Basic definition: a product which becomes more valuable as more people use it
Eg: Uber became more valuable as more drivers & riders used it, telephone is also a networked product
The product’s network is distinct from the actual product. Product is physical/software; network is made of people
Most successful tech products make use of the network effect
Network effects are made up of the network and the effect
The network is defined by the people and their creations. Companies don’t own anything in this network; they just connect different people really well
The effect part is how the network makes more action more valuable
It is now really difficult to launch a product, but network effects lends itself to virality (which plays well in the zero-sum attention era)
Network effects act as a great defensive moat. It is easy to copy a product, but it is not easy to copy the network of a product
A brief history
Metcalfe’s Law: the value of a network grows by the square of the number of users in the network
This is very simplistic and leaves out a lot about network quality, how to start a network, etc. This was the model of thinking in the 90s
A better mental model is that of social animals
Below a certain threshold of users, the network is not very usable. Above the threshold, the network can grow
The network has a carrying capacity. You can do various things (eg. spam detection, bot deletion) to grow the carrying capacity of a network
At a certain point, if the network overextends, it can collapse spectacularly
Cold start theory
5 primary stages in building a network:
The cold start problem: most new networks fail. You need to find a way to provide value immediately to users in the network
Tipping point: atomic network is established and provides value, which immediately makes new networks easier to create
Escape velocity: companies scale up to create sustain network growth, trying to improve its acquisition, engagement and monetization effects
Hitting the ceiling: negative forces in the network multiply and growth slows. This is usually solved and another growth spurt occurs, over and over again
The moat: networks are used as defense/offense against other companies
The Cold Start Problem
Tiny speck
Your first priority should be to build a singular, small atomic network that is self-sustaining
Eg: Slack built a networked product. First used IRC for internal comms when team was building Glitch. Then focused on IRC tool and seeded it within friends’ companies, which was the atomic network
Anti-network effects
If the network is too small, people will leave. This leaves us with the classic chicken-and-egg problem
If you plot network size on the x-axis and various metrics on the y-axis, you will see a kink in the network. That kink is your threshold that you need to pass
If that threshold is high, it will be hard but far more defensible if you can succeed
You need to have density and interconnectedness in your first network. 10 people in the same team using Slack is far better than 10 people in 10 different teams
The atomic network
Networked products often start with small networks and gradually build up (eg. one campus, one company, one city)
Eg: Bank of America pioneered credit card and started launch in Fresno by sending a few thousand people a free credit card and signing up small merchants
This expanded throughout California
This small atomic network needs to have enough density and interconnectedness to overcome anti-network effects
Usually networked products are extremely minimal, extremely focused on the atomic network and use unscalable techniques to grow
You can use growth hacks to launch and build an atomic network
Networked products usually target a niche, making them seem like a toy. This is an underestimation
To start a network, you need to focus on the smallest network possible (eg. Uber focused on a very small radius around a particular train station at a particular time)
If the atomic network doesn’t require lots of people, it’s easy to go viral and grow, but its not as defensible
Building one atomic network makes the next one even easier
The hard side
The “hard side” of your network are the users who contribute the most value but are the hardest to retain
Eg: contributors on Wikipedia, content creators on YouTube, doc creators, app developers
Usually governed by power law
You need to keep these users happy in order to generate network effects within your atomic network
Eg: Only 0.02% of total viewers on Wikipedia consistently contribute
Networks have hard sides because there are parts of the network that require work. Contributors expect value as well and will try multiple resources before settling
You need to understand the motivations of the hard side of users so you can cater to them right from the beginning
Solve a hard problem
Your first step in creating an atomic network is attracting the hard side. This requires satisfying a hard problem for this group of users
Eg: Tinder solved the issue for attractive women by combining location-based dating, trust signals via Facebook and easy swiping to drastically reduce the work needed to do online dating
Previous iterations of dating software made it difficult for attractive women to select dates
For most marketplace software, the supply side is usually the hard side
In order to figure out the hard problem for the hard side, you need to understand which problems are they engaging with but has not been addressed well
