Global vs. local network effects

How to figure out whether a platform’s network effects are global or local when things are not as obvious as for Airbnb (clearly global) or TaskRabbit (clearly local)?

All other things equal, global network effects are stronger and more defensible than country-level network effects, which in turn are stronger and more defensible than city-level (or local) network effects. Many times, it is obvious whether a platform’s network effects are global or local in nature. Sometimes, however, things are less clear-cut. We’ll discuss examples of each and provide useful questions that investors can ask to determine how global or local a given startup’s network effects might be. These questions should also be useful for entrepreneurs in assessing the strength of their network effects.

Let’s start with two obvious examples: TaskRabbit vs. Airbnb.

TaskRabbit’s network effects are local. Meaning that they only exist within limited geographical areas. A tasker in Boston only cares about TaskRabbit clients in the Boston area. And vice versa, only Boston area taskers are relevant to a Boston-based client. Contrast this with Airbnb. Its network effects are clearly global. A host in Paris, France may receive guests from any country around the world, so the value of Airbnb to such a host is clearly higher the more travelers Airbnb has in all countries, not just in France. Conversely, a traveler from Brazil derives higher value from Airbnb when the latter has more hosts in all the countries the traveler might want to visit.

Note that the notion of local vs. global network effects is not about whether the company is local or global. TaskRabbit is present in several countries in North America and Europe, yet its network effects are local. This might seem like an obvious point to make, but we have seen startups with local network effects that are eager to claim a national or international presence, implicitly (and incorrectly) suggesting their network effects are stronger because they are active in many cities.

What about ride-hailing apps like Uber? At first glance, their network effects appear mostly local: Uber drivers in San Francisco only care about Uber riders in San Francisco and vice versa. However, to the extent that Uber riders travel and use Uber in various cities, that creates some amount of country-level network effects. San Francisco’s Uber drivers may also care about Uber’s Boston riders if many of them go to San Francisco. Thus, Uber’s network effects extend across cities which receive significant numbers of visitors from each other. That being said, we still expect the majority of Uber riders in a given city to be locals. And Uber’s global network effects are likely considerably weaker because a much smaller fraction of Uber riders in a given country are visitors from a different country.

Why does the distinction between local vs. global network effects matter? Because all other things equal, global network effects are broader and last longer as the platform expands, and ultimately provide far more defensibility for platforms than local network effects. Think of it this way. Every time TaskRabbit launches in a new city (let alone a new country), its network effects essentially start from zero. A tasker in Austin, Texas couldn’t care less about TaskRabbit’s installed base of clients in other cities. Nor does a client in Vancouver care about the many taskers that TaskRabbit has in US cities. Contrast this with Airbnb. Every time Airbnb launches in a new city or country, its entire installed base of hosts and travelers from all existing countries and cities is relevant and gives it a massive advantage over any local competitor. Once Airbnb’s network effects are working globally, any new entrant needs to enter and build liquidity globally to compete. This is a large part of the reason Airbnb has few (if any) meaningful international competitors. Meanwhile, TaskRabbit has many competitors, both in the US (Thumbtack, Handy) and internationally (Airtasker in Australia, Bark in the UK, now also in the US). And Uber ended up losing many of its international competitive battles: against Didi in China, against Grab in South-East Asia, etc. Uber’s large installed base of riders and drivers in the US gave it no real network effects advantage in those regions, and ultimately the regional competitors won out because they had the advantage of better local knowledge.

As a quick application of these distinctions, consider Boatsetter, a marketplace that allows users to rent private boats (with or without a boat captain) from boat owners. Does it enjoy local or global network effects? We don’t have any particular insight into the behavior of Boatsetter renters, but a key question we would ask in order to evaluate the strength and defensibility of Boatsetter’s network effects is: how many Boatsetter renters are local vs. visitors? The higher the percentage of Boatsetter renters that are out-of-town visitors, the more global the network effects are, and so the stronger and more defensible they are.

To further refine our understanding of local vs. global network effects, consider Shappi, a platform which allows buyers in South America (Ecuador only for now, presumably expanding to other countries soon) to purchase products only available in the US and have them delivered by verified travelers. In short, a buyer in Ecuador might buy an iPhone online and have it delivered to Shappi’s US warehouse. Shappi then assigns a verified traveler that plans to travel to Ecuador soon (travelers share their planned trips with Shappi), ships the iPhone to the traveler, then the traveler takes the iPhone to Ecuador when traveling there and delivers it to a Shappi representative, who then delivers it to the buyer. Typically, a traveler would deliver multiple products (for multiple buyers) to Ecuador and Shappi pays them based on the number and weight of the products they deliver.

