The Cold Start Problem: How to Start and Scale Network Effects

Why this book exists

Network-effect products — products that become more valuable as more people use them — are one of the most important categories in modern tech. The largest companies in the world (Meta, Google, Microsoft, Amazon) have built moats around network effects, and the most exciting startups (Tinder, Zoom, Slack, Notion, Airbnb, Uber, TikTok) have grown by exploiting them. Yet the playbook for actually launching and growing a network-effect product is different from the playbook for traditional SaaS or consumer products, and most of the existing growth literature underweights or ignores network-specific dynamics.

Andrew Chen spent the 2010s in the middle of this category. As VP of Growth at Uber, he led the team that scaled rider and driver networks across hundreds of cities. As a general partner at Andreessen Horowitz, he invested in and worked with dozens of network-effect companies across consumer social, marketplaces, and developer tools. The pattern recognition from both operator and investor perspectives produced a specific framework for how network-effect products work — a framework that diverges meaningfully from the standard PMF-and-growth literature.

The Cold Start Problem is the book that operationalizes the framework. It is the most thorough single resource on network-effect product strategy in print and has become required reading for founders and PMs building anything with network dynamics.

The network effect lifecycle

The book is structured around a five-stage lifecycle that network-effect products move through:

Stage 1: The cold start. The earliest stage, when the network has too few users on one or both sides to be valuable. Users who arrive find the product empty or sparsely populated; they leave; the network never gets going. Most network-effect products die in the cold start.

Stage 2: The tipping point. The network reaches enough density to become self-reinforcing. New users find value because existing users are there; existing users find more value as new users arrive; the network grows on its own momentum rather than requiring founder-led acquisition.

Stage 3: Escape velocity. The network grows through compounding loops — acquisition loops, engagement loops, economic loops — each reinforcing the others. Growth becomes exponential rather than linear.

Stage 4: Hitting the ceiling. The network saturates its initial market, faces engagement quality degradation at scale, attracts regulatory pressure, or encounters competitive response. Growth slows; the team must navigate the new challenges.

Stage 5: The moat. The mature network has built defensible advantages — user base, data flywheel, ecosystem lock-in — that competitors cannot replicate. The network produces dominant market share and sustained profitability.

The lifecycle framework matters because the right strategies differ dramatically at each stage. Tactics that work in the cold start (manual concierge curation, geographic concentration, friend-of-founder seeding) are wrong at escape velocity. Tactics that work at scale (platform partnerships, ecosystem development, regulatory engagement) are wrong in the cold start. The book provides stage-appropriate guidance for each.

The cold start: where most products die

Chen devotes the largest portion of the book to the cold start, because it is where most network-effect products fail. The problem is the chicken-and-egg dynamic: a marketplace needs both buyers and sellers, but neither will join until the other is there. A social network needs people to follow and people to be followed by, but neither shows up first. A communications tool needs people to communicate with, but the first user has no one to message.

The book describes specific cold start strategies that have worked:

The atomic network. Pick the smallest possible network that can be useful and dominate it. Tinder launched at a single college sorority where users already knew each other and could check the app to find matches among their immediate social circle. Once that atomic network worked, Tinder expanded to other sororities, then fraternities, then colleges, then beyond. The atomic network produced the initial density that traditional cold start strategies could not.

The hard side first. In multi-sided networks, one side is usually harder to recruit than the other. In ride-sharing, drivers are harder than riders; in marketplaces, sellers are harder than buyers; in dating apps, the underrepresented gender is harder than the overrepresented. Cold start strategies should focus disproportionately on the hard side because the easy side will follow once the hard side is in place.

Concierge curation. Manually fulfill the network functions in the earliest stages. Airbnb's founders went door-to-door in New York taking professional photos of listings; Reddit's founders created fake user accounts to populate the early site with content; LinkedIn's founders manually built initial professional connections. The manual work is unscalable but bootstraps the network past the cold start.

Single-player mode. Build value into the product for individual users so they get value even when the network is sparse. Instagram's photo filters made the app valuable even when no one was following you; Slack's archiving and search made the tool useful even for a small team. Single-player mode survives the cold start because users do not need the network to find value.