Segment your underserved audience as much as you can and look at their side hustles and things they do in their free time
The killer product
Networked products are different because they facilitate relationships between users in a network, which is very different from traditional software
These products also need to balance the needs between hard side users and easy side users
The richness of a networked product comes from the quality of the network and the ability for network nodes to interact with each other, not the features of the product itself
Because of this, networked product are usually quite simple
Usually products that are trying to build networks use a freemium model to reduce the barrier of building a network
Magic moments
A product that has solved the cold start problem can start to deliver on its core value prop
Eg: Clubhouse in its heyday had so many chat rooms that you could use spontaneously
You can also try to quantify how often non-magical moments are happening, termed as “zeroes"
"Zeroes” are the worst possible experience on the software. Users who go through this often churn out. Keep a close watch on these metrics
Magic moments require a great product and a great network
The Tipping Point
Tinder
Once you have gotten the atomic network down, you need to start scaling and building new networks in order to reap network effects
Tinder’s initial strategy was to throw parties for popular, hyperconnected students on college campuses. The catch was that all partiers had to have Tinder
They scaled this approach by continuing to throw parties around campuses or other areas of dense social connections
The tipping point is when you transition from creating atomic networks to tipping entire markets over to your software. Momentum is on your side
Invite-only
Invite mechanics are useful because a curated network will invite people similar to them, making a stronger network. It’s like a copy-paste mechanism
LinkedIn used this mechanism to make a curated network of professionals who would actually connect with others. This made sure people could be trusted and the network quality was high
Another pro for this mechanism is that you can create a better onboarding experience because they were invited, and so there is already some trust
Invites also create hype and can help teams build infrastructure for more people as the network grows
To curate these high quality networks, use ratings, high-touch onboarding or collect more information during waitlist signups
Come for the tool, stay for the network
If you can provide good value for just 1 user but then is complemented with a great network, this can kickstart growth
Instagram was initially hyped out for its great photo filtering experience, but it always had a network. Now, most pictures don’t have filters and people stayed for the social network
This strategy greatly eases the cold start problem because the threshold for utility is quite low
Major clusters of tool + network combo:
Create + share
Organize + collaborate
System of record + keep up to date with others
Look up + contribute with others
Transitioning a product from tool to network is tough! There must be clear, direct links between the network and the product
Paying up for launch
Many companies drive up cost in order to get their network into a tipping point, at which point money is not needed to provide momentum
Eg: when coupons were just beginning, companies would give out free samples of their product, which incentivized grocers (the hard side) to stock up
This eventually led to regular customer traffic → regular stocking at groceries
Uber did this to get drivers in a new city: they guaranteed a base salary and used give/get referrals to incentivize people to join
You should only use money for growth once you have a product and market figured out and know exactly how you want to scale your atomic networks
Crypto did the same by incentivizing initial miners with large sums of Bitcoin and gradually tapering off rewards for miners
Microsoft did the same by partnering up with IBM to sell its first OS. Even though there was a cost of doing so, it helped them establish a network
Unprofitability in the short term may lead to dominance in the long term
Flintstoning
Flintstoning: incomplete product is made complete by having a Wizard-of-Oz functionality of humans running various functions
Once a network is formed, you then automate yourself out
Usually, you try to imitate the hard side as you prop up your network. Reddit did this by having the founders post interesting articles
Flintstoning can be combined with software to make it scalable
Eg: PayPal used bots on Ebay that would only pay through PayPal to force EBay to use PayPal
Once the cold start problem is solved, stop flintstoning. It may hurt
Always be hustlin’
You need to bre creative when figuring out how to reach the tipping point
Escape Velocity
Dropbox
When the cold start problem and the network tips over, your product will start to have runaway hockey-stick growth
This doesn’t mean everything is great: you have to maintain high levels of growth and maintain network quality
When Dropbox hit escape velocity, it focused on certain user segments (esp. high-value businesses) in order to maintain runway for big projects
The trio of forces
When a product hits escape velocity, you will most likely need to scale up your team
The network effect is made up of:
Acquisition effect: the network makes it easier for new users to sign up.