Are Shappi’s network effects local or global? In some sense, they appear global because they cross international borders: Shappi enables interactions between buyers in Ecuador and travelers from the US to Ecuador. A buyer from Ecuador values Shappi more when Shappi has a larger network of travelers from the US to Ecuador. And vice versa, a traveler from the US to Ecuador values signing up with Shappi more when it has more Ecuador-based buyers, because that means opportunities for higher payout.

However, it is quite apparent that the network effects are mainly tied to specific “routes”, such as US-Ecuador. Assuming Shappi expands to other South American countries, a buyer in Columbia would care about travelers from the US to Columbia, who are most likely distinct from the travelers from the US to Ecuador. Conversely, travelers will most likely not change their travel plans based on Shappi delivery opportunities. A traveler who plans to go from the US to Columbia only cares about the number of Columbia-based Shappi buyers and not about the Ecuador-based Shappi buyers.

This means Shappi is not nearly as defensible as a platform with truly global network effects like Airbnb. One could start a competitor to Shappi for the US-Peru route and be evenly matched with Shappi. Meaning that Shappi’s existing network of buyers and travelers for the US-Ecuador route gives it no advantage on the US-Peru route. In this sense, Shappi looks more like TaskRabbit: its network effects are largely “local”, where here “local” means specific to a route.

One could argue that some travelers may go to different destinations, so a Shappi traveler that went to Ecuador this month might go to Columbia next month. First, we doubt there are many such travelers that use Shappi. If there are, Shappi would be well advised to focus on expanding to routes that share travelers in common. Second, even if there were many such travelers, Shappi buyers still only care about US travelers coming to their specific country, so the network effects remain far less global than Airbnb’s.

Key take-aways

There are many shades of gray between local and global network effects. In general, whether a platform’s network effects are closer to local or global depends on several factors, including the nature of the products or services being traded, transportation costs, language barriers, the share of users that use the platform in more than one location, etc.

Thus, marketplaces for digital products have inherently global network effects – see for example Open Sea, Uniswap. So do marketplaces for digital services or services that can be performed remotely – see Upwork, Fiverr, Magic.

Meanwhile, when it comes to marketplaces for physical products, the key question to ask is: how much demand for trade is there between different locations? An important factor that determines the answer is the magnitude of transportation costs relative to the value of the products being traded. Etsy has global network effects because it facilitates the trade of relatively small items (e.g. jewelry) which can be shipped internationally at reasonably low rates. B2B wholesale marketplaces like Alibaba and Faire.com also have global network effects because they facilitate the trade of bulk orders. By contrast, the network effects of car marketplaces for individual consumers (e.g. CarGurus, eBay Motors) are largely local.

Finally, for in-person services, the questions to ask in order to determine whether a platform’s network effects are closer to local or global are:

• Do participants on one side interact with participants on the other side from many different locations or just one? (see the discussion of Shappi above)

• For every given location (e.g. city), what is the share of participants on each side that are local vs. visitors from out-of-town (or out-of-country)?

More fundamentally, in our view all these cases boil down to figuring out whether the platform in question accumulates a network effect advantage as it expands to new locations. To what extent does the network from one location provide an advantage when “opening up” a new location? The answer is “a lot” for platforms with global network effects like Airbnb, Alibaba, Etsy and Upwork, vs. “not much if at all” for platforms like TaskRabbit and Shappi. For ride-hailing platforms like Uber or food-delivery platforms like DoorDash, the answer is “somewhat” when it comes to opening up new cities in the US, and “not much” when it comes to opening up new countries.

This discussion also suggests that companies can be strategic in how they expand, in order to maximize the instances of global network effects as much as possible. For example, Gable connects companies that wish to rent collaborative workspaces for their remote employees with hosts who own suitable spaces. On the surface, Gable’s network effects appear to be mostly local, but it can choose to pursue companies that are present in multiple cities and expand to those cities where its existing company-clients have workers, thus deliberately taking advantage of inter-city network effects. And in some cases companies can take action to at least partially expand the reach of their network effects from local to global. For example, eBay Motors offers remote inspection and shipping services via partnerships, which can facilitate more cross-country transactions.