Geographic concentration. For location-dependent networks, dominate one city before expanding to others. Uber launched in San Francisco only; DoorDash launched in Stanford-area neighborhoods only; Airbnb launched in specific neighborhoods of specific cities. Concentration produces local network density that geographic dispersion cannot.

Vertical concentration. For non-location-dependent networks, dominate one segment before expanding. Slack initially targeted tech startups; LinkedIn initially targeted tech professionals; Substack initially targeted journalists with specific large audiences. Vertical concentration is the digital equivalent of geographic concentration.

The cold start strategies share a common pattern: pick a narrow initial network where density is achievable, dominate it, then expand. Founders who try to launch broadly almost always fail; founders who narrow ruthlessly often succeed.

The tipping point

Once an atomic network is dense enough to self-sustain, the product reaches the tipping point. New users find value because existing users are present; existing users find more value as new users arrive; growth begins to compound.

The tipping point is observable in retention and engagement data. Cohort retention curves flatten at a non-trivial percentage; engagement-per-user grows or stabilizes rather than declining; organic referrals begin to outpace paid acquisition. The team can tell they have reached the tipping point because the product begins to grow without their direct intervention.

After the tipping point, the strategic priority shifts from cold-starting to scaling. The team can begin to expand to adjacent atomic networks, knowing that the same pattern that worked in the initial network will replicate (with adjustments) in adjacent ones. The bowling pin expansion strategy from Crossing the Chasm applies, with the additional dimension that each new atomic network needs to be cold-started itself.

The book describes specific signs that the tipping point has been reached and how to verify them quantitatively. Teams that mistake the tipping point — declaring it too early or missing it when it occurs — make wrong strategic decisions. The diagnostic discipline matters.

Escape velocity through compounding loops

After the tipping point, sustained exponential growth requires multiple growth loops that compound. The book identifies several loop types:

Acquisition loops. Existing users bring in new users through referrals, viral mechanics, content sharing, or invitations. Dropbox's referral program, Hotmail's email signature, Pinterest's content shareability, and LinkedIn's connection requests are canonical examples.

Engagement loops. Existing users return more frequently because of network activity. Facebook's notifications, Instagram's feed updates, Slack's real-time messages, and TikTok's algorithmic feed are engagement loops.

Economic loops. The network's economics improve as it grows. Marketplace take rates can rise as the network's value to participants grows; ad rates can rise as audience attention concentrates; subscription pricing can rise as the network's value justifies higher prices.

Content loops. User-generated content attracts more users, who create more content, which attracts more users. YouTube, Pinterest, Reddit, and TikTok all operate on content loops.

Data loops. More usage produces more data, which improves the product (better recommendations, better search, better matching), which attracts more usage. Google's search ranking, Netflix's recommendations, and Spotify's discover-weekly all rely on data loops.

Escape velocity comes when multiple loops are active simultaneously. A product with only one loop grows but is fragile to disruption of that loop. A product with three or four loops is much more resilient and grows much faster. The strategic question at the escape velocity stage is which loops can be added or strengthened to accelerate compounding.

Hitting the ceiling

Networks do not grow forever. The book identifies several ceilings that network-effect products hit:

Market saturation. The network reaches a significant fraction of its addressable market and growth slows. Facebook in mature markets, Twitter at peak years, LinkedIn for senior professionals all hit saturation.

Engagement quality degradation. As the network grows, the average quality of interactions can decline (more spam, more low-quality content, more bad actors). Twitter's troll problem, Facebook's misinformation problem, and many online communities' descent into toxicity are quality degradation issues at scale.

Algorithm overfitting. Algorithmic feeds optimized for engagement can produce echo chambers, addictive patterns, or polarization that ultimately harm the network. The book treats algorithm management as a strategic concern, not just a technical one.

Competitive response. Once a network reaches significant scale, it attracts competitors. Snap responded to Instagram, Threads responded to Twitter, TikTok competed with YouTube Shorts and Instagram Reels. The competitive response can erode network growth.

