The Network Effects Map | NFX Case Study: Uber

‍by James Currier, NFX

(To sign up for NFX essays on network effects, growth, and psychology, visit:

Though many founders are easily misled to focus only on growth, the actual value of a business largely rests on defensibility.

Over the last few years we’ve developed a Network Effects Map to help founders visualize what most do not see, in order to build an informed defensibility strategy. The way we delineate different ideas with this Map is not right or wrong. It’s one of our evolving methods for helping founders dissect the often-hidden yet highly intentional forces behind technology’s most successful companies.

The map provides a vocabulary for discussing complex and interrelated concepts, with the express purpose of helping operating teams take action to improve. Without action, it’s just entertainment.

To help founders understand and make use of the map, we’ll feature it in a series of posts where we’ll take an “NFX-ray” of the defensibility of well-known companies.

First up is Uber.

The Network Effects Map of Defensibility

As we’ve discussed before, in the digital world we believe there are four remaining “moats” or defensibilities:

  1. Network effects, the idea that the more users/data/nodes you have using your product, the more valuable it becomes for all users. We believe network effects is the most important defensibility because a) it’s native to the digital world, b) it’s available to startups of all types, c) there’s larger optionality because there are at least 13 different types to use when building your company, and d) network effects have been responsible for 70% of value creation in technology companies since the Internet made it easy to network anything.
  2. Embedding a product into the way that customers operate, like Oracle or Workday do, so the product is hard to rip out and replace.
  3. Scale makes it hard for competitors to compete with pricing and logistics, as with Amazon.
  4. Brand increases the psychological cost of switching to a competitor, because customers feel an emotional or reputational attachment to the company, like with Apple or IBM.

It is important to realize that you’re not choosing just one defensibility for your startup; instead, think about a company like a painting. Each of the defensibilities, including the 13 network effects, are a color. Some works of art are monochromatic, others use the full spectrum. Some businesses get big with one defensibility, others have many.


This video uses the Network Effects Map to reveal how we see Uber trying to build its defensibility.


Here are the five main points:

  1. Uber is vulnerable because its core network effect, a two-sided marketplace between drivers and riders, is asymptotic. It’s relatively weak. This is because there’s not much added value to a rider getting picked up in 3 minutes versus 4 minutes. So, beyond that point, adding to the supply of drivers has diminishing returns. Other companies can enter a city’s ride sharing market if they can reach a sufficient threshold to provide comparable pickup times.
  2. The threshold to enter a market isn’t very high because both sides of the marketplace will “multi-tenant,” meaning they will use multiple services. The drivers will drive for Uber, Lyft and maybe more. The riders will add another app to their phones and flip between the apps to get the best price/service.
  3. In the face of a weak core network effect, Uber has done a good job of adding other defensibilities like scale and brand. Brand in particular. They have stayed in the press for a whole variety of reasons for 7 years now. With scale, they have raised so much money, they can continue to build and scale regardless of profits, which theoretically drives their utilization up and cost per ride down. (This may or may not be true past a certain threshold.)
  4. Their aim in the future should be to try and add stronger types of network effects. I would argue they are already trying to do that with Uber Commute, which has a true direct network effect (non-asymptotic). Possibly they could add a two-sided platform network effect based on the data they are collecting, which others could tap into, but I haven’t seen them try that yet.
  5. If they don’t add a stronger network effect, Uber will continue to be vulnerable, not only to Lyft, but also possibly to new entrants. The natural stable state of the ridesharing market is probably duopoly or oligopoly until or unless one of the players develops a stronger type of network effect.

How a weak core network effect makes Uber vulnerable

Let’s further analyze some of the points above.

Uber is using one type network effect as its central defensibility. Some types are orders of magnitude more powerful than others.

The core attribute of a network effect is that each new product user makes the product more valuable to all existing users. The more value added by each additional user, the more powerful the network effect.

Uber does have this property up to a point. Uber operates as a double-sided marketplace, with a supply-side of drivers and a passenger demand-side. The more local drivers join Uber, the more useful the app becomes to local passengers as pickup times decline. And vice versa, as more riders equal more profits from a driver’s standpoint.

But after a certain number of drivers join, a point of diminishing returns is reached. This is where Uber is at risk. The difference between getting picked up in two minutes or four minutes isn’t a huge deal to the passengers — it only delivers marginally more utility.

With a true two-sided marketplace, each additional user on the supply side makes the product much more valuable on the demand side, and vice versa. With Uber and other ride sharing companies, each additional user on the supply side brings diminishing returns for passengers past a certain point, and even a has a negative impact on other drivers. This kind of network effect is called asymptotic — it is not very defensible and therefore weaker than many other kinds.

This vulnerability is compounded by low switching costs for both drivers and passengers. It’s called “multi-tenanting” when supply or demand is willing to live in two or more places. Many drivers end up running both Lyft and Uber at the same time because the penalty for doing so is low. They’re also aware of the inverse relationship between the number of other Uber drivers on the platform and their own profits. Uber itself admits that, in New York City, one of its best penetrated markets, drivers still only spend about half their time with passengers in the car.

Passengers also have low switching costs, and lose little by firing up their Lyft and Uber apps at the same time to check arrival time. (Amazon used to have this problem until Prime made their service much stickier, making them more defensible.)

In the future, a third ride-sharing company could come along and build up a following, now that Uber has done the hard and expensive work of selling the world on the benefits of ridesharing. Over time, Uber’s advantages of aggressiveness and speed will diminish, as the market growth and penetration settles down. How many new entrants we see will depend on supply/demand threshold levels in each market and companies’ access to capital.

How Uber is addressing the problem, and what they might do next

Having realized that it’s vulnerable at a time when the stakes are high, Uber is furiously building other defensibilities.

Perhaps the most notable one is self driving cars. If they can achieve this, Uber will theoretically be able to defend themselves with a “threshold barrier” to competition, since they have a massive cash advantage from scale. In other words, they will be able to buy more self driving cars than competitors because they’re so big, providing them with a superior supply advantage for price/service. As a result, the cash threshold for new market entrants will be too intimidating — although this will be more of a scale effect than a network effect. Once safe from competition behind that high cash threshold, the theory goes, they could raise prices and profits. (We’re skeptical self driving cars are coming as quickly as Uber is telling their most recent investors, but that’s for another post.)

In reality, one of the more promising defensibilities Uber is building is Uber Commute. This is called a direct network effect where the more commuters, the higher the value for all commuters, and that value keeps increasing to a very large number of users. Uber Commute seems to be getting uptake right now. Uber Pool also has some of this characteristic, but not as strong.

Another move Uber has made to shore up defensibility is “embedding” Uber in other apps. Users of Facebook Messenger or Google Maps automatically see Uber suggested as a ride service, making it hard to compete with.

Brand is another moat that they’re using effectively — barely a day goes by without some news about Uber. Even before the scandals that have plagued them of late, Uber did a great job at making noise, be it in the press, in popular culture, and even in their company redesign.

In fact, they’ve even made an effective use of the rare “social” network effect of “language.” They encourage users to talk about “ubering to the airport” and “ubering home.” Lyft is far less often elevated by this verbal use of their brand name.

The Map

The Network Effects Map reveals the moves Uber has made in the past and should probably make in the future. Having started 10 network effect companies that created more than $10 billion in value, we not only are true believers in the magical results of network effects, but we’re moreover dedicated to exposing the mechanics of how they actually function.

Founders with ambition to build the next Google, Facebook, or Slack will undoubtedly benefit from using the map to build the best possible defensibility strategy.

In the end, defensibility is what matters most.

To sign up for NFX essays on network effects, growth, and psychology, visit: