Self Help

The Cold Start Problem - Andrew Chen

Author Photo

Matheus Puppe

· 55 min read



  • The introduction provides context about Uber’s operations at their San Francisco headquarters in December 2015. It describes the busy, global work environment with reminders of their international presence.

  • It introduces the concept of Uber operating a complex, global network of smaller networks representing each city. Each city network had to be started, scaled, and defended against competitors.

  • A major weekly meeting called NACS (North American Championship Series) was described, where the CEO and executives would deeply review metrics and operations for each major city market.

  • The level of granular detail discussed, even breaking cities down into neighborhoods, reflected Uber’s need to carefully manage each local supply-demand network.

  • Each city represented an individual network within Uber’s global network of networks. The health and growth of each individual network was critically important to evaluate and influence through tactics like marketing.

  • The introduction establishes the theme of Uber operating as a complex system of interconnected local networks, rather than just as a single overall business. Careful management of networks was key to their operations and strategy.

  • The passage describes a meeting of the North American Championship Series (NACS) team at Uber to discuss issues in key markets like San Francisco, Los Angeles, and San Diego. Drivers were switching to Lyft due to higher referral bonuses.

  • This was creating long wait times for riders and higher cancellation rates, frustrating both riders and drivers. The networks were becoming imbalanced.

  • TK (the head of the NACS team) advocates for a $750/$750 driver referral bonus to quickly add more drivers. Others discuss alternative solutions but TK wants immediate action.

  • The RGM teams agree to implement the higher referral bonus. The author helps change the product and rollout is planned for the weekend.

  • These weekly NACS meetings helped Uber respond nimbly to issues across markets in North America as the business rapidly scaled.

  • In 2018, the author transitions from Uber to a new career in venture capital at Andreessen Horowitz, drawing on experience from Uber and as a tech blogger/investor.

  • As a VC, the author meets with entrepreneurs pitching new startup ideas and business models leveraging concepts like network effects, viral growth, economies of scale discussed in this book.

  • The author realized they did not fully understand the concept of “network effects” despite hearing it referenced frequently in the tech industry. They set out to research and write a comprehensive book on the topic.

  • Network effects are a critical concept for technology companies but are often discussed superficially. The author interviews over 100 founders to learn more about how networks develop from early stages through growth.

  • There is a need for a universal framework to understand network effects across different product categories and industry examples. The book aims to provide practical insights and metrics beyond existing high-level discussions.

  • The core framework covers the full network life cycle from creating initial network effects (the “cold start problem”) through ongoing scaling and harnessing effects. It draws on theories and historical examples as well as founder experiences.

  • The author has applied their network effects knowledge through investments in over two dozen startups focused on connecting users. They seek to write the definitive guide to help practitioners apply the concepts to their own products.

So in summary, the passage outlines the motivation for researching and writing a comprehensive book on network effects to fill gaps in understanding of this important topic for technology companies.

  • A network effect describes when a product becomes more valuable as more people use it. Classic examples are the telephone network and social networks.

  • Theodore Vail, president of AT&T in 1900, noted that a telephone is useless without being connected to other phones, and its value increases with the number of connections. This early insight described the core concept of a network effect.

  • For modern products, the “network” refers to the people connecting through an app or service, while the “product” is the software itself. Successful network effects require both.

  • Many of today’s largest tech companies have billions of users due to network effects, including Facebook, Google, Microsoft, TikTok, AliPay and others.

  • A network effect has two parts - the “network” of users, and the “effect” where value increases as the network grows. Engagement and growth tend to be stronger as more people participate.

  • To identify a network effect, ask if the product connects users and if its value proposition strengthens as the network size increases. The strength of network effects varies by product.

  • The passage discusses the importance of network effects in driving the success of some of the largest tech companies today like Facebook, Google, etc. Understanding network effects is critical for entrepreneurs, competitors, and others in the tech ecosystem.

  • Launching new tech products has become incredibly challenging. There is intense competition for limited user attention from millions of apps and services. Marketing channels are saturated and expensive. Network effects provide one of the few barriers to competition.

  • The concepts of network effects have their roots in the origins of telecommunications over a century ago. But our modern understanding stems from the dot-com boom of the late 1990s when the commercial internet first emerged and network-driven businesses like Yahoo, eBay, Amazon took off.

  • During the dot-com boom, venture capital flooded into startups pursuing a “winner-take-all” vision, believing the company that attracted the most users and became the dominant network in its category would be nearly unstoppable. This led to huge valuations and hype, though many companies ultimately failed to deliver.

  • In summary, the passage outlines how network effects have become a defining force in technology and defines the challenges of launching new products in today’s hyper-competitive landscape dominated by large network-driven incumbents.

  • Metcalfe’s Law, which was popularized during the dot-com boom, states that the value of a network grows as the square of the number of connected nodes. It was used to justify high valuations of early network-based startups.

  • However, Metcalfe’s Law is an oversimplification that doesn’t account for key factors like network quality, engagement, multi-sided networks, and degradation from overcrowding. It has not held up well compared to real-world network building.

  • A better model comes from population dynamics of social animals like meerkats. Populations face an “Allee threshold” - below a critical size, benefits decrease and the population is at risk of collapse. Above the threshold, populations can grow exponentially.

  • But growth is limited by carrying capacity - the maximum sustainable population given environmental resources. For networks, “overpopulation” above carrying capacity degrades the experience through issues like too many messages, posts or listings.

  • This “Meerkat’s Law” model more accurately captures the phases of network growth and challenges of maintaining quality and usability as a network scales up in size. It provides a better framework than Metcalfe’s Law for understanding and managing network effects.

Here is a summary of the key points about capacity for users from the passage:

  • Network effects products often have an “Allee threshold” or “tipping point” where the number of users/activity needs to reach a critical mass for the network to become self-sustaining and grow on its own. Below this threshold, the network is at risk of collapsing.

  • The sardine fishing industry in Monterey, CA collapsed after overfishing pushed the sardine population below its tipping point/Allee threshold. The population then spiraled down and never recovered.

  • Similar dynamics can occur with technology products - if user counts or activity dip below the tipping point, the network effects start to unwind and it risks full collapse.

  • Drawing an “Allee curve” can show how a network’s value/conversion rates change based on the number of users/drivers/listings etc. It will be low at first until passing the tipping point, then increase up to a saturation point with diminishing returns.

  • For a product like Uber, value increases as more drivers are added up to a point, but eventually additional density doesn’t improve the experience much further once saturation is reached.

  • Founders need to focus on reaching the critical mass or tipping point needed to sustain growth, whether that’s a minimum number of users, listings, drivers or other activity depending on the specific network. Getting past the “cold start problem” is key.

  • The article describes a theoretical framework called the Cold Start Theory, which outlines five stages for creating, scaling, and defending network effects for a new product.

  • The first stage is the Cold Start Problem - building an initial tiny network that is self-sustaining. This is very difficult to do.

  • The story of a failed product called Tiny Speck is used to illustrate the challenges of the Cold Start Problem stage. Tiny Speck took 4+ years to launch a beta version and shut down, despite being a talented team.