Regulatory pressure. Mature networks attract regulatory attention. GDPR in Europe, app store policies from Apple and Google, antitrust scrutiny in the U.S. and Europe have constrained the growth of large networks.

The book provides strategies for managing each ceiling. Market saturation requires international expansion or vertical extension; quality degradation requires moderation investment and community guidelines; algorithm overfitting requires explicit objective rebalancing; competitive response requires differentiation and network depth; regulatory pressure requires proactive engagement.

The ceilings are real and predictable. Networks that anticipate them and prepare navigate past them; networks that ignore them suffer slowdowns or declines.

The moat

Mature networks build moats that competitors cannot replicate. The moat components include:

User base. The accumulated user base itself is a moat. Replicating it would require getting all those users to switch, which is hard when the network they use already has all their connections, content, or transaction history.

Data flywheel. The accumulated data produces a product quality advantage that competitors cannot match without years of equivalent data accumulation. Google's search index, Netflix's recommendation engine, and Tinder's matching algorithm all benefit from data moats.

Ecosystem. The third-party developers, partners, content creators, and complementary services that build around a network become a moat. Apple's app developer ecosystem, Salesforce's AppExchange partners, and YouTube's creator economy are ecosystem moats.

Brand and habit. Users develop habits of using the network for specific purposes that competitors must dislodge. Google for search, YouTube for video, Facebook for personal connections, LinkedIn for professional networking — the habits are hard to break.

Cross-side network effects. When the network is two-sided, both sides reinforce each other and competitors must build both sides simultaneously. Marketplaces, dating apps, and platform businesses benefit from cross-side moats.

The strongest moats combine multiple components. Mature networks like Facebook, Google, and LinkedIn have user base, data, ecosystem, brand, and cross-side moats all working together. Replicating any of them is hard; replicating all of them is essentially impossible without a fundamental technology shift.

Case studies from the book

The book is rich with case studies. Some highlights:

Tinder's atomic network strategy. The founders launched at a single sorority and expanded geographically. The case shows how cold start strategies actually look in execution.

Uber's hard side strategy. Uber focused on driver recruitment in each new city, knowing that riders would follow. The case shows the systematic approach to hard-side cold start.

Slack's single-player mode. Slack invested in features that made small teams valuable users even without large-network density. The case shows how single-player value bootstraps the network.

Zoom's pandemic escape velocity. Zoom benefited from a sudden category shift (COVID lockdowns) that produced an exogenous tipping point. The case shows how exogenous events can compress the lifecycle.

Clubhouse's failure to escape. Clubhouse achieved a tipping point during the pandemic but failed to compound loops and ultimately could not retain users as the world reopened. The case shows that tipping is necessary but not sufficient.

TikTok's algorithm-as-cold-start solution. TikTok solved the cold start by making the algorithm so good that single users could enjoy the content without any social network. The algorithm did the heavy lifting that social cold start usually requires.

Each case study illustrates the framework in action and provides patterns that subsequent network-effect companies have adapted.

What the book does badly

The book has limitations worth naming:

It is heavily consumer-focused. B2B network effects (Slack, Notion, Figma, some dev tools) are covered but less deeply than consumer examples. B2B founders should adapt the frameworks accordingly.

It is light on technical infrastructure. Building network-effect products requires specific technical infrastructure (real-time messaging, scalable matching algorithms, global content delivery) that the book mentions but does not cover in depth.

Some examples have aged poorly. Clubhouse, Vine, and other examples that were prominent at publication have since declined. The patterns are still valid but the specific company stories are now historical curiosities.

It under-covers ethical dimensions. Network-effect products can produce significant social externalities (addiction, polarization, mental health harms) that the book treats as challenges to navigate rather than fundamental concerns. Modern thinking on these issues is more critical.

These critiques do not undermine the core value but suggest readers should bring critical engagement.