Engagement effect: dense networks allow for more stickiness and usage by users
Economic effect: network improves monetization
These work in loops and will grow stronger as the network grows in strength, creating a positive feedback loop
The engagement effect
Cohort retention curves are extremely important, as they determine how sticky your product is
Most products have a 60% retention after day 1, 30% after day 7 and 15% after day 30. Networked products sometimes beat these numbers because the product becomes more sticky
Sometimes, product use cases will become more apparent with a larger network. The product should try to make these use cases clear
To figure out how to nudge users to be more engaging, you need to segment based on levels of enagagement and figure out the different needs for users
Once you have segmented users, you can run different A/B tests to see if the cohorts have different behaviour based on changes you made
Based off this, you can find characteristics that separate great users from mediocre users
Networked products often have engagement loops, where actions make actions further down more likely and it creates a cycle
Eg: When a content creator posts on Instagram, they get likes & comments from followers, which gives social validation for the creator to post more
If you understand your product’s engagement loop, try finding ways to improve each step of the loop
The enagagement effect can often reactivate users if they see that their network is using the product
The acquisition effect
Networked products will often have viral growth which are baked into the product itself
Eg: PayPal & Dropbox products are only useful if you have people to send money/documents to
How do you make acquisition easy? Figure out the funnel and optimize via A/B tests
To measure this effect, we can use the viral factor: out of the users in a particular time period, how many referred others into the app?
Retention usually helps a lot with acquisition
A network can have a strong acquistion effect and a weak engagement effect
Usually, networked products that have the acquisition effect have health
ier networks because there is interdependency between nodes in the network
The economic effect
Sometimes driven by data network effects, where the value & costs of customers are better understood as the network grows larger
Eg: credit bureaus expanded from local to international, which helped improve the ability for bureaus to determine if someone was trustworthy for credit
Bigger network → more data → more personalized
Bigger networks often mean that any spend can be extended to far more users.
You can structure premiums on your service such that it is more useful with more people, so bigger networks naturally go premium
Networks provide big defense because the switching cost becomes high
The Ceiling
Twitch
As product grows to scale, growth starts to slow, so product teams need to find new ways to create growth
Justin.tv focused in on gamers and streamer technology, which kicked off intense growth after stalling
Facebook used growth teams to grow after plateauing at 90M users
Many B2B startups go afer big enterprises during times of stagnation
Rocketship growth
Generally, rocketship companies will get 2MARR−>triple−>triple−>double−>double−>double(gettingto144M)
This is based off the assumption that the company wants to get 1B valuation in 10 years (typical in VC-funded businesses)
This is quite hard to do for any company since the pace of growth needs to be roughly the same throughout all years
When you hit a ceiling, it’s imperative to find a way to reignite growth as it can dry up investment, engineers leave, etc.