  • Escape Velocity is the next stage, where growth must be sustained through strengthening acquisition, engagement, and economic network effects.

  • Eventually products will hit a ceiling and growth will stall due to issues like acquisition costs rising. Managing problems like spam is important.

  • In the maturing stage, network effects are leveraged to fend off competitors through asymmetric network-based competition strategies.

  • The theory aims to provide a framework for starting, scaling, and defending the network effect across different technology products and historical examples. Understanding these dynamics is important for building successful networked businesses.

  • Tiny Speck launched in 2009 to build an online multiplayer game called Glitch, raising $17 million. However, the game was not well-received and failed to retain users.

  • To collaborate internally, Tiny Speck built a chat tool on top of IRC to enable searchable messages, file sharing, etc. This became very useful for the team.

  • When Glitch failed, they decided to rebuild the internal chat tool as a standalone product called Slack that any company could use.

  • They privately beta tested Slack with friends at other startups, getting feedback and iterating. This helped form early “atomic networks” of sustained users.

  • As more and larger teams adopted Slack, they learned how usage patterns scaled up from dozens to hundreds of users. Constant feedback helped improve the product.

  • After years of development and beta testing, Slack eventually launched publicly and became very successful, valued at $26 billion when acquired by Salesforce. This was a remarkable turnaround for Tiny Speck after their first product Glitch failed.

  • Anti-network effects occur when a new network lacks enough users, causing a negative feedback loop where few users leave because their friends/coworkers aren’t on it yet. This is the “cold start problem” facing all new networked products.

  • To overcome cold start, a product needs to reach a minimum critical mass called the “atomic network” - the smallest viable network where users will stick around.

  • Networks often have two “sides” - an easier side to attract initially (like content consumers) and a harder side that does most of the work (like content creators). Success requires attracting the hard side.

  • To attract the hard side, a product must “solve a hard problem” by being compelling enough for key users. Tinder did this for attractive users.

  • Additionally, the most successful networked products are often “killer products” that are dead simple and emphasize interactions between users, like early Zoom.

  • When a product solves cold start, it creates “magic moments” where users find an active network is already built out upon joining. Network effects then take over and tipping point occurs.

  • According to Slack’s CEO, the minimum viable network is 3 users, where stable groups can form and the product can be considered a customer. Two users works but isn’t as robust.

Here are the key points about how creating an initial “atomic network” can help solve the Cold Start Problem:

  • Companies like Slack, Uber, and early credit cards started very small, in a single location/campus/company, to establish an initial engaged network.

  • This small, self-sustaining “atomic network” is stable enough that a second one can be built off it. And those networks allow building even more over time.

  • For offline networks, cities are a common unit of atomic networks due to geographic constraints. But online networks may start within colleges or individual companies.

  • The first credit card targeted a single town of 250k people, deemed the minimum critical mass needed. Mass mailing 60k cards without applications got the network going.

  • Starting with a tight, interconnected atomic network allows engagement, retention and usage metrics to improve gradually as more users are added, solving the cold start problem from the bottom up. Understanding how to build these initial small networks is key to long-term success.

  • Bank of America issued 60,000 credit cards to residents of Fresno, California on the same day. This created an “atomic network” of cardholders ready to use the cards.

  • They also signed up small, local merchants in Fresno to accept the cards. More than 300 Fresno merchants signed up in the first three months.

  • Within a year, Bank of America had expanded the network to other major California cities like San Francisco, Sacramento, and Los Angeles. They had issued 2 million cards and onboarded 20,000 merchants.

  • The key to their success was starting small in Fresno and focusing on saturation and density there before expanding. This created the “atomic network” - the smallest viable network that could then grow organically on its own as more users and merchants joined.

  • Building atomic networks is a common strategy for new digital networks like credit cards, games, collaboration tools, etc. It involves starting with a very niche, targeted group and focusing on building density within that group to achieve critical mass and momentum.

  • Many enterprise products like HR or project management tools require all or most of a company’s employees to use the product for it to have value. This makes viral growth difficult and a top-down enterprise sales approach may work better.

  • Growing networks city by city or campus by campus creates “atomic networks” with dense connections between users. This improves engagement, viral growth, and retention compared to spreading users randomly across a geography.

  • Every network has a “hard side” - the minority of users that create disproportionate value. For social networks, this is content creators. For marketplaces, it’s sellers. These hard side users must be engaged and retained for the network to function.

  • Looking at Wikipedia, which was created entirely by volunteers, shows that active contributors make up a tiny fraction of total users, around 0.02%. Understanding the motivations of these hard side contributors is important.

  • The hard side does more work but also creates more value for the network. Products must have a value proposition and engaging experience tailored specifically to their needs from the start.

The passage discusses the hard side of social content apps and networks, which are the small percentage of highly engaged users who create the majority of content. It describes the 1/10/100 rule, where 1% of users create content, 10% actively engage, and 100% benefit as consumers.

Content creators are motivated by different needs like self-expression, status, communication or talent/entertainment. Their motivation determines how difficult it is to participate - anyone can share a selfie, but learning a complex TikTok dance takes much more effort.

These content creators, or the “hard side”, are critical for these networks and apps to survive. Without them generating content, the network falls apart. Understanding what motivates different hard sides, like status for Instagram influencers or community for Wikipedia editors, is important for product design.

The key is solving important problems and needs for the hard side. The passage uses the example of online dating apps, explaining how each generation improved the experience for attractive users (the hard side), like reducing messages for women on Tinder. Solving the hard side’s problem is critical to attract them and get the network started.

  • The previous generation of online dating sites were like doing a second job, requiring users to fill out lengthy profiles and send messages in the evenings. Tinder made dating more casual and fun by allowing quick swiping on profiles.

  • Tinder addressed issues with sorting through large numbers of matches by integrating with Facebook. This showed mutual friends and ensured users were only seeing people nearby, building trust in the system.

  • For marketplaces, the “hard side” or supply side is usually more difficult to scale - these are the workers, small businesses, and sellers that provide the products/services. Initial focus is on bringing enough supply onto the platform.

  • Early ridesharing apps like Sidecar and then Uber recruited large numbers of non-commercial drivers by allowing anyone to sign up, following the model of the nonprofit Homobiles service. This drastically expanded the potential driver pool.

  • The hard side of networks often comes from people engaging in hobbies and side hustles during nights/weekends, representing underutilized time, assets, skills. Looking at niche segments within this can identify opportunities to start disruption.

Here are the key points about what makes for a great networked product idea:

  • Networked products focus on facilitating experiences between users, not just a user’s interaction with the software itself. They rely on network effects to grow.

  • They are often simple, doing one thing well rather than trying to offer many features. Simplicity is actually hard to achieve and allows the product idea to spread virally.

  • The most important features involve connecting users to relevant people and content on the platform.

  • Simplicity unlocks new “atomic networks” of just two or more users interacting in a new way, like Zoom did for video calls.

  • Simple, meme-like ideas are easy to describe and spread from user to user, fueling growth through word-of-mouth.

  • Critics often argue simplicity means lack of technology differentiation, but the interface belies complex underpinnings needed to power the network.