How to use the book in practice

The most effective adoption pattern for a founder or PM building a network-effect product:

1. Read the book once cover to cover. Absorb the five-stage lifecycle and the strategies for each stage.
2. Diagnose your current stage. Be honest about where you are. Most teams overestimate their stage.
3. Apply stage-appropriate strategies. In the cold start, focus on atomic networks and the hard side. After tipping, focus on loop compounding. Near the ceiling, focus on quality and competitive defense.
4. Track stage-appropriate metrics. Different metrics matter at different stages. Liquidity in marketplaces, content density in social, transaction frequency in commerce networks.
5. Re-read relevant chapters as you progress. As your product moves through stages, return to the chapters most relevant to your current stage.

Founders who follow this pattern make better strategic decisions for their network-effect products than founders who rely on traditional growth literature alone.

The book's place in the modern PM and founder canon

The Cold Start Problem is one of the most-recommended books for network-effect founders and PMs. It pairs with:

• Platform Revolution by Parker, Van Alstyne, and Choudary — the strategic framework for platform businesses.
• Modern Monopolies by Moazed and Johnson — the dynamics of platform businesses at scale.
• Hacking Growth by Sean Ellis and Morgan Brown — the cross-functional growth methodology that complements network-specific tactics.
• Crossing the Chasm by Geoffrey Moore — the broader technology adoption framework that network-effect products operate within.

Together these texts form a coherent curriculum for network-effect product strategy.

On the specific applicability to AI products

A category not central to the book but increasingly important: AI products can have meaningful network effects through data accumulation. More users producing more queries produce more training data, which improves the model, which attracts more users. ChatGPT's growth has been partly a network-effect story.

For AI PMs, the book's frameworks apply with adjustments. The cold start for an AI product is solved by the underlying model quality rather than by network density (the model is useful even without other users), but the data flywheel that improves the model over time is a network effect. The escape velocity stage involves compounding loops that include both traditional acquisition loops and the data loop.

AI PMs should pair this book with current AI strategy resources to navigate the AI-specific dynamics.

On the moral dimensions of network-effect work

A topic the book underemphasizes but which deserves explicit attention: network-effect products have produced some of the most consequential moral and social impacts of the past two decades. Social networks have shaped elections, mental health crises, and global information ecosystems. Marketplaces have transformed labor markets, sometimes in ways that benefit workers and sometimes in ways that harm them. Communications platforms have enabled both productive collaboration and bullying campaigns.

The frameworks in the book are morally neutral — they describe how networks work, not whether they should be built. Founders and PMs building network-effect products should bring their own moral reasoning to the strategic decisions. What kind of network do you want to build? What externalities will it produce? What responsibilities does the team have to its users and to society broader?

These questions are not addressed in the book but should be in the practitioner's head. Networks shape society; the people who build them carry responsibility for what they create.

Closing thought

Network-effect products are among the most powerful business categories ever created and among the most strategically distinctive to navigate. The standard growth and PMF literature treats network dynamics as a special case; The Cold Start Problem treats them as the main case and provides the dedicated playbook.

For founders and PMs building anything with network effects, this book is essential reading. The frameworks have been validated across dozens of successful network-effect companies and continue to inform the strategy of every new entrant. Read it once for the lifecycle and the strategies; return to it as your product moves through stages; supplement with current resources for emerging dynamics.

The book is one of the most useful additions to the PM canon of the 2020s. The category it addresses — network-effect products — remains one of the most important, and the framework it provides remains the best operational guide.

A worked example: launching a marketplace

Consider a founder launching a marketplace for local services (home cleaning, lawn care, handyman work). They apply the framework:

Stage diagnosis: the launch is in the cold start. There are no users on either side; both supply (service providers) and demand (homeowners) need to be cold-started simultaneously.

Atomic network selection: the founder picks a single small town as the launch market. The town has roughly 30,000 households (enough to support meaningful transaction volume) and a connected community (where word-of-mouth can propagate quickly).

Hard side identification: service providers are the hard side. Homeowners can sign up easily but will not stay engaged without providers; providers must commit time to the platform without certainty of demand.

Cold start strategy: the founder personally recruits 30 service providers in the launch town through door-to-door outreach and local business association partnerships. The founder provides white-glove onboarding, helps providers set up profiles, and guarantees a minimum volume of bookings in the first month. The investment in the hard side is intensive but produces the initial supply density.

