)/LTV in the Farfetch-specific context below.
Farfetch themselves define LTV in the following way:
LTV means cumulative Platform Order Contribution, calculated as gross profit less demand generation expense, excluding demand generation expense attributable to any new consumer acquisition, over a period of time attributable to a particular consumer cohort since the acquisition of those consumers divided by the number of consumers acquired during the cohort period. Each consumer cohort is defined as consumers who have been acquired during a specific period.
Farfetch F1
In effect, the nuance in this definition means that they index Churn by ‘excluding demand generation expense attributable to any new customer acquisition’ meaning that they are looking at the performance of each Cohort over an extended period of time.
Because New Consumers spend less than Existing Customers, we get a blended ‘Marketplace Order Contribution’ that is an average between the two on a yearly basis. We can see how the % Adjusted Marketplace Revenue number trends up from 33% to 44% from 2015 to 2017, effectively demonstrating the effect of loyal customers on their ARPU, or as they would say Order Contribution Margin. Theta themselves noted this in their own analysis.
We infer that customers who maintain their relationships with Farfetch spend about 20% more per year.
Our most conservative variable profitability assumptions imply Farfetch customers generate approximately $500 in marginal profits after they are acquired, in net present value terms.
Theta CLV – Farfetch IPO
The calculation that Farfetch themselves broke out for LTV is shown below.
In determining how successful our consumer acquisition and retention strategy is, we closely monitor the initial Consumer Acquisition Cost (“CAC”), the Lifetime Value of a Consumer (“LTV”) and Platform Order Contribution Margin. These performance indicators enable us to assess the strength of the short-term and long-term consumer unit economics.
Platform Order Contribution Margin means Platform Order Contribution as a percentage of Adjusted Platform Revenue during a particular financial period.
Farfetch F1
We talked about the meaning of ‘Adjusted Platform Revenue’ in the Farfetch portion of ARPU above, and ‘Platform Order Contribution Margin’ in the section. Adjusted Platform Revenue was $311.8 Million in 2017, while Platform Order Contribution Margin was 44%, which would equal $137.2 Million ($311.8 x 44% = $137.2 M) if multiplied together. The actual Platform Order Contribution of $127.4 Million was reported for 2017, so while close, there is still some discrepancy in the exact math.
As you can see in the graphic above, they measure the LTV/CAC ratio after 6 months as 1.72X for the 2017 Cohort. The company spent $69.2 Million in ‘Demand Generation Expense’ in 2017. While not an exact match because of the differing time periods, we can see that taking Platform Order Contribution/CAC for 2017 gives us 1.84 ($127.4 M/$69.2 M = 1.84), which gets us in the range of Farfetch’s actual LTV calculation for its 1st 2017 Cohort.
If we extrapolate other Cohort data and apply it to Farfetch’s 2017 Cohort, we would expect it to have crossed the 3X CLV/LTV ratio within 24 months. This 3X or 300% is generally seen as the threshold to indicate an excellent ROI (Return on Investment) on CAC investment.
Even today, Farfetch continues to break out Cohort LTV data to measure repeat purchasing and consumer behaviour in relation to CAC, as seen from the most recent Earnings Transcript as articulated by the company’s CEO.
We now have cohort data from Q2 and it’s incredible data. So the retention is very strong, we have lifetime value for those customers, which are higher than the previous 10 quarters. Those customers — the cost of acquisition of those Q2 customers has been paid back in less than six months therefore, and the Q3 cohort of customers is, as I said, showing very strong repurchase behavior, spend per customer behavior.
Farfetch Q4 ’20 Earnings Transcript
Impact on Modelling
When you are looking at a business from the top down, you look at numbers such as Revenue, Growth Rates, and the opportunity on the horizon long-term such as TAM (Total Addressable Market).
In contrast, the combined data from ARPU, CAC, LTV and their derivatives help create a picture from the bottom-up perspective. This creates a view on how each customer contributes to profitability on a marginal basis over time.
This creates models where the central focus is strategy as opposed to accounting or finance. In accounting, the goal is typically to create retrospective statements (Income Statement, Balance Sheet, etc) for both compliance and analytical purposes.
In finance, the goal is to typically to create forward-looking models that focus on metrics such as IRR (Internal Rate of Return), projected EBITDA, etc. for the purposes of determining whether or not a given company merits an investment or not.
When looking at this type of data from a strategy perspective, however, we are analyzing the cashflow of the business and whether not certain actions (ie. Customer Acquisition strategies, Partnership strategies, etc) should be undertaken to improve cashflow metrics on a monthly/quarterly/annual basis. Improved cashflow opens opportunities to make moves in the market independent of investment.
The analysis to reach these conclusions, however, is complex and sits outside of what would be considered GAAP (Generally Accepted Accounting Principles).
This is why the concept of business modelling can be valuable, as it helps to frame the analysis in a strategic way to unlock insights in real-time.
Naturally this type of data can be further leveraged to raise professional investment, but the actual synthesis of that data and modelling requires much more advanced modelling as we have seen in this post.