  • Focusing on facilitating connections between users is more important for growth than technical features or functionality.

  • Many viral consumer products like WhatsApp, Facebook, and early versions of Uber were built quickly by college students or small teams and only laterprofessionalized as the product and user base grew.

  • This trend has spread to the enterprise space, with products like Dropbox initially aimed at consumers but adopted widely in workplaces. Other products like Slack were created by founders with consumer backgrounds.

  • There is cross-pollination between consumer and enterprise as skills and ideas transfer between the spaces. Things like emojis, livestreaming, and on-demand marketplaces start in niche consumer use and spread widely.

  • Many successful networked products use a “freemium” model to grow easily without barriers. This allowed viral growth through word-of-mouth as users could try the product for free. Zoom adopted this model to drive initial growth.

  • New computing platforms like smartphones create opportunities as behaviors change and new interface paradigms emerge. This resets the playing field and allows new killer products to emerge that take advantage of new technologies. Zoom benefited from trends like widespread internet access and a shift to remote work accelerated by the pandemic.

  • Zoom was facing the ‘cold start problem’ early on, where the network wasn’t active enough and the product experience wasn’t quite working.

  • Two important steps to solving the cold start problem were 1) getting early adopter organizations like schools and programs to adopt Zoom independently, which would help grow the network, and 2) focusing on building “atomic networks” of initial connected groups that could spread the use of the product to others.

  • Once the basic product and initial networks were established, companies aim to create “magic moments” - when the experience of using the product truly shines due to a filled-out, active network connecting people in meaningful ways. This delivers the core value proposition.

  • The author provides the example of Clubhouse, noting its early empty, ghost town-like state before finding product-market fit. Over time, magic moments emerged as networks formed among tech communities and later Black creative communities, solving the cold start problem and fueling growth.

  • Launching at the right time, during the pandemic focus on connectivity, also helped Clubhouse achieve momentum according to interviews with early employees and investors cited.

So in summary, addressing the cold start problem involves developing the basic product, finding initial connected groups, and allowing magic moments to emerge organically over time as the network becomes more robust and delivering the core value. Examples from Zoom and Clubhouse illustrate solving this challenge.

  • Tinder became a dominant dating app by solving the “cold start problem” of building out online dating networks at scale. Dating apps face unique challenges due to hyperlocal nature of dating and high churn when matches are made.

  • Early on, Tinder focused on dense urban areas in the US to build localized “atomic networks” that took off. They knew how to replicate this success launching on new college campuses.

  • The bigger challenge was expanding to multiple geographies and demographics simultaneously to really scale. This required solving the cold start problem repeatedly across many new dating networks of different regions and groups.

  • Through repeated trial and error, Tinder figured out how to extend from one successful atomic network to two, then many more. When this scaling process became repeatable, it helped Tinder hit the “tipping point” to take over the online dating market worldwide.

  • By effectively addressing the cold start problem at both localized and global scales, Tinder was able to unprecedentedly scale an online dating product to tens of millions of users and billions of daily swipes, establishing the norms for mobile-first dating apps.

Here is a summary of how Tinder grew from a small campus app to a global dating platform:

  • Tinder launched in 2012 at the University of Southern California (USC), where the founders had gone to college. It started very simple, without the iconic swiping feature.

  • Early growth was slow, so the founders promoted the app by throwing a birthday party and requiring attendees to download Tinder first. This was a success, generating their highest ever one-day spike in downloads.

  • The key was launching at USC provided an initial network of social, connected students all using Tinder at once. Matches began happening quickly.

  • Tinder reached the “tipping point” where viral growth took over. They scaled the campus party launch tactic to other colleges, recruiting student ambassadors.

  • By throwing parties at Greek life and other campus groups, Tinder was able to rapidly expand to new student networks across the US.

  • Once they reached 20,000 users in a market, the app would see “escape velocity” and take over that region. They scaled internationally using a similar strategy.

  • Within a few years, Tinder became one of the top apps and found a business model as the highest grossing non-game app, redefining online dating around swiping and mobile use.

In summary, Tinder’s success came from strategically launching on college campuses to bootstrap initial viral growth, which allowed them to tip entire student and regional markets at a repeatable scale.

  • The article discusses the “invite-only” strategy for launching networked products like LinkedIn, Facebook, Gmail, etc.

  • With an invite-only launch, users can only join if they are invited by an existing user, limiting growth initially. This allows teams to fix bugs and scale infrastructure before fully launching.

  • More importantly, invite mechanics automatically “copy and paste” the initial curated network. If the first users invite others, it spreads the network on its own.

  • LinkedIn founder Reid Hoffman seeded LinkedIn initially with invitations to his connections, mid-level professionals. This curated the initial network to be relevant and useful.

  • Invites from this initial network rapidly grew LinkedIn’s membership in the first week through the “copy and paste” mechanism, without much other marketing.

  • Invite-only provides a better welcome experience for new users as they already have at least one connection on the platform from their inviter.

So in summary, the invite-only strategy focuses on curating a relevant initial network and leveraging their invites to automatically scale the product through network effects.

  • Invite-only launches can be beneficial for networking products by allowing the early network to gel as a community and develop strong, dense connections through word-of-mouth and organic virality. This results in a high-quality initial user base.

  • Early adopters tend to have large contact lists, inviting others with similar large networks. This quickly connects many new users to each other from the start.

  • Products like LinkedIn optimized their invite mechanics over time to mine contacts and suggest more connections, growing the network density.

  • Exclusivity from invite-only launches creates buzz and attracts attention. Gmail’s launch this way wasn’t initially planned but became a “marketing decision.” Early access provides status and benefits.

  • Curating a high-quality initial network is important for trust-based categories like marketplaces and ride-sharing. Uber personally interviewed all early drivers.

  • Waitlists allow scaling user intake while curating the network through social media engagement or info collection.

  • The strategy of “come for the tool, stay for the network” focuses on building utility first before emphasizing the social/networking aspects over time.

  • Hipstamatic was one of the earliest popular photo apps on the iPhone, allowing users to apply vintage filters to photos. However, it had some friction like requiring multiple taps to view filters and a delay between photos.

  • Instagram launched in 2010 and focused solely on photo sharing and filters. It was easier to use than Hipstamatic and also included social networking features like profiles, feeds, and sharing to other platforms.

  • Instagram grew massively through its viral sharing and network effects as more people joined. It is seen as a prime example of the “come for the tool, stay for the network” strategy - attracting users initially with photo tools but keeping them engaged in the social network.

  • While photo filters drew people to Instagram at first, the social features became more important over time. Instagram demonstrates how great tools can help bootstrap a network, but the network is what ultimately sustains long-term growth and value.

Here are the key points summarized from the passage:

  • Paying up front through subsidies like coupons can help drive growth for network products by reaching the tipping point faster. This was done successfully by early companies like Coca-Cola.

  • Coupons in particular help address the chicken-and-egg problem for new products trying to get into grocery stores. Grocery stores won’t carry a new product if consumers don’t ask for it, but consumers can’t try it if stores don’t carry it.

