Nicolas will show us how to analyze marketplaces live on Wednesday, February 19th! RSVP here.
There are many reasons why venture capitalists love marketplace business models, mainly due to their ability to build network effects or marketplace liquidity that can act to increase moats or barriers to entry as marketplaces scale. However, as the resident Headline “numbers guy” who relies heavily on validating product-market fit and unit economics via cohort analysis, I have scratched my head for years about how to properly analyze marketplace businesses. Unlike traditional SaaS or consumer-subscription businesses, marketplaces have two separate sets of customers: the demand-side customers and the supply-side customers, which necessitates analyzing each customer base separately and running two separate analyses. Consequently, there are two separate retention dynamics, customer acquisition costs, and–if you think about marketplace businesses like me–even two variable costs and gross margins.
Let me walk you through my approach.
In simple terms, tech-enabled marketplaces are online platforms that connect buyers and sellers, leveraging technology to facilitate transactions. These marketplaces often provide additional services such as payment processing, logistics, and customer support to enhance the user experience and build trust. Examples of these marketplaces span various industries, such as e-commerce platforms like Amazon and eBay, service platforms like Uber and Airbnb, and freelancing platforms like Upwork and Fiverr.
To run any business through a bottoms-up SaaS metric analysis, I require the historical monthly profit and loss (P&L) statement, as well as the historical revenue transactions. From this, I can build revenue and customer cohorts. These cohorts enable me to break down the P&L based on customer behavior. This is similar to marketplace business, but there are two distinct differences:
- For the P&L, one requires the sales and marketing costs (investments to acquire new customers) to be broken out into demand-side and supply-side acquisition spend, meaning one bucket is attributable to acquiring demand-side customers, and one bucket is attributable to acquiring supply-side customers.
- For the revenue or gross merchandise value (GMV, the total value of all goods transacted on the marketplace) transactions, one will require both the demand-side and the supply-side customer IDs associated with each transaction.
See an example of the data request that we at Headline would use for a marketplace business outlined below:
1. Historical monthly P&L with the following four columns broken out:
- Revenue (either net revenue or GMV revenue)
- All payouts/variable costs/cost of goods sold that scale linearly to revenue
- Sales & Marketing, including salaries of all sales and marketing staff attributable to demand acquisition
- Sales & Marketing, including salaries of all sales and marketing staff attributable to supply acquisition
- G&A, all other below-the-line general administrative expenditures, including research and development expenditures
- EBITDA/Net Income
2. CSV with historical transactions with the following columns:
- demand_customer_id, a code for each demand-side marketplace participant that is consistent across all transactions
- supply_customer_id, a code for each supply-side marketplace participant that is consistent across all transactions
- date, the date of the transaction in mm/dd/yyyy format
- transaction_value, the monetary value of the transaction amount (either net or GMV revenue)
- tag1 (optional), in some cases, it is helpful to slice and dice data by additional attributes, essentially any metadata you have about the customer or product.
Note that marketplaces will report their top-line traction on either a GMV basis, which translates to the value of total products/services that are transacted on the marketplace in a given period, or net revenue, which is the portion of revenue that the marketplace or business retains via its take-rate on GMV. In other words, net revenue equals GMV less any payouts to marketplace participants.
Logic to Analyzing a Marketplace Business
Given that we have two types of customer IDs, we can run two sets of cohorts:
- Demand-side cohorts, which are cohorted around the month in which a given demand_customer_id became active for the first time
- Supply-side cohorts, which are cohorted around the month in which a given supply_customer_id became active for the first time.
The net revenue and/or GMV total should be identical between two sets of cohorts, while everything else (size of each individual cohort, retention, unit economics, etc.) will differ. If you are not familiar with how we at Headline think about cohorts, check out this video.
We enrich these cohorts with two separate types of P&L: demand-side P&L and supply-side P&L. We summarize the P&Ls, which consist of:
- Net revenue/GMV (be conscious of what you are using as this needs to match the revenue transaction and will impact what you include in the variable expenses/COGS)
- Cost of Goods Sold (all variable spend that increase in-line with revenue)
- Sales & Marketing/Customer Acquisition Spend (S&M)
- Research and Development (R&D)
- General and Administrative (G&A)
- EBITDA/Net-Income
Four line items of this summarized P&L will be identical across the demand and supply side P&L, which are:
- Net revenue/GMV
- R&D
- G&A
- EBITDA/Net-Income
The two line items that will differ between the two P&Ls are the Variable Costs/COGS and the Sales & Marketing (S&M).
- The S&M is straightforward: if you run the demand-side customer cohorts/IDs, you will run the S&M with all S&M expenditures applicable for acquiring demand-side customers. Conversely, if you run the supply-side customer cohorts/IDs, you will run the S&M with all S&M expenditures applicable for acquiring supply-side customers.
