Network effects in finance
This post is the first of a series of collaborations between Fintech Ruminations and other like-minded friends, colleagues, and contributors. This post was co-authored with Aika Ussenova, Strategy manager at Wise, fellow Fintech Nerd and author of Aika's newsletter - a Forbes-featured newsletter on Fintech.
What are network effects?
“There are many products for which the utility that a user derives from consumption of the good increases with the number of other agents consuming the good.”
Michael L. Katz and Carl Shapiro
Network effects is probably one of the most used (and often abused) expressions in technology today. In the tech ecosystem it has been used to support massive startup valuations and to explain outlandish success cases.
The goal of our post is to dig deeper into the role networks effects play in fintech to provide a sustained competitive advantage.
Network effect is defined as a phenomenon by which the utility of a product or a service for a user depends on the number of users using it: the more users using a product, the more value each user will get.
The concept, formalized in the brilliant paper “Network Externalities, Competition, and Compatibility” published in 1985 by Michael Katz and Carl Shapiro, has proved to be formidably successful in many markets, particularly in the internet economy.
Social networks like Facebook, LinkedIn and Twitter; messaging platforms like Whatsapp, WeChat or Telegram; two-sided marketplaces like Uber, AirBnb or eBay: all these highly successful businesses have network effects at the core of their growth.
In the case of Facebook, it is pretty obvious that the tool would be ultimately a ghost town if no users were there. Facebook’s real asset and value prop for the users is the network, and each user joining the network is adding an extra bit of value to existing users: you can connect with one more person. This is what Shapiro and Katz define as a direct network effect.
In the case of AirBnB or Uber, the network effect is less obvious. A Uber rider doesn’t directly benefit from another rider joining the platform, but clearly benefits from a new driver joining the platform. In this case we have what is defined as an indirect network effect: there are 2 interdependent groups and the utility of one group grows as the other group grows.
In the examples above, the product doesn’t really exist without a network, the network is a core element of the product. There are, however, other examples of products in which the network effect wasn’t part of the initial core value of the product, but was added as an added bonus. This is the case of products like Figma or Dropbox or Instagram: users come for the tool but stay for the network.
What are NOT network effects?
The fact that many successful internet and tech businesses are built on network effects, does not mean that every successful product must include a network effect. Likewise having a mass adoption doesn’t mean that a product has a network effect in action.
The foremost example of this is scale. A company like Amazon has massive scale which guarantees incredible economies of scale. But these economies of scale are not network effects. Economies of scale allow the supplier to diminish the cost of providing a service (on the offer side), while network effects amplify the utility for the user (on the demand side).
On a brutal term: the gain in utility that a user obtains from an extra user buying on Amazon is not even comparable to the utility gain that the same users achieve if they both use Whatsapp.
The case of viral content is a bit trickier. Does it bring value to a user that somebody else is watching the same viral video? In a way, this brings value because the user can ultimately share an experience with the other person and this is a relevant part of virality. But, on the other hand, having another person watching the same video, doesn’t add anything to the specific experience in itself. So ultimately this is not a network effect comparable to having another contact on Whatsapp.
Network effects in finance
Network effects are very prominent for some of the more successful fintech companies. But it isn’t an obvious path to grow, and in many cases network effects become a by-product of a successful strategy, rather than a prerequisite to viability.
Network effects are critical for companies like Venmo and CashApp. They are also important indirectly for platforms or marketplaces such as Klarna, Amex, or DeFi lending protocols where they cultivate demand and supply loops. For the majority of companies though network effects really transform into economies of scale. No direct node relationship is necessary, but economies of scale allow to build a better product. Banks and neobanks are examples of such business models.
For these businesses, customer ownership is critical. Put simply, it is hard to build any strong network effects for a B2B business and usually successful B2B companies benefit from economies of scale directly. That’s why here we will focus on consumer fintech examples.
Not surprisingly the extent of how important network effects to businesses correlates directly with the breadth of services provided by these companies. So we broke down fintech business models and mapped them against two axes: network effects and jobs served.