  • Marketing pioneer Claude Hopkins used coupons successfully to get his client Van Camp’s Milk powder into grocery stores. Coupons helped create enough initial demand to convince stores to carry the new product.

  • The goal of subsidies is to drive enough early growth and positive network effects to reach the tipping point, after which the product can pull back subsidies and focus on profitability with a large established user base and network effects driving the business.

So in summary, paying upfront through subsidies like coupons can help products overcome the challenges of network growth by creating enough initial demand to kickstart the network and reach critical mass. The goal is establishing a large network that then takes on a life of its own.

  • Van Camp’s milk devised a plan to promote their product by inserting coupons in newspaper ads for a 10 cent can of milk that could be redeemed at any store. They paid grocers the retail price for redeemed coupons.

  • The clever part was focusing on grocers (the “hard side” of the network) with the coupons. This incentivized grocers to stock the product to redeem coupons, helping bootstrap the supply side of the network.

  • They tested this in smaller cities then expanded to New York, gaining 97% distribution through letters to grocers explaining the upcoming coupon promotion. Over 1.4 million coupons were redeemed, costing $146k but establishing the brand in homes and capturing the market.

  • Uber faced a similar chicken-and-egg problem and began by subsidizing drivers with hourly guarantees to bootstrap the supply side. They then leveraged referral programs and word of mouth to recruit more drivers as the network grew.

  • Financial subsidies and leveraging the existing network through referrals/word of mouth are effective strategies for address cold start problems and accelerating growth in two-sided markets. Cryptocurrencies like Bitcoin also use economic incentives to bootstrap participation.

  • Early product launches often lack key features that require significant development work like account deletion, content moderation, referrals, etc.

  • “Flintstoning” is where companies manually perform these functions themselves in the interim using back-end tools, rather than having the feature fully built out.

  • This allows the product to get to market faster while still addressing user needs, getting feedback, and building initial networks of users and content.

  • Reddit is cited as an example - when it first launched, only the founders posted content. But this manual approach allowed them to bootstrap initial communities until other users started joining.

  • Over time as momentum builds, Flintstoning techniques evolve into automated product features as the company scales. The goal is to manually fill critical network gaps until self-sustaining.

  • It’s a way for startups to solve the “Cold Start Problem” and get a networked product off the ground before fully developing every desired feature.

So in summary, Flintstoning refers to manually performing functions that are not yet automated, to quickly bootstrap networks and get user feedback as a workaround prior to complete product development.

  • In the early days of Reddit, the founders Steve Huffman and Alexis Ohanian would post content themselves using dummy accounts to make it seem like there was an active community on the site.

  • Over time, Steve wrote code to automatically scrape news sites and post content under fake usernames. This helped scale their efforts but still required Steve’s attention. Once when he went camping for a month without submitting links, the front page went blank.

  • Other companies like Yelp, Quora, food delivery apps, and B2B marketplaces also use “Flintstoning” by manually filling in key parts of the network themselves until it grows organically.

  • Flintstoning can involve a spectrum from fully manual to hybrid human-automated to fully automated approaches. Technology is layered on to create leverage over time.

  • At the extreme, platforms may build entire internal studios/companies to create key network content themselves, as Nintendo does by developing flagship games to launch new consoles.

  • The goal is to transition Flintstoning efforts from manual to automated as the network grows, and eventually let organic users/sellers drive it rather than internal fake accounts or content.

  • Uber held a team retreat in Las Vegas for over 4,000 employees to celebrate hitting $10 billion in gross revenue, another major growth milestone.

  • The retreat was branded “X to the X” and featured performances by David Guetta, Kygo, and Beyoncé to boost morale.

  • The Operations team, comprising thousands of “boots on the ground” workers, deserved most credit for Uber’s growth. They launched new cities and grew ridership through aggressive street teams and manual tactics.

  • Early creativity and localized experiments, like Uber Ice Cream trucks, puppies/kittens for rent, and targeted promotions, helped tip individual markets past the critical threshold despite being unsustainable long-term.

  • Uber created a culture that rewarded city teams for quickly executing and iterating on new ideas to overcome cold start problems in each new location.

  • Similar early hustle through personal networks, meeting customers where they are, and press was important for fast-growing B2B startups to find their first customers.

  • Dropbox faced the challenge of maintaining momentum and growth after initially solving its Cold Start Problem and reaching escape velocity. It needed to continue innovating and evolving to fuel further growth.

  • Dropbox focused on building out more collaborative and productivity features like version history, comments, previews and public links to files/folders to increase user engagement and cross-use. This expanded its utility beyond basic file syncing.

  • It cultivated atomic networks and viral growth within organizations by enhancing its sharing and collaboration capabilities among work colleagues. Many businesses informally or formally adopted Dropbox enterprise-wide.

  • Dropbox also experimented with new revenue streams like referring business users to paid Dropbox for Business plans. It tested a referral program for existing users to earn potential payouts.

  • While facing competition from Google Drive, Box, Microsoft and others, Dropbox maintained its early innovator advantage through continual product improvements and its massive network effects with over 500 million users by its successful 2018 IPO.

  • By focusing on fueling further growth after initially solving its cold start problem, Dropbox scaled rapidly and became one of the fastest SaaS companies to reach $1 billion in annual recurring revenue. Its growth trajectory looked like a classic hockey stick.

  • In 2012, Dropbox had grown to 100 million registered users but needed to focus on monetization to justify its $4 billion valuation from investors.

  • It had a small sales team focused on self-serve upgrades, but needed to better target large enterprises for recurring revenue.

  • Cloud hosting costs were ballooning as usage grew exponentially, forcing Dropbox to build its own data centers to save costs.

  • A Growth and Monetization team was formed to directly drive new revenue opportunities through products, pricing, and targeting high-value users like businesses.

  • Data analysis showed some users were more valuable than others based on collaboration and sharing patterns. Priority shifted from all users to “high-value actives.”

  • Dropbox realized many valuable users were businesses and started building more enterprise features and targeting companies directly through existing user networks.

  • File type analysis showed documents, spreadsheets and presentations - not photos - were driving the most engagement and collaboration within businesses.

  • This prompted a strategic shift to focus on serving the business sector with collaboration tools for core office files.

  • Dropbox’s journey from founding to IPO in 2018 involved important middle chapters beyond just the origin story and IPO.

  • During the years before and after 2012, Dropbox learned about its most valuable users, introduced key enterprise features, and expanded marketing efforts. These helped scale the network effects and reach “escape velocity.”

  • Escape velocity refers to sustaining high growth over time by amplifying network effects. This is harder than the initial startup phase and requires extensive coordinated efforts across thousands of employees.

  • The textbook definition of network effects is simplified and vague. To be actionable, it needs to broken down into the “trio of forces” - engagement, acquisition, and economic effects.

  • The engagement effect is about stickiness and usage levels increasing as the network grows. Acquisition effect is viral growth powering new customer acquisition. Economic effect is how business models improve through premium features, pricing, etc. as the network scales.

  • Focusing product efforts on these concrete network effects allows teams to meaningfully strengthen them through specific initiatives like referral programs or elevated collaborative features.

  • Engagement, acquisition, and economic (monetization) effects are key metrics that product teams care about - active users and revenue.

