Uber and Airbnb are two of the most iconic companies from the last decade. Both companies created entirely new markets via a marketplace model and were originally considered to be part of the same “sharing economy”. Since then, their paths have diverged. While they were both wildly successful, Uber (and Lyft) required far more funding to create value than Airbnb did.
Uber and Lyft are valued at roughly 2–3 times total equity funding raised (incl. funds raised in their IPO). Airbnb is far more capital efficient, with a valuation-to-funding multiple of 9–10x. In other words, Airbnb created 3–5 times more value than Uber or Lyft for every dollar of funding raised. This is critical for investors to understand as it directly affects returns.
Despite the common “sharing economy” tag applied to them, Uber and Airbnb are built on very different models. Both run marketplaces that connect underutilized assets (and later, professionals) to consumers, with a self-reinforcing network effect, i.e. the addition of a supplier makes the product more valuable for all customers, and vice versa. They were both founded in Silicon Valley and were venture capital funded. However, that is where the similarities end.
Hyperlocal Network Effects
Uber’s model relies on hyperlocal network effects, i.e. the addition of a unit of supply (a driver) makes the product more valuable for the demand side (riders) within a small geographic radius. So when Uber acquired a driver in a city, it only helped it grow organically within that city (usually, within a small part of that city). And when Uber expanded to other cities, they had to re-invest in driver acquisition without the benefit of any latent demand. They had no drivers and so did not have riders to attract them organically (commonly called the “cold start” problem). This was complicated further by the fact that Uber’s success catalysed competitors in other markets who then created their own local driver networks before Uber could enter them (e.g. Didi in China, Ola in India, Careem in the Middle East, Grab in South East Asia, etc.). Since local competitors had an established hyperlocal network effect, it became even more expensive for Uber to enter and operate in these markets. As a result, Uber was forced to eventually sell many of its regional units to local competitors and acquire others.
These dynamics pressure unit economics and increase the amount of capital required to address a given market. This frequently results in investors overestimating the addressable market that be targeted sustainably (see: SoftBank Vision Fund). Since unit economics are less scalable, the valuation-to-funding multiple for hyperlocal marketplaces tends to be low. And consequently, this is why many local marketplaces tend to be fragmented by geography (from C2C commerce marketplaces like Shpock or Letgo to food delivery services like DoorDash or Deliveroo).
Cross-Border Network Effects
Airbnb’s model, on the other hand, is built on cross-border network effects, i.e. the addition of a unit of supply (a host) makes the product more valuable for the demand side (guests) across geographic boundaries. For example, a new host in New York could attract tourists visiting the city from anywhere in the world. When Airbnb expanded to other markets, untapped demand from existing guests made it easy and cost-effective to attract hosts in new markets. Existing Airbnb users could book properties in the new destination and this organically attracted hosts in the new destination, i.e. Airbnb did not face the “cold start” problem when expanding to new markets. So once Airbnb began to scale their demand and supply sides, they faced very little competition from other, regional startups who had limited supply in other regions. This made Airbnb’s unit economics immensely scalable, lowered capital requirements and resulted in a high valuation-to-funding multiple. As a result, Airbnb is heading towards an IPO with largely healthy unit economics while Uber is still at least a year away from profitability (and sustainability remains an open question).
Convoy, which has a valuation-to-funding multiple of 4–5x, is another great example of cross-border network effects. Convoy connects shippers (companies shipping physical products globally) to truckers who transport shipments from ports/airports to other destinations. For example, adding a unit of supply (a trucker) in New York, makes the product more valuable for any shipper across the globe who wants to ship to the Northeastern part of the United States. And when Convoy expands to other countries, it can leverage untapped, latent demand from shippers to attract new truckers. Existing Convoy customers like Unilever can book truckers in the new market, which will organically attract local truckers.
Both Airbnb and Convoy are examples of network effects without any boundaries, but not all cross-border network effects need to be truly global. Some markets could be naturally constrained by factors like regulations, cultural preferences, etc. and exhibit “regional network effects”.
For example, Numbrs operates a financial marketplace in Germany and boasts a valuation-to-funding multiple comparable to Airbnb. As Numbrs expands to other countries (e.g. the UK), it will need to attract new suppliers because of regulatory and licensing requirements for financial services firms without the benefit of latent demand in the new market. However, adding a new unit of supply here (a financial product) will help attract users anywhere in the UK (the entire market bound by the new regulatory framework), and not just a small portion of it. Scaling these marketplaces is far more efficient as compared to those with hyperlocal network effects, but less efficient when compared with truly global marketplaces. So as Numbrs expands to other geographies, we should expect its valuation-to-funding multiple to decline but still be well ahead of hyperlocal marketplaces.
Our Investment Thesis for Marketplaces
Geographic restrictions to network effects are a critical, but not the only, input to evaluate the potential of marketplace startups. Other factors to consider include frequency of interactions/transactions, marginal costs, side-switching, multi-tenanting, size of market adjacencies, fragmentation of supply and demand, supply differentiation, SaaS components, etc. I will write about some of these intricacies in the future. But assuming these factors remain equal, we view marketplaces with cross-border network effects as better bets than those that rely on hyperlocal network effects.
A good place to start looking for these types of marketplaces is to think of real-world communities or networks that already transact or communicate across geographic boundaries. There are some markets that naturally gravitate to this sort of a model. Sizeable industries like manufacturing and industrials deal with suppliers all over the world (e.g. Scoutbee). Similarly, markets like financial services and early stage startup funding connect supply (financial products or investors) and demand (consumers or investors) within reasonably large markets and geographic areas. These could even be found in industries as diverse as medical devices or genomics. Our thesis is deliberately sector agnostic and “horizontal”, so that we can help founders leverage winning business model patterns and solve critical problems that we may be unaware of. If you are an early stage founder building a marketplace that fits this description, email me here.
Of course, this does not mean we will ignore every startup reliant on hyperlocal network effects. There are key problems in sectors like childcare and elderly care that can only be addressed with local service-oriented marketplaces. But as we have seen, hyperlocal network effects tend to shrink the size of the market that can be addressed sustainably. In many of these cases, venture capital (and the growth expectations it carries) may not be the right fit for entrepreneurs.