Demand bootstrapping: the founder advertises locally and offers a 50% discount to homeowners for their first booking. Initial bookings are subsidized; the goal is to seed the network with successful transactions that produce reviews and word-of-mouth.

Tipping point pursuit: the founder tracks supply density (number of active providers), demand frequency (bookings per week), and net growth (new providers and homeowners minus churned ones). After three months, the metrics indicate the network is approaching the tipping point — providers report enough bookings to stay engaged without founder intervention, and homeowners book without subsidy.

Adjacent expansion: with the launch town past the tipping point, the founder expands to two adjacent towns using a similar playbook but with reduced founder time per market (the learnings from the first town allow more efficient launch).

Loop compounding: over the next year, the founder builds multiple loops — a referral program for providers and homeowners, an engagement loop where homeowners are reminded to book follow-up services, an economic loop where the platform's take rate increases gradually as the network's value to providers grows.

Result: after 18 months, the marketplace covers a region of 15 towns, has thousands of providers and tens of thousands of homeowners, and is growing exponentially. The cold start playbook produced the foundation; the subsequent stages produced the scale.

This pattern recurs in successful network-effect launches. The book provides the manual; the founder must execute with discipline.

A closing reflection on the importance of the category

Network-effect businesses produce winner-take-most or winner-take-all outcomes more often than any other business category. The discipline of building them well is therefore one of the highest-leverage skills in modern technology. Read this book seriously and let it shape your strategic thinking; the long-term return on the time investment is exceptional.

On the recommended pairing with adjacent reads

For a complete network-effect-and-platform curriculum, pair this book with: Platform Revolution by Parker, Van Alstyne, and Choudary for the broader platform strategy framework; Modern Monopolies by Moazed and Johnson for the platform economics depth; The Power of Platforms by Cusumano, Gawer, and Yoffie for the academic treatment; and Sangeet Choudary's Platform Scale for the design patterns. Together these texts cover the full theoretical and operational landscape of platform and network-effect businesses.

On the personal toll of network-effect work

A topic the book mentions only in passing but which deserves explicit acknowledgment: network-effect product work is unusually demanding. The cold start phase requires intense founder-led effort that does not scale. The escape velocity phase requires managing exponential growth that breaks most systems and people. The ceiling phase requires navigating complex strategic challenges with high stakes.

Founders building network-effect products should expect multi-year periods of high intensity. The work is rewarding when it succeeds — the products that win in this category produce enormous impact and value — but the path is harder than most founders expect. Personal sustainability practices, strong co-founder relationships, and realistic expectations matter.

The book's framework is helpful operationally; the personal dimension is on the founder to navigate. Read other resources (Jerry Colonna's Reboot, Tony Schwartz's energy management work, Reid Hoffman on resilience) for the personal side.

On the relationship to investor expectations

Network-effect products attract specific kinds of investors with specific expectations. Investors in network-effect companies typically expect: a clear theory of the network mechanic and why it produces a moat, evidence of cold start progress before significant funding, a credible path from initial network to scale, and unit economics that work at scale even if not at the cold start.

Founders raising for network-effect products should be able to articulate the network theory, the lifecycle stage they are in, the strategies they are using, and the expected progression. Investors who specialize in network effects (like Andreessen Horowitz, where the author worked) ask sharper questions than generalist investors; founders who have read this book can answer those questions credibly.

For PMs joining venture-backed network-effect companies, understanding investor expectations also matters. The investor pressure shapes strategy in ways that PMs without context may not recognize. Reading the book provides that context.

On the rhythm of network expansion

A pattern worth highlighting: successful network-effect companies expand on a rhythm. They dominate the first atomic network, then expand to a small number of adjacent networks, then build the infrastructure for broader expansion, then scale broadly. The rhythm typically involves periods of intense expansion followed by periods of consolidation.

Trying to skip the rhythm — expanding to many networks simultaneously before any is dominated — usually produces shallow penetration across many areas instead of dominant position in any. The book's recommendation: respect the rhythm even when it feels slow. The discipline of sequential expansion produces stronger long-term outcomes than parallel expansion.