  • Active users are a function of new user signups and retention of existing users. Revenue depends on active users and average revenue per user (ARPU).

  • Networked products can leverage network effects to increase these metrics over time through a “growth accounting equation.”

  • As the network grows and hits “escape velocity,” the engagement, acquisition, and economic effects become more powerful, driving up new users, retention, and monetization.

  • This creates a accumulating advantage for networked products over traditional ones that don’t benefit from network effects.

  • While the effects can be described separately, in practice they reinforce each other - stronger engagement drives more viral growth, etc. Amplifying one effect often increases the others.

So in summary, network effects are uniquely able to increase the key metrics of active users and revenue over time through reciprocal reinforcement of the engagement, acquisition, and economic networks as the user base grows. This creates accelerating growth for networked products.

  • Teams need to segment users into higher and lower value groups, similar to what Dropbox did. Categorization shouldn’t just be based on monetary value, but other factors like frequency of use, lifetime value, use cases, etc.

  • LinkedIn segmented users based on engagement frequency over different time periods (e.g. active in last 7 days, 6 days, etc.). This allowed them to understand each group’s needs and find ways to increase engagement.

  • The “levers” to increase engagement are different for infrequent versus power users. Early users may just need more connections, while power users need advanced features.

  • Studying power/HVA users can reveal what makes them unique, but you need A/B tests before assuming a correlation is causation and applying it to all users.

  • Teams can introduce features, content, and incentives tailored to each group based on segmentation insights to encourage higher-value actions like installing across multiple devices or collaborating more.

  • The goal is to accelerate the “engagement loop” for a product - making each step like content sharing/viewing/interaction more effective to increase overall engagement and network effects.

  • Reactivating lapsed users can be a major growth lever for mature products with large numbers of past users. Simple personalization like sending weekly digests or notifications of friends’ activity can significantly boost reactivation rates.

  • Making the password recovery process easy is also important, as many users fail recovery attempts. Reactivation should be treated with the same priority as new user onboarding.

  • Viral user acquisition, also known as the Acquisition Effect, is one of the most powerful forces in tech. It allows networks to attract new users as they grow in scale.

  • PayPal is a prime example, originally growing slowly but then spreading virally through eBay. Offering referral bonuses accelerated this viral spread, recouping the costs through increased engagement on the platform.

  • Truly viral growth is product-driven and leverages existing network connections, unlike promotional “viral” campaigns. It allows free, scalable user acquisition without expensive marketing. This viral user acquisition is critical for startups to fuel growth beyond initial users.

The key takeaways are that reactivation and viral user growth can both be major growth levers through personalization, easy processes and leveraging network effects within existing user bases. This helps avoid overspending on new user acquisition.

  • Networked products like Dropbox, PayPal, Slack, and Instagram are uniquely positioned to embed viral growth directly into the product experience through features like folder sharing, user payments, inviting colleagues, and sharing photos with connections.

  • This “product/network duo” allows the product to attract users to the network while the network brings more value back to the product, fueling viral growth.

  • Viral growth loops can be broken down step-by-step and each step optimized through A/B testing to improve conversion rates, such as Uber optimizing their driver referral process.

  • The “viral factor” measures the effectiveness of viral growth by calculating the ratio between user cohorts, with higher ratios indicating more efficient growth.

  • Tracking and improving the viral factor through product changes and testing is key to accelerating acquisition. A factor above 1 indicates true viral growth.

  • Long-term retention also strengthens viral loops by fueling more frequent sharing and invitations over time. One-time users are less effective for growth.

  • While acquisition can grow independently, true success requires both acquisition and engagement effects working together to create a sustainable network.

  • Early business networks like credit bureaus saw a “data network effect” where more data from more sources improved risk analysis and lending decisions, attracting even more members and data in a virtuous cycle.

  • As Uber grew, it shifted its focus from “growth at all costs” through subsidies to improving “efficiency over subsidy” by refining its unit economics and driver incentives to reduce spending.

  • Launching a new network often requires subsidies upfront like funding content creators, but as the network aggregates more niche audiences, it can support content in a more efficient way than early subsidies.

  • The “economic effect” refers more broadly to how a network’s business model, like its ability to do premium pricing or take on more risk, improves over time as the network grows in size and strength due to network effects.

So in summary, the economic network effect describes how a business model and unit economics naturally get more efficient and profitable as the user network expands due to network effects, reducing the need for early heavy subsidies to kickstart the network. Data-driven improvements and a larger customer base unlock these advantages.

Here is a summary of the key points regarding the economic network effect from the passage:

  • The economic network effect refers to how the business model and economics of a networked product or service can improve as the network grows larger over time.

  • As the network grows, it can more efficiently subsidize participation through things like promotions, discounts, and incentives since the costs are spread over more users. This leads to lower subsidies or “burn” per user.

  • A larger network is also able to generate more demand over time through things like increased density of users. This allows providers to earn more revenue per hour and rely less on subsidies.

  • Conversion rates from free to paid users can also increase as the network grows since premium/paid features become more valuable as more people participate in the network.

  • The economic network effect provides a competitive advantage for larger networks as it strengthens their business model and makes them harder to displace compared to smaller competitors. It allows them to maintain pricing power and profitability as the network scales up.

  • This network effect, along with others around acquisition and engagement, help explain why many digital platforms and services experience strong dominance once they reach a certain scale in the market.

  •, the predecessor to Twitch, had grown to millions of users but hit a growth ceiling by late 2010. The company was profitable but not growing at all.

  • To break through this ceiling, the team took a few approaches - developing a mobile video app (Socialcam), continuing work on the core product, and spinning out a new focus on gaming under Emmett Shear and Kevin Lin called (which later became Twitch).

  • Twitch focused specifically on gamer streamers rather than general audience. Key changes included better streaming quality tools, organizing content by game, sorting by popularity, and monetization like tips. Partnerships and events like TwitchCon were also developed.

  • These changes made streaming engaging even for those with small audiences. The goal was to attract streamers away from YouTube and help them build audiences and eventually make streaming a full-time career through the platform.

  • By focusing intensely on serving streamers, Twitch was able to break through’s growth ceiling and scale rapidly, eventually becoming a major platform and being acquired by Amazon for $970 million.

  • Twitch’s original strategy of focusing on helping streamers build content and audiences was largely successful even after changing their name from Xarth.

  • Within a month of launching with this focus on streamers, Twitch had 8 million unique viewers. Within a year it doubled to 20 million viewers, and continued doubling as it grew to become one of the most highly trafficked websites.

  • Individual streamers now have over 5 million followers and can earn millions in revenue per year. The name Xarth also lives on as the name of Twitch’s main boardroom.

  • The story shows how transforming a product like into focusing more narrowly on games streaming unlocked tremendous growth beyond the initial success. Addressing the needs of content creators/streamers was a key part of Twitch’s strategy and success.

  • All networked products face eventual challenges of slowing growth as the market becomes saturated. The chapter goes on to discuss how products hit a “ceiling” and the challenges of dealing with plateaus in growth over time.

Here are the key points about the Rocketship Growth Rate:

  • It refers to the precise pace of growth needed for a startup to break out and reach a billion dollar valuation within 7-10 years.

  • For SaaS companies specifically, it involves tripling revenue each year until hitting $144M in annual recurring revenue. This trajectory allows hitting $100M in annual revenue needed for a $1B valuation within 6-9 years.

  • The growth rate can be calculated using an equation that factors in the target revenue, starting revenue, and number of years to achieve the target.

  • For marketplaces, the target is often $200M in net revenue within 6 years from $1M starting revenue, requiring average annual growth of 2.4x.

  • Achieving rocketship growth is very difficult as growth rates naturally decrease over time. Teams have to continuously increase ambition, resources, and market focus to sustain very high triple-digit percentage growth.

  • Hitting growth ceilings is dangerous as it leads to loss of momentum, defections, and difficulty raising capital at higher valuations. All continue the pressure to reignite growth.

  • Even successful rocketships will see growth rates slow over time as markets saturate. Maintaining high growth for 7-10 years straight is exceedingly challenging.

So in summary, the rocketship growth rate defines the exponential growth trajectory needed to build a billion dollar startup, but sustaining that level of growth over the long run is very tough to achieve.

Networked products experience a phenomenon called network saturation as they grow large, which causes their growth rate to plateau over time in addition to traditional market saturation.

Network saturation occurs as the incremental value of each new connection to the network diminishes. For example, on a marketplace like eBay, the 10th listing of a product is not nearly as useful to a search as the 1st few listings. On social networks, each additional friend beyond a certain number contributes less and less to engagement.

As a result, the exponential-looking growth curve of a networked product is actually composed of many individual lines of business layered on top of each other to counteract saturation effects. For eBay, this included launching buy-it-now listings, international expansion, payments integration, and seller tools.

To keep growing, networked products must constantly evolve their product, target markets, and feature set to bring in new layers of users and revenues. This helps offset both network saturation within the core user base and traditional market saturation as the total available market is tapped.

  • Bangaly Kaba, former head of growth at Instagram, developed the concept of “Adjacent Users” - users who are aware of a product but don’t fully understand it or see how it fits into their lives.

  • When he joined Instagram in 2016, growth had slowed despite 400 million users. He took an approach of systematically evaluating Instagram’s “network of networks” to identify subpar user experiences.

  • There were likely multiple sets of “nonfunctional adjacent networks” at any given time, requiring different solutions like improving features for low-end Android phones or attracting more popular creators/content.

  • Solving the experience for one set of adjacent users would reveal the next set, and the process had to repeat itself as the product and markets evolved over time.

  • New “formats” like Stories allowed existing Instagram users to engage in new ways, increasing usage without needing new users.

  • Expanding to new geographies like international markets provided fresh users but also required solving the “cold start problem” of launching in each new region.

  • Adjacent regions were easier to expand to, while faraway geographies required essentially restarting the network from zero in addition to localization challenges. Achieving growth through new formats, geographies, and adjacent users is an ongoing process of evaluating networks to identify limitations.

  • Marketing channels like paid ads, email, social media inevitably degrade over time as click-through rates and engagement decrease, regardless of the specific channel. This occurs due to factors like consumer acclimation, saturation, and users tuning out messages.

  • This “Law of Shitty Clickthroughs” poses an existential threat to products relying on network effects, as degrading marketing channels can significantly weaken the loops that bring in new users and drive engagement. A 50% drop in invite conversions through email, for example, could result in an 80% reduction in total new users.

  • As acquisition channels plateau, companies need to layer on new growth strategies to sustain momentum. Simply acquiring startups can help introduce new avenues for growth. However, acquisitions also face challenges like integration difficulties, high costs, and government scrutiny. Diversifying growth strategies remains an ongoing challenge for successful companies.

The key points are that marketing effectiveness inevitably declines over time, endangering network effects, and companies must continually explore new approaches to user acquisition and engagement as initial channels degrade. Acquisitions can provide new growth, but come with their own obstacles to overcome as well.

  • Products with network effects often face challenges in keeping both sides of their network (hard and easy sides) aligned and satisfied over time.

  • The hard side, like Uber drivers or sellers on eBay/Airbnb, tends to be a scarcer group that provides a vital service but can feel taken advantage of by changes that benefit the larger easy side (riders/buyers).

  • As the hard side gains importance to the business, tensions can rise if their needs and treatment by the company do not improve accordingly. Protests and revolts by core hard side members pose an existential threat.

  • Well-organized revolts by major players on the hard side have the potential to seriously damage or even kill a product, as seen with Vine losing top creators to coordinated demands.

  • Maintaining balance between both sides of the network and adapting to the hard side’s changing role over time is a constant challenge, especially as their scarcity increases their influence and bargaining power within the system.

  • Vine turned down a monetary request from some of its top creators to monetize the platform more, and a few years later shut down the service entirely.

  • Many digital networks see concentration where a small percentage of users drive the majority of activity and revenue. For example, the top 15% of Uber drivers accounted for over 40% of trips. The top iOS apps are mostly from a few large developers like Google, Facebook, etc.

  • This concentration occurs through positive feedback loops that reward high-quality content and engagement. Good creators/sellers/providers get more exposure and success, while poor ones see less engagement and may eventually stop using the platform.

  • Networks want to encourage professionalization, where successful amateur users transition to full-time professional roles. This improves capacity and allows the network to scale. Methods include training, enterprise features, customer success teams, etc.

  • Uber’s XChange leasing program, which financed vehicles for drivers, lost $525M as it attracted less reliable drivers who did not make payments and misused the cars. Professionalization efforts carry risks if not properly managed.

  • Overall, professionalization leads to more quality, consistency, and scale, though it may eventually misalign incentives as power users demand more concessions from the network. Networks have no choice but to embrace this evolution.

  • Microsoft and Nintendo initially only offered their apps and content on their own platforms, but eventually realized mobile was too big to ignore and released apps for iOS/Android.

  • As networks grow large and diverse, they’re often described as “economies” like the gig, attention, and creator economies, with whole ecosystems emerging around sectors.

  • For networks to scale, they must focus on scaling up the “hard side” (users/drivers/creators) as new user growth slows. This requires professionalizing onboarding and support to reduce churn.

  • Usenet was one of the earliest online communities and faced many issues modern networks do, like spam, trolling, and unmanaged growth.

  • Its collapse was hastened by the “Eternal September” event in 1993, when millions of new AOL users joined without understanding Usenet culture, overwhelming the system and changing its nature permanently as it struggled to manage the influx.

  • Usenet faced issues with spam, pornography, and trolling as it grew in scale, contributing to its decline starting in 1993. This serves as a lesson for other networked products.

  • As networks get bigger, they experience “context collapse” - when many different social contexts converge into one shared space. This inhibits people from freely sharing content as they don’t know how it will be perceived by different audiences.

  • Early niche networks start with a shared culture/norms but as outsiders join, the context collapses and makes long-time users uncomfortable sharing. This drives anti-network effects that counteract growth.

  • Products like messaging apps are resistant to context collapse as users communicate in separate isolated threads. Features that allow splitting a large network into smaller topical communities or channels within an app can also help preserve separate contexts.

  • Downvoting, moderation and other community feedback tools help address issues like spam and trolling that contributed to Usenet’s decline by allowing users to collaboratively shape the content and experience.

  • YouTube faced the problem of having too many videos as it grew, making it difficult for users to find what they wanted to watch. This is an example of a broader issue called “overcrowding” that can hurt network effects.

  • When a networked product has too much content, comments, threads, emails, followers, players in a game, it becomes unusable as there is too much to deal with.

  • YouTube co-founder Steve Chen was interviewed about how they scaled the product to keep content discoverable as the number of videos increased over time.

  • YouTube started in 2005 and initially only allowed dating video uploads, but soon expanded as users uploaded all types of videos. As the volume grew, discoverability became a challenge they had to address through features like search, categories, recommendations etc.

  • Addressing overcrowding and keeping content discovery functional is a key challenge for many networked products as user bases and content volume increases massively over time.

  • ite was originally intended as a YouTube-style dating site where people could upload videos of themselves as part of their profile.

  • Within a few weeks, the founders realized it would work better as a general video sharing site without the dating focus.

  • The first video uploaded was a 19-second clip of YouTube co-founder Jawed Karim at the zoo.

  • In the early days, they focused on getting content and solving the “cold start problem” by getting people to upload and share videos. There was no sophisticated recommendation system.

  • As the site grew, they started organizing videos into categories, country-specific top video lists, and trending lists chosen by YouTube.

  • User comments were also a big part of early participation and growth on the site.

  • Discovery became a challenge as the number of videos grew tremendously quickly within the first year, surpassing all of their internal milestones.

  • They had to transition from manual curation to popularity-based sorting to algorithmic recommendations to deal with the overcrowding issue.

  • Popular creators with a lot of early attention faced less challenges gaining more views and standing out, creating obstacles for new creators breaking in.

YouTube and other popular platforms like LinkedIn and TikTok face the challenge of overcrowding as their networks and amount of content grow tremendously over time. This can make it hard for users to find relevant content and for creators and buyers/sellers to connect with their intended audiences.

To address this, the platforms rely heavily on algorithms and machine learning to surface the most relevant content and connections. Features like search, recommendations, related videos, people you may know, and personalized feeds aim to alleviate overcrowding by better matching users with what interests them. These relevance algorithms are shaped by both explicit user signals like likes and follows as well as implicit signals from users’ broader behaviors.

While algorithms help, they are not a perfect solution and come with their own unintended consequences if not optimized properly. The platforms are in a constant battle against overcrowding as their networks continue growing rapidly. Early on, YouTube and others relied more on basic user-driven organization, but over time built more sophisticated algorithmic tools to handle the massive scales involved. Keeping networks balanced and healthy as they grow is an ongoing challenge even for the largest platforms.

  • In 2011, Airbnb encountered its first direct competitor called Wimdu, which launched focused on the European market.

  • Wimdu was founded by Rocket Internet, known for copying successful US businesses. Wimdu’s website and messaging closely mimicked Airbnb.

  • Wimdu received $90 million in funding, the largest investment in a European startup at the time. It rapidly hired 400 employees and listed thousands of properties.

  • Wimdu posed a serious threat as it moved aggressively into Europe. It built listings by scraping Airbnb and convincing Airbnb hosts to also list on Wimdu.

  • Within a year, Wimdu claimed 50,000 listings across 100+ countries and $130 million in projected annual revenue. However, its listings were of lower quality compared to Airbnb.

  • While Wimdu grew quickly, it struggled to deliver the same experience quality as Airbnb. By 2014 it began laying off employees and accepting it lost the European market lead.

  • Airbnb saw Wimdu as its first true competitor and rallied to scale its operations in Europe to counter Wimdu’s rise.

  • Airbnb was facing competition in Europe from Wimdu, a home sharing platform focused solely on that region.

  • To fend off Wimdu, Airbnb mobilized its product teams to rapidly improve support for international regions. This included translating the platform, adding more currencies, buying local domain names, and improving the mobile experience.

  • Airbnb also scaled up paid marketing in Europe and hired its first Head of International to accelerate expansion. They launched localized websites and an integrated marketing campaign across several European countries over four months.

  • This coordinated internationalization of the product and expansion into Europe through marketing and offices helped Airbnb defeat Wimdu by gaining density in the European networks faster than the competitor. Having a global network gave Airbnb an advantage over Wimdu, which was focused only on Europe.

  • The story demonstrates how network-based platforms can defeat competitors by quickly scaling their own network presence in key regions before rivals do. It’s a battle of gaining enough localized network density versus a competitor.

  • Networked products like marketplaces, social networks and collaboration tools face winner-take-all competition due to network effects. As one product gains dominance within individual networks, it can start dominating the overall market.

  • It’s not about who has the biggest network or most features initially. Quality of the network matters more. Startups can disrupt larger players by targeting underserved niches and building high quality, engaged networks in those areas.

  • All networked products have network effects, not just the incumbent. Competition depends on who does the best job of scaling network effects through acquisition, engagement and monetization.

  • As competition intensifies, it becomes zero-sum and can trigger a vicious cycle where losing networks experience an exponential collapse as users and value flee to the winner. This is what ultimately led to the demise of competitors like MySpace.

  • Large and small players experience asymmetric competition due to different stages of the cold start framework. Startups can focus on top-line growth while larger players worry more about profitability and gravitational pull as networks mature. Both will use different competitive strategies reflecting these asymmetries.

  • Startups have fewer resources than larger companies, but they have advantages in speed and flexibility that allow them to quickly pivot and try different business models until they find product-market fit. This trial-and-error process is part of how startups like YouTube, Twitter, etc. were able to grow.

  • Larger companies have more resources, manpower, and existing assets/customers. However, it’s harder for them to execute quickly, take risks, and launch new initiatives due to processes, risk aversion, and the need to align new products with the existing business model. As companies grow very large with tens of thousands of employees, this effect becomes more pronounced.

  • Small startups can “cherry pick” underserved audiences from large incumbent companies like Craigslist by focusing on a specific use case or category. If they’re able to attract those users and reach critical mass, they have a new standalone network that is hard for the incumbent to fight. This allows startups to effectively “unbundle” larger networks.

  • When looking for opportunities to challenge incumbents, startups should look for the “soft spots” - parts of the incumbent’s network that are most vulnerable due to poor service, lack of features, disengaged users, etc. Focusing on these niches allows a startup to more easily access network effects and grow their own “atomic” standalone network.

  • The “Big Bang Launch” is a strategy used by larger companies to quickly overwhelm opponents with a big launch using their scale and resources. However, this often fails for networked products.

  • A classic Big Bang Launch involves a major announcement event, widespread media coverage, promotion across the company’s main products/channels to send over many users at once, and large marketing pushes.

  • The intent is to launch the best possible product to as many people as possible from the start, building the network from the top influencers and nodes down.

  • However, for networked products, this approach is often a trap. While convenient for established companies due to their resources, it does not account for the unique challenges of building networks, which require more organic, grounded growth to reach critical mass.