Networked products are fare more likely to sustain rocketship growth
Saturation
Sometimes, your product will be able to serve everyone until there are very few left. Focus should shift on revenue generation and layering services
Eg: Ebay started offering new ways of buying items (fixed prices rather than auctions) and kept layering new products → revenue growth
A network’s growth can stop due to market saturation (almost everyone is on it) and network saturation (diminishing returns of new nodes in the network)
Can counteract by thinking about the set of adjacent users who are not as healthy as the core users of the product
Can also have new formats for the same product offering (eg. Snap stories)
Can launch in new geographies
Works well if geographies share users, but beware of different cultures and attitudes
Acquisition is another strategy that companies can use to combat saturation
The law of shitty clickthroughs
Law: marketing channel performance degrades over time, which will impact growth trajectories
Why? Humans ignore ads after seeing it for so long
This is a serious problem for networks because this will have huge impact on the bottom of the funnel and can even impact engagement
Solution is to layer on new marketing channels all the time and focus on making the network healthy
When the network revolts
Networks have a backbone of power users who start to professionalize and eventually will push back against the product
These network professionals are either home-grown or were professional before they entered
There’s no choice but to embrace these professionals, even through their protests
Eternal September
Usenet ws like an early version of Reddit. When more consumers got access to the Internet, they flooded Usenet with bad content which killed it
Every network has netiquette, but if too many networks brought in at once, netiquette can break down since there’s too many contexts
You can build tools that enable people to create different contexts and avoid issues (eg. private stories, finstas, Slack warning about time zones)
Can leverage the network to monitor good behaviour (eg. Reddit uses downvotes to know if something is bad)
Centralized software usually outperforms here; decentralized software can’t coordinate to fight off bad actors
Overcrowding
When a network becomes too crowded, it becomes much harder to use
Eg: YouTube faced this issue and initially relied on manual curation; now it uses algorithms
Another issue that happens when a network is overcrowded is that new node additions become much more difficult
Networks naturally have lots of data, which means that developing good algorithms for curation is much easier
The Moat
Wimdu versus AirBnB
Wimdu was a European clone of AirBnB and was much larger and had more employees. They failed because their supply was nowhere near as good as AirBnB
AirBnB fought against Wimdu by quickly internationalizing their product and team
This section is all about how networks compete
Vicious cycle, virtuous cycle
The cold start problem acts as an initial defense for networked products
It is quite difficult to compete against a network as it keeps growing and steals supply from you. It’s even more difficult when the network is integrated well
Uber is easier to tackle than AirBnB because Uber’s networks are primarily city-based, while AirBnB are global
Network competition leads to winner-takes-all, so it is a high-stakes competition
This is not necessarily a product competition, but more of a network dynamics competition (eg. which network grows faster and has better quality)
Just because you have a network doesn’t mean that you have a magical defense. It’s all about which network is amplifying their acquisition, engagement and monetization effects better
Network competition leads to vicious cycles for losing companies since the network quality degrades as people leave
Cherry picking
When networked products get large, it might not be able to serve all parts of the network equally well. Startups can notice dissatisfied groups and use them as their audience
These upstarts can build a denser subnetwork than the larger network and people will switch over. Network quality & density >>> network size
This is very painful for large networks because it is hard to regain networks & they need to solve the Cold Start Problem again, but this time with competition
If you cherry pick to the point where the network you develop is dependent on a larger network, you run into dependency issues
Big bang failures
Big companies love to go all out when launching networked products with a big, wide launch. This often fails (eg. Google+)
It usually failes because the networks that develop are weak and shallow
Big bang approaches for launch are usually not targeted and use broadcast channels, which leads to lots of users but weak interconnections
Harder to assess if network is growing well
Larger companies try to target the biggest market possible, but smaller startups can go for niches and build far more resilient networks
Competing over the hard side
When networks compete, it’s usually over the hard side
A good strategy that many companies used is to attract hard side individuals from competitor’s network to theirs
Uber used incentives and bonuses to attract drivers from other apps to theirs
Competitive intelligence is worth it’s weight in gold. Uber used anonymous reports from credit card bureaus, email agencies and even reverse engineered APIs for intelligence
If products have similar networks and not significantly differentiated, you cannot really have a winner-takes-all
Bundling
Bigger products can cross-sell other products easily, which can help them quickly get over the cold start problem
Bundling doesn’t work if the products it is cross-selling is poor quality
Good bundling allows different products to talk to each other and use each other’s networks
Eg: Instagram and Facebook are now heavily integrated, such it’s easier for networks to communicate with each other
Bundling is powerful but it can create clutter and poor product decisions