Why Network Effects Matter Less Than They Used To
When we teach strategy to MBA students, they want magic bullets, things they can do to make their companies thrive forever. For a long time we emphasized “network effects” as a potential secret sauce for business models. Economists use “network effects” to describe contexts where a good or service offers increasing benefits the more users it has. Network effects can be direct: for example, Slack becomes more useful as other people also use Slack. Network effects can also be indirect, meaning that one set of users benefits as more of another type of users joins a platform. For example, AirBnB would not be useful for travelers if there were no apartment-owners using the platform. Similarly, home-owners would not want to use AirBnB if travelers weren’t using it to find a place to stay.
We have long taught that network effects can provide market power and sustained or even self-reinforcing competitive advantage (the best kind). The more users you got, the larger your user base was, and the more compelling your proposition became for attracting new users.
At the tail end of the dot-com boom in Silicon Valley, I wrote my dissertation on network effects. Entrepreneurs and business leaders were excited about them too. But it now seems they are not the panacea we first thought.
Take Microsoft. In 2000, people thought they had unassailable network effects; they were the poster child, attracting attention from envious business people and from regulators. Who would ever desert the Microsoft ecosystem and not use Word and Excel? This expectation that consumers were welded to the Windows operating system meant that developers and computer manufacturers all felt they had incentives to focus on contributing to the Microsoft ecosystem. However in 2018, Microsoft is struggling to retain customers within its ecosystem, and developers are offering many products that don’t rely on Windows. (I wrote this article in Google Docs, not Microsoft Word.)
What has changed? Why don’t network effects work as they used to?
For one thing, today’s network effects are not tied to a particular piece of hardware, like a desktop computer. Since 2000 and the desktop era, we have seen the evolution of multiple different devices, such as smartphones, tablets, and digital assistants such as Alexa. This means that network effects are no longer intertwined with a particular piece of hardware, as was the case with the desktop computer in the 1990s. Instead, any notions of scale for technology companies depend on user profiles that can be ported to multiple different hardware platforms.
This means that platforms which exhibit network effects may be purely digital. Social networks, ride-hailing apps, or digital marketplaces do not depend on any one type of hardware, and as a consequence it costs very little for users to try new ones out. Having five different social media apps on my phone is not a problem at all. Having five different desktops with different operating systems, on the other hand, is clunky. (Something similar is true for the supplier side. It’s generally cheaper for a developer to write code for a new platform than to build and ship a new device.)
The effect on competition is clear. Think about Lyft and Uber. Ride-hailing is characterized by fierce competition and firms burning through obscene amounts of venture capital in an effort to reach scale. However, users can easily install both Lyft and Uber apps on their phone and judge in the moment which is cheaper. Likewise, on the driver side, many drivers have both Lyft and Uber installed, and choose to operate on whichever platform is offering them the more profitable ride.
These examples remind us that network effects only really work as a source of competitive advantage if your product is also “sticky.” Scale will not bring future competitive advantage through network effects if your customers can all leave tomorrow.
Some people argue that digital platforms can be made sticky through customers owning data through a platform. The theory is that data stored in one place can lead to lock-in which in turn will power up network effects.
However, history belies this. Look at the iTunes Store, which was once my favorite teaching example of a sticky digital platform exhibiting network effects. Once you had a library of data in the form of mp3s, I used to say, there was no way that you would shift to another platform. Who wants to have their music library in multiple places? How could you come up with the right party playlist? Music providers knew they would have to be there to reach listeners and listeners would naturally gravitate to iTunes because they knew that all the music they might want was there. Then Spotify entered, and showed us just how short-term this advantage was. What does a music library of mp3s matter when you can stream any song at all at any time?
Courtesy : Harvard Business Review