Network effects spectrum
Consider the models or companies we mentioned above.
These companies are mapped on the spectrum of how important network effects are for them vs types of network effects they’ve built:
P2P payments: P2P transfer success is entirely based on direct network effect: the more users you get, the better the service you can offer. Having people you transact with on the app is synonymous with the product itself. Examples include Venmo and CashApp.
Two-sided networks: For two-sided networks like P2P lending marketplaces, BNPL providers, or platforms with consumers and merchants, indirect network effects are fairly obvious: the more users the company adds on one side of the platform, the better the service will be for the other side of the platform, and vice versa. These go in loops. Examples include companies like Klarna, Paypal, Amex, Square+CashApp and P2P lending businesses.
This also includes DeFi protocols which have a pure marketplace nature: they connect at least two sets of users that fulfill different needs on the protocol. It is the case of DEXs - such as Uniswap or Sushiswap - and Interest-rate protocols, more simplistic defined as lending protocols, such as Compound or Aave. In these cases, the protocol enormously benefits from indirect network effects: the more users are present on one side, the better it is for the other side of the protocol. A perfect example to describe these mechanics is Compound, one of the oldest lending protocols in DeFi. On Compound users can borrow crypto assets from other users that are lending them, paying an interest rate. The more users borrow on Compound, the better it is for the lenders (both in terms of pricing and liquidity)
Social investing: Social investing applications - like Public, Partyround - are a new form of investment strategy that emerged over the last decade and it ultimately merges social networks and investment apps. A user can discover other investors, see what their portfolio is and follow their strategies. Essentially, these apps add a coordination layer to an uncoordinated group of investors: in theory, the coordination of some agents should per-se bring some value to each user and the more users are coordinated, the bigger the value is. This is what happened when the meme-stocks phenomenon started: a group of retail investors managed to coordinate themselves in a very effective way, mainly through the subreddit wallstreetbets, and achieve enormous gains. But what happened with AMC and GameStop is something extremely hard to achieve at scale in other more liquid asset classes, which are what retail investors traditionally invest in. As such, we believe that the emergence of network effects for social investing applications will be limited to specific asset classes for limited time windows.
Banks: In the case of neobanks, we argue that direct network effects do not really exist: as a Revolut user I have a very limited benefit when another user joins the product. The main benefit is the economy of scale that the company can achieve from spreading the fixed cost of serving extra users - but this is ultimately a benefit for Revolut, that can or cannot be transferred to the individual user.
For someone like Venmo losing the hold on the social graph may mean the end of the product. A bank on the other hand is as useful, whether its customers transact with each other or not. It isn’t surprising - Venmo provides a single service, whereas banks serve all financial jobs. So network effects combined with breadth of service make an interesting map.
Jobs served spectrum
Now let’s map the same companies and business models on the jobs served spectrum in order to show the depth of their service - from one job to many.
The weaker the network effects, the more jobs the company tends to serve. This is not a coincidence. The companies that have direct network effects can ramp up and grow very fast by having just a simple one use-case product. On the other hand, the companies that provide deep and broad services, depend on network effects to a much lesser extent. However, there are some interesting dislocations pointing towards business model equilibrium moving into serving more jobs and thus diminished importance of network effects.
P2P payments: P2P transfer products like Venmo serve only one job. They depend the most on the network effects, which are fundamental to their viability as a business. But businesses that depend just on network effects can unravel fast. The growth of CashApp overshadowing Venmo is proving that. Incidentally CashApp deepened its product stack quite fast to serve more jobs, on the way diminishing the importance of network effects for its viability.
Two-sided networks: BNPL providers like Klarna and Affirm initially served just one job - shopping. But both expanded in order to own the primary customer relationship by providing more services - from current accounts, to cards, to building a shopping destination. They're moving down the slope where serving more jobs correlates with lower importance of network effects. Owning demand directly will lead to a stronger position vis a vis supply.
Consumer-merchant platforms such as PayPal, Square+CashApp now serve the full scope of financial services directly comparable to banks. On the chart above they are quite dislocated vs the slope too. But that also shows that with more jobs served deeper, each part of the network becomes sufficient on its own.
With regards to P2P lending businesses we now know that they have failed. Most of the P2P lending pioneers transformed to be banks - SoFi, Lending Club, Zopa. That was a direct response to their own dislocation on the slope above - serving a single and in-frequent job while needing to maintain two sides of a network that are dependent on each other is unviable. As a result, they moved across to a richer banking proposition and towards diminishing the impact of the network effects.
On the same spectrum is DeFi lending. For now, because the product in itself has more inherent frequency attached [more hype, more early support, deeply engaged user-base], the risks of P2P lending above are distant. However, it might mean that DeFi lending would still need to develop other services and jobs to sustain long-term.
Social investing: Companies like Public do serve one job, but similar to the DeFi protocols above, this job is frequent. People actively trade, rather than passively hold income. So job’s depth overlaps with frequency of engagement. However, other single-job trading platforms from Coinbase to Robinhood broadened the set of jobs served. Granted they don’t have the social element that Public aims to build, nevertheless expanding the menu of jobs appears to be a path forward for these businesses too.
Banks: Banks classically serve all financial jobs, and so only really have economies of scale. They don’t need consumers interacting with each other to sustain the business model. However, the scale and adoption of a banking product then results in network effects appearing post-factum. Monzo says each customer has 30+ contacts on Monzo, Revolut is a popular P2P payment solution between its customers. But that is an after-effect of the product adoption, vs a prerequisite required to succeed. Moreover it is unclear how these emerging network effects can help these businesses fundamentally. One benefit is that they could increase switching costs, but that is also not a sustainable moat given the competitive landscape of neobanking.
What is the ultimate path?
Network effects naturally occur more where people interact with each other through the product. Where the value of the product to people is connecting them to each other, network effects are key. Where the value of the product is internal and connecting people to each other becomes a smaller part of the product, then the value and hence the impact of network effects diminishes.
To the extent that business models evolve into providing fuller service to customers coming into the product, the value of the network effects becomes less important. People might come for the network, but stay for the tool. Staying for the tool is an important aspect of consumer fintech - strategically that has been the business model that works in consumer fintech. The key reason is retention of customers or LTV/CAC that are particularly key in consumer fintech. Higher life-time value of a customer means companies can afford to spend more to acquire that customer. Network effects help to acquire customers en masse and cheaper - in the case with CashApp they acquire customers for just $5 on average. But in fintech, higher LTV is possible only if customers have a long-enough lifetime and do more during that lifetime. We live in times when neobanks are launching stock trading, and stock trading apps are launching accounts.
The question is what retains the customers. Initially that might be the network effects. But unlike social apps, financial products are primarily utilities. Customers hire these providers for specific jobs - to send, save, spend. But the individual specific job reaches the frequency asymptote very quickly [perhaps that’s why a lot of neobanks start with spending cards which serves a naturally more frequent job]. Businesses then expand the jobs served to retain customers for other things. This is true not only for neobanks, but also for specialists who started with an initial wedge, but expanded on the breadth of services eventually.
In fact, on the extreme side, utility of the product becomes so strong that classic network effects entirely diminish and transform into economies of scale. Look at the evolution of pure peer-to-peer network businesses. CashApp was a P2P payment app for the first few years, which allowed it to grow the user-base very quickly. Then it started adding a lot more utility in the form of cards, accounts, stocks, bitcoin, and lending. Engagement of the product increased and the company said that customers using cards are 2-3x more valuable. The P2P element of the product is secondary if not diminished entirely.
The final destination of this trajectory is that 2-sided networks then become superseded by the economy of scale. Embracing the necessity to provide more jobs, all business models eventually converge on being a version of a bank or what these days is called a financial super-app.
Ben Thompson writes that maps can take two forms: some give direction, and others provide context for what has already happened. The above ‘network effects - jobs served’ map is a context of what happened, but it nevertheless reveals what seems to be the ultimate successful consumer fintech path on a long-enough time scale.