Using network effects to create defensible barriers
Facebook. Amazon. Google. Apple. These are some of the most valuable companies in the world. On the face of it, they provide different products — social networking, online shopping, search engine, hardware and software. Yet, they share one thing that has contributed to their tremendous success.
Network Effects.
A platform or product is said to exhibit network effects when the more that people that use it, the more valuable it becomes to every user. So, more the number of buyers on Amazon, greater is the number of third-party sellers attracted to the platform, thus offering more choices to consumers. More the number of prime subscribers, greater is the incentive for Amazon to build more fulfillment centers (FCs) and improve their supply chain. More FCs, translates to quicker delivery, thus attracting more people to join the prime subscription. Virtuous cycle created.
Today, network effect is the holy grail for all digital product companies. Companies try to attract an initial base of users with an offering and over time get these users to participate in a network. Once a critical mass of users is achieved, the network creates incremental value for the users and a defensive moat for the company.
Once a critical mass is reached, the value offered by a network far exceeds the cost of adding new customers.
Which accounts for the tremendous valuation of companies such as Facebook, Amazon and Uber. However, there are also subtle differences in the network effects enjoyed by the said companies. Broadly speaking, network effects can be categorized as follows.
Marketplaces. A marketplace is usually two-sided with buyers and sellers transacting in either products or services. Most common examples of a Marketplace are Amazon, Uber and Airbnb. Once a critical mass is reached, it is very difficult to break the monopoly enjoyed by a marketplace. A disadvantage inherent in a marketplace is the option open to a seller or buyer to simultaneously use a competing product. If I can get a cab using Lyft or Ola or any of Uber’s other rivals, there would be a very minimal switching cost for me.
Social Networks. Facebook, Twitter, Instagram, LinkedIn etc. enjoy the network effect of a marketplace mostly without the threat of competing networks. Unless all my friends were to leave Facebook and join a rival network, I would have no incentive to shun Facebook. As long as my professional network continues to be on LinkedIn, I will have little interest in a similar product even if it were to be better designed.
Content Networks. Spotify, which is today valued at $26 billion, owes its exponential growth to network effects. It facilitates the discovery and sharing of new podcasts and personal stations, thus attracting more customers. Netflix, Medium and Amazon Prime are other examples of content networks. More the number of writers contributing high quality articles to Medium, more will be the number of people flocking to join Medium.
Data Network Effect. Considering that ‘data is the new oil’, companies constantly try to extract value from the data they collect to gain competitive advantages. The more customers you have, the more data you collect and hence thus get even more customers. In reality though, it is easy for companies to overestimate the value of their data. For example, Netflix is famously known to use the tons of viewing data they collect to provide better recommendations to customers. These recommendations are based on your own viewing patterns along with those of users who watch similar shows.
However, the true value of Netflix is not based on its recommendations but rather the library of original programs that it develops.
Hypothetically, a new upstart network could tomorrow create or buy an exciting new range of shows, movies and documentaries and be able to wean customers away from Netflix. Even without all the data that Netflix owns. Thus, Netflix exhibits only a weak Data Network Effect. In contrast to Netflix, stands Yelp.
In 2004, Jeremy Stoppelman was struck down with the flu. He searched online to find a good doctor, but all he could find was general information which did not really help him. Stoppleman thought that if he could get his friends and acquaintances to start reviewing local services, it would solve a common pain point. Thus was born Yelp, a social review site, where people review restaurants and other local businesses. Today Yelp has more than 180m reviews and tens of millions of users. What Yelp managed to do was to create a virtuous cycle of individual consumers and business owners, both benefiting from the network effect. The crowdsourced data on its platform offers Yelp a defensible position because it will not be easy for any other company to level the playing field by purchasing data from other sources.
The most powerful network effects are built when companies use customer generated data to shut out competition. Let’s take a look at the criteria that product leaders can use to judge if it can build itself a data enabled defensive moat.
Using data to build defensive moats
How can you judge if you have an opportunity to build a data powered defensive moat? The first important criteria is the speed with which data can be put to use. Value of customer data is greater when the insights from that data can be quickly translated into product improvements. The longer the time it takes to learn from customer data, the higher the likelihood that competitors will build or buy their own data and innovate.
In 2018, Google Maps had a 67% of the market share compared to Apple Map’s 11%. This dominance isn’t due to the design features of Google Maps which can be easily copied and even improved upon. Rather, Google leverages its huge user base to learn from real-time user data and predict traffic conditions and recommend optimal routes more accurately. Google also uses crowd sourced data to predict how crowded a bus or train will be even before you board it. This data enabled learning allows Google to continually and rapidly improve its algorithms, thus incentivizing more people to use Google Maps. Result is a data powered network effect.
The second important criteria is the rate at which value of additional data starts tapering off. The sooner this is, the weaker is the network effect. Consider the case of speech recognition software. Products like Alexa and Siri need to be trained on the voice samples of a limited set of people for any language. Here, data enabled learning cannot provide any significant advantage.
In contrast, let’s look at the business of Advanced Driver Assistance Systems (ADAS). Self-drive vehicles are touted as the future of driving and companies such as Tesla, Google (Waymo) and Uber are pouring in billions of dollars to build and support these vehicles. ADAS which provide collision detection warnings are an important component of these vehicles. It is critical that these systems be fail-safe before they are incorporated by a car manufacturer. Let’s assume that after a significant amount of testing, an ADAS is able to achieve 99% accuracy in its predictions. From here on, every single decimal point gain in accuracy would require a lot more testing. However, give the life or death implications, even a small increase in accuracy would be highly sought after by car manufacturers. This is why companies like Mobileye are able to defend against competitors as every additional data point has an associated value.
The third criteria to be considered is the availability of proprietary data.If a company has data that is not available to its competitors then this grants strategic and sustainable advantages. Using this data, a company can come up with unique insights which can become a value multiplier. IBM Watson partners with Mayo Clinic to study patient data and predict the probability of breast cancer. Timely predictions help save lives and also reduce treatment cost by catching the disease in an early stage. The more the number of patient records that is analyzed, the better is the accuracy of diagnosis. A competitor would find it difficult to provide a similar product, as patient data records are sensitive and cannot be bought or generated easily.
To create a defensive moat, companies have to constantly and rapidly learn from customer data. The data itself would have to be proprietary so that competitors cannot generate it. And the product improvements would have to be based on insights that can be gained only by having a significant amount of user data.
Takeaways
In the future, the most valuable businesses will be those built on the back of network effects and enhanced by data enabled learning. As product leaders it is important to identify the opportunities for network effects to maintain a strong competitive position.
- Consider the number of users who also use other similar products. This is usually true for most marketplaces (Uber, Amazon); a significant percentage of your users also consuming a rival product or service would need a continuous investment in both supply side and demand side to maintain your position.
- Also look at the exit cost for your current users. The higher you can make the exit cost (unique data generated value offerings, users having to rebuild their social network) the stronger the network effects.
- Check if adding new users to your network, increase the value proposition for your existing users. If true, this will lead to stronger network effects.
- Try to keep the learning cycle between getting customer data and acting on the insights as short as possible.
- Proprietary data sets are tougher to compete against. If you are in a market that already has a competitor with a strong data enabled network effect, check if the data can be generated through AI. This will allow you to negate your competitors network advantage and build up your own product.
I hope this article was helpful. If it was, feel free to follow me on Twitter where I share thoughts and articles on product management and leadership. You can also check out my startup https://komenco.in, which is dedicated to helping products scale.