Discover more from Course Notes: Continuous Business Learning
The Cold Start Problem 1/7
I’ve been waiting for Andrew’s book [Amazon Link] for a while now. He’s got a wonderful blog, which I used to read with a lot of pleasure. This time I’ve decided to summarise his book for my records and for your entertainment.
1/ What’s a Network Effect, Anyway?
It describes what happens when products become more valuable as more people use them. [MK: logically, some aspect of these people’s engagement with the product – like, share, interaction between each other – improves the core product metrics. For two-sided marketplaces value is created interdependently by users and sellers, drivers and passengers, etc.]
The network is a group of people using the product to interact with each other. YouTube: a network of content creators provides content for viewers to consume – via the platform with the recommendation algorithm. Content creators are on the network (in part) because they need the audience, too. These people don’t own the asset (platform), but own the content.
The effect is the means of increasing value as more people start using the product. Increasing value means increasing engagement leading to faster growth or higher monetisation or something else.
If a product has a network of users interacting with each other, it’s probably a networked product. If either acquisition, retention, or monetisation become easier / better as the product growth – the effect is probably strong.
New products are being launched every day despite an obvious fact that every single space is saturated, cutting through the noise is close to impossible and users are not loyal. There are only 24 hours in a day, and maybe adding network effects to one’s app can capture a larger chunk of their time (engagement) or make them stay longer (retention).
Competition for attention has shifted from fighting boredom (waiting for a bus) to doing it and much more in an addictive way. Building version 1 of the app is fast and cheap (meaning that it will be cloned within days after the first signs of traction emerge).
Paid customer acquisition has become very expensive as incumbent players have learned how to monetise traffic (and app installs) really well, thus driving prices up for everyone, which hurts young startups the most.
Network effects shape user behaviours, effectively creating defensive barriers (or moats – if you like looking down instead of up) and making switching to copycats harder, pointless, or both. [MK: Networks of users don’t migrate to competitors simply they have rolled out a similar feature. One can argue, though, that those who haven’t made it into top content creators on one network are more likely to migrate to a competing network to do their own land grab. But it’s a limited use case.]
Some products get adopted from within the organisations (Slack, Dropbox, etc. being the notable examples). Once a critical mass of users emerges – companies have to adopt these tools for everyone.
2/ A Brief History
The dotcom boom gave birth to such myths as “the first mover advantage” (with a handful of exceptions, a smart follower has way more advantage), or “winner takes all” (value has become very complex and owning the entire value creation chain is impossible and impractical).
Metcalfe’s Law (“the systemic value of compatibly communicating devices grows as a square of their number”) was incorrectly extrapolated to apply to networked products and has partially caused the dotcom bust. It fails to take into account the quality of the network (active vs registered users), network degradation (lagging software or harmful content), and the multi-sidedness of many networks (buyers vs sellers).
Meerkat’s Law (made up by the author) describes the fluctuations of social animals’ populations based on the occupied area. [MK: in a weird twist of things, my Masters (Math) thesis was about the fluctuations of animals’ populations based on the availability of food and predators. Hits sooo close to home.] Applied to networked products, this means that if the population (user base) growth is below a certain safe threshold, the network dies (lower engagement, people delete the app). Unhinged growth leads to exceeding the carrying capacity of the network resulting in lower engagement (too many messages —> fatigue, too many goods in a marketplace —> poor choice and lower purchase frequency). Reducing fatigue, creating discovery mechanisms and fighting spam increases the carrying capacity (saturation).
In two-sided marketplaces the Meercat’s Law applies separately to buyers and sellers. Initially the value grows as the number of both grows; at some point the law of the diminishing returns kicks in and having more sellers carrying the same product is not going to improve the buyer’s experience but will have fewer sales among all sellers.
3/ Cold Start Theory
It’s a framework for creating value with network effects, consisting of five primary stages.
The cold start problem. Most new networks fail due to the lack of content (images, posts, goods sold) or contacts (meaningful interactions, buyers for the goods). A disbalance of users and content early on creates “anti-network effects” damaging the product. This balance is unique for every product, and sometimes this problem can be initially solved cheaper and/or faster than investing in a full-scale product – via the atomic network (the minimum network size for a meaningful value-creating interaction between users).
Tipping point. This is the point when the network starts growing on its own, capturing like audiences and expanding from one population to another. It occurs in consumer apps and SaaS products.
Escape velocity. A newly created network needs stabilisation and nurturing, so the company proactively invests not just in further growth (which is still needed), but also in making sure users keep receiving even more value than before. There are three types of the company’s activities at this stage:
Acquisition – getting new users efficiently and cheaply via viral growth. Can be via referrals, content sharing, recommendations, etc.
Engagement – increasing interactions between users and networks grow (avoiding loose ends and gaps in user experience). Increasing the number of uses of the product means moving users up the “engagement ladder” via gamification [MK: does anyone still remember what it is?], incentives and product features.
Monetisation – higher revenue per user or per user group (SaaS) and conversion rates (efficiency).
Hitting the ceiling. A rapidly growing network wants to both grow and tear itself apart from the inside. [MK: it’s such an elegant saying…] Eventually, headwinds start to blow and growth stalls. The most common reason is the uneconomically high customer acquisition costs driven either by competition or saturation (every new paid customer is more expensive than the previous one). All acquisitions channels eventually grow stale; engagement can be tainted by spam, fraud, context collapse, change of interests, etc. Growth usually goes from hitting a ceiling to hitting another ceiling and so on (unless the new ceilings suspiciously start resembling steps on a staircase leading down). This problem is as fundamental as it is common and solving it over time requires making increasingly complex decisions. [MK: From a risk management perspective, certain risks like spam and fraud need to be actively managed but can’t be completely eliminated unless every single user is thrown off the platform.]
The moat. As the product matures and becomes visibly successful (in at least some aspects), the goal is to use network effects to fend off competitors. The sad truth is that competitors usually grow thanks to the same tailwinds that made your product successful, so the nature of the network-based competition shifts to the ecosystem-based competition. [MK: ecosystems also never fail to … fail] As copying each other’s moves doesn’t make much sense, competition evolves into an asymmetrical fight, mainly driven by the size of the network available. The smaller the network – the more defensible it needs to be, and it’s very preferable to have the unit economics and user engagement trajectories to look convincingly steep.
MK: I’m old enough to remember the “installed base” competitive advantage where companies used to measure their moats in the number of devices or operating systems with a useful life of several years (and also trying to bold a few things on top to increase stickiness). Good times…