- The variable costs/COGS are more complex and should include the following:
- All marketplace payouts to the supply side, or third party payouts if a dashboard is run with GMV transactions. If run with net revenue transactions, this will be left out
- Payment processing fees can be significant for marketplaces as you get charged ~2% on GMV. Take the example of a marketplace with a 20% take rate on GMV and a 2% payment processing fee. This means that for a $100 transaction, the company will earn a $20 take-rate and pay $2 in transaction fees. The transaction fee will take up 10% of the take rate/margin. Therefore, the formula to get to the true variable cost margin of payment processing on net revenue is:
payment processing fee % x (1 / marketplace take-rate %)
Note that some marketplaces might pass on the payment processing cost to marketplace participants. In that case, you would either capture this in revenue and include its processing costs in COGS or just leave it out of both; garbage in, garbage out. - Customer success costs to facilitate transactions and deal with disputes.
- Fulfillment/shipping costs, which can be applicable for physical marketplaces but may be passed on to marketplace participants
- Storage/warehousing costs of products, which can be applicable to physical goods marketplaces like the RealReal
- The amortized cost of acquiring the other side of the marketplace (the big topic)
The Amortized Cost of Acquiring the Other Side of the Marketplace
To explain this concept, let's start out with a practical example. If you are an artisan store that sells the products of local artisans, you have the cost for you to procure the products to fill your shelves. Each artisan has a limited number of products that they can possibly sell to you. If you want to double your revenue, you will most likely need to double your inventory as a means of working with twice as many suppliers. The cost/effort to procure these products would double then as well. The procurement function, therefore, acts as a variable cost that scales linearly with revenue.
I think about marketplaces as a store with a single line of shelves in the middle and two entrances: one entrance for the demand and one entrance for the supply. Take Uber as an example. Uber, on one side of the store/shelf, has passengers (demand), and on the other side, you have drivers (supply). The demand walks into the store looking at the shelf for a product, which in their mind is a driver with a car that takes them from point A to point B. The supplier walks into the store from the other side, looking at the shelf for a product, which in their mind are passengers to take from point A to point B.
If you now run the numbers on this business from the demand-side, with demand_customer_ids and demand-side cohorts, the cost of stocking the shelf (procuring the products) is the cost of acquiring the supply; in Uber’s case, acquiring the drivers. Conversely, if you run the numbers on this business from the supply-side, with supply_customer_ids and supply-side cohorts, the cost of stocking the shelf (procuring the products) is the cost of acquiring the demand; in Uber’s case, acquiring the passengers.
There is one core difference between the artisan store and Uber (a marketplace), though. When the artisan sells a product, that product is gone, and they need to restock it. In Uber’s case, when a passenger or a driver completes a trip, those individuals will retain with the business at a given rate. It is, therefore, fair to amortize the cost of acquiring the demand and supply cost over the average lifetime of the demand and supply, respectively.
The formula to get the lifetime of supply and demand is the same, which is:
Demand/Supply lifetime in months = 1 / average monthly logo churn rate
I like using the average monthly churn rate over the last couple of months to minimize potential noise. Please understand that this lifetime number is somewhat fictional with limited roots in reality, as supply/demand ideally retain asymptotically. However, this is a great ball-park calculation to get to a single lifetime number. You then separately create an amortization schedule of the cost of acquiring the demand and supply, spread out over the calculated lifetime of the demand/supply. Take the amortized monthly total of the supply acquisition cost and put it into the variable costs/COGS of the demand-side dashboard. Conversely, take the amortized monthly total of the demand acquisition cost and put it into the variable costs/COGS of the supply-side dashboard.
See the example math for a demo marketplace P&L in this GSheet
I also added a walkthrough video, which you can view >> here
With this math, you’ll end up with the P&L input required for a fully loaded demand-side and supply-side analysis. We at Headline leverage these inputs to run our analysis via our automated financial analysis tool, Deepdive.
What if the Gross Margins in the Dashboard are Negative?
The most important data point for an early-stage marketplace is not its current gross margin but the trend line of the gross margin. As a marketplace grows, its network effects grow, which should increase the pricing power of the marketplace or the ability to upsell participants with additional services, which together will grow the margin of the marketplace. The same network effects might even lower customer acquisition costs, as either the demand or supply side gets more value from the marketplace, which reduces the relative amortized cost of acquisition layered into variable costs, further increasing the gross margin.
If this trend line cannot be identified, either the marketplace is not building network effects, or the effect of these cannot be seen in the data yet. In those scenarios, an operator or investor might need to make assumptions that the business might get there and improve its margins in the long-term. Operators and investors should be aware of this when pricing the marketplace, as it does act as an additional risk factor.
In Summary…
Tech-enabled marketplaces are complex entities that connect buyers and sellers through an online platform. Their unique structure, involving two distinct customer bases—demand-side and supply-side—require a specific way to analyze them within Deepdive. They require separate Deepdive dashboards for each the demand and supply side, with two distinct customer acquisition budgets and a nuanced understanding of variable expenses, including the amortized cost of acquiring the opposite side of the marketplace.
Despite these complexities, the potential for network effects to enhance value and improve margins over time makes marketplaces a compelling but intricate investment opportunity. Understanding these dynamics is crucial for accurately assessing and pricing such businesses.
Deepdive is a free-to-use, privacy-first analytics tool that is open to everyone. Try it out!