  • A notable example failure is Google+, which launched with much fanfare but failed to gain sustainable traction as a social network due to over-reliance on a Big Bang Launch approach rather than grassroots network formation.

  • The article argues that Google+‘s wide launch strategy was exactly the wrong way to build a successful social network. A wide launch leads to many weak networks that are not stable on their own.

  • While Google+ claimed hundreds of millions of users early on, these numbers masked that the network was largely empty and inactive. Users signed up but did not engage meaningfully on the platform.

  • For a network to succeed, it needs to start with focused, dense, engaged atomic networks that grow organically through viral effects. A big bang launch produces many scattered, unconnected users before the product is ready to sustain viral growth.

  • Contrast this with how Snapchat, Instagram, Twitch, and TikTok succeeded - by innovating in their features for creators first within smaller communities before widespread adoption.

  • Starting small allows a product to gain traction through word-of-mouth within existing networks, giving new users value from connected others already on the platform. It’s better to build quality networks from the bottom-up than rely on top-line metrics from a big launch.

  • While counterintuitive, history shows the largest networks like Facebook and Uber started by focusing on small, niche atomic networks first before widespread scaling - despite initial doubts about addressing small markets. This paradoxical approach is necessary to leverage network effects.

  • At the time, Airbnb was focused on low-end accommodations like air mattresses and breakfasts, not more developed accommodation listings. They had ideas to expand but had not made much progress yet.

  • Investors could not envision air mattresses as the next hotel room and did not pursue investing in Airbnb at that early stage.

  • However, others saw the potential in Airbnb’s team and vision to build an “eBay of spaces” and provided funding.

  • It is easy to underestimate early-stage startups that are focused on narrow initial products or networks, but fail to see how they could expand into much larger adjacent markets through network effects.

  • Airbnb eventually disrupted the entire hotel industry despite starting small with air mattress listings. Other examples include chat products that start small but dominate communication markets.

  • Larger, established companies often try to “jump from zero to the tipping point” with a big launch due to internal pressures, rather than starting small like startups do. However, this path often leads to failure due to inability to establish early product-market fit.

  • Uber fiercely competed against rivals like Lyft and Sidecar through coordinated strategies developed in “war rooms” or competitive meetings. This combined product improvements, incentives for drivers and riders, and other tactics applied at a hyper-local level to gain dominance in key cities.

  • Uber focused much of its competitive efforts on targeted bonuses and incentives for drivers, as drivers were primarily motivated by earnings.

  • Their bonuses were aimed at “flipping” dual-app drivers (those using multiple rideshare apps) from rivals’ networks to Uber’s exclusively through large, special weekly bonuses that required driving many hours solely on Uber.

  • Uber developed sophisticated methods to identify dual-app drivers through manual tagging, behavioral signals, and machine learning models. These drivers were then targeted with various incentive offers.

  • The goal of the incentive structures was to get drivers to commit to Uber for the full week by hitting trip quotas that would be difficult to achieve across multiple apps. Over $50M/week was sometimes spent on incentives in major regions.

  • While improving the product experience was also important, Uber’s main competitive lever was using targeted financial incentives to gain an advantage in the “hard side” of the network by attracting key drivers from rivals’ networks.

  • This strategy helped Uber defeat many early competitors but met limitations as rivals like Lyft and DoorDash still achieved success, and Uber exited some foreign markets.

  • When Uber and a competitor have around 50% market share each in a city, or when Uber is the smaller player, it does not benefit from network effects and must differentiate through other means. This is challenging in transportation markets that are essentially commodity services.

  • DoorDash found early success by starting in suburban markets with less competition, building up networks there before expanding into cities. This head start allowed it to surpass competitors in urban areas.

  • Network effects don’t guarantee winner-take-all outcomes. Products compete as interconnected networks, so even dominant networks like Uber’s could only achieve parity with competitors in some cities.

  • Bundling new products with existing networks can help solve the cold start problem by providing initial users and traction. However, the new product still needs to be high quality or users won’t stick around.

  • Examples like Internet Explorer and Bing show that distribution advantages through bundling are not enough - the product itself must be significantly better. Microsoft Office succeeded due to bundled distribution and leapfrogging competitors through product improvements.

  • Modern bundling is more about cross-promotion through clicks, APIs, announcements, etc. to rapidly build new network effects. But this only works if it crystallizes into real engaged user networks for the new product.

  • Instagram was able to greatly increase user retention by tapping into Facebook’s larger social graph and recommending users’ real friends from Facebook to follow on Instagram. This created stronger, denser networks on Instagram compared to just recommending celebrities.

  • Microsoft also leveraged its ecosystem and network effects to compete, not just on product features. It focused on attracting and retaining developers through tools like Visual Basic that made Windows app development accessible. It ensured backwards compatibility so existing apps always worked on new Windows versions.

  • When Netscape launched the first popular web browser, Microsoft recognized the threat and rushed out its own browser, Internet Explorer. It pursued a strategy of embedding internet capabilities directly into Windows apps rather than just the browser. This grew IE’s usage and forced web developers to support it.

  • Bundling strategies like pre-installing IE worked to grow IE’s market share dominant but drew antitrust scrutiny. Bundling can also add clutter to products and result in worse design. It stops working once network effects are overcome by a stronger competitor’s product.

  • The passage discusses network effects and how they have impacted industries like mobile operating systems, web browsers, office suites, and more.

  • Microsoft struggled to replicate its Windows desktop network effects in the mobile world. Google gave Android away for free, driving adoption and eventually dominating the app ecosystem through advertising revenues.

  • Microsoft has also lost the browser market to Chrome and faces challenges in office suites from startups. Bundling strategies like adding chat to Office haven’t led to clear victories over competitors like Slack.

  • Bundling alone is not a silver bullet strategy. Google bundled products like Google+ and Maps without strong retention. Uber bundled food delivery but still trails DoorDash.

  • Many early Uber employees have since started their own companies or joined others, spreading the lessons they learned about network effects, scaling through growth, competing fiercely, and more. Their impact and that of other “alumni” from networked companies will drive innovation.

  • Network effects will become more central across industries as technology transforms the economy in coming years. Cryptocurrency in particular could redefine many categories by infusing network effects throughout software.

  • The author thanks numerous individuals who provided feedback, research assistance, interviews and advice for the book project over multiple years. This includes employees and founders from companies like Uber, YouTube, Slack, Airbnb, Dropbox, etc.

  • They specifically thank their sister, editor Hollis Heimbouch, agent Chris Parris-Lamb and research collaborators like Bubba Murarka for their significant contributions.

  • Many CEOs, product leaders and venture capitalists were interviewed to gain insights about building iconic products and companies.

  • Fellow authors also provided advice as a first-time writer.

  • In conclusion, the author expresses gratitude to all those who helped make the multi-year book project possible through their guidance, feedback and participation in interviews. It seems to have been a major collaborative effort drawing on the experiences of people from across the tech industry.

This passage does not provide a clear summary. It lists references/citations from 64-93 without any context or explanation. A summary would synthesize and discuss the main ideas from the sources, not simply list the references.

Author Photo

About Matheus Puppe