For founders impatient to grow fast, the rhythm is uncomfortable. But the founders who maintain the rhythm consistently produce better outcomes than the founders who try to compress it. Network effects do not compress; they unfold at their own pace.

On the role of timing and macro context

A topic the book covers somewhat lightly but which is worth expanding: timing matters enormously for network-effect products. Some categories that failed in one decade succeed in the next when conditions change. Video calling failed many times before Zoom; social audio (Clubhouse, Twitter Spaces) succeeded briefly and faded; collaborative work tools that struggled pre-pandemic exploded during it.

The macro context — technology readiness, user behavior changes, regulatory landscape, economic conditions — shapes whether a given network-effect product can succeed at a given time. Founders should think carefully about whether the macro context favors their specific category and adjust strategy accordingly.

This does not mean waiting for perfect conditions; perfect conditions never arrive. It does mean being honest about whether the conditions are favorable enough to make the bet worth it. Some bets are better made later; some bets are best made now even with imperfect conditions.

Annotated highlights worth marking

• The five-stage lifecycle framework — the conceptual heart of the book.
• The atomic network and hard-side strategies in the cold start chapter.
• The tipping point chapter and the quantitative signals that indicate it.
• The compounding loops chapter on escape velocity.
• The case studies of Tinder, Uber, Slack, and Zoom — the most detailed worked examples.

On network economics and unit economics

A topic the book covers but which deserves more attention than most network-effect founders give it: network-effect products often have non-obvious unit economics. The cost to acquire a user on each side, the contribution margin per transaction, the take rate that is sustainable, and the customer lifetime value all interact in complex ways.

For marketplaces specifically: the take rate must balance vendor willingness to participate (lower take rate is more attractive to vendors) against marketplace economics (higher take rate produces more revenue per transaction). Most successful marketplaces settle into take rates in the 10-30% range depending on category, with the right answer driven by what value the marketplace adds beyond pure matching.

For social and content networks: monetization is usually through advertising, and the relevant unit economics are ad CPM (cost per thousand impressions) and time spent per user. Networks with high time-spent and quality audience attention command premium CPMs; networks with low engagement or low-quality audience command discount CPMs.

For communications and collaboration networks: monetization is usually through subscription, and unit economics depend on willingness to pay and retention. Network depth (how essential the product is to the user's workflow) determines willingness to pay; network breadth (how many of the user's collaborators are on the platform) determines retention.

The book provides general guidance on these dynamics but each category requires deeper category-specific analysis. Founders should pair the book with category-specific economic analysis to make sound monetization decisions.

On the failure modes the book identifies

A useful chapter the book devotes to specific failure modes that network-effect products commonly experience:

The empty room. Users arrive but find the network sparse and leave. The cold start was not solved before broader launch.

The chicken-and-egg standoff. Neither side will join without the other. The team did not pick a hard side to focus on.

The shallow network. The network grew but engagement is low; users sign up but do not return. The product has acquired users without the value mechanism that produces retention.

The toxic network. As the network grew, the quality of interactions declined. Trolls, spam, bad actors, and low-quality content dominated. The team did not invest in moderation early enough.

The locked-in network that stops growing. The existing users are engaged but new user acquisition has stalled. The network reached a ceiling that the team has not addressed.

The disrupted network. A competitor launched a better-designed alternative and existing users started defecting. The moat was weaker than the team assumed.

Each failure mode has specific remedies. Recognizing which mode is killing your product is the first step to addressing it. The book provides diagnostic guidance for each.

On the relationship between network depth and breadth

A subtle strategic dimension the book covers: networks have both breadth (how many users) and depth (how dense and engaged the connections among users are). Strategy must account for both.

Some networks compete primarily on breadth — having a critical mass of users is what matters, and depth follows from breadth. Facebook arguably operates this way: the value is having most people on it, and the depth of interactions is secondary.

Other networks compete primarily on depth — having engaged active users matters more than total count. Substack arguably operates this way: the value is having engaged paying subscribers for specific writers, not having millions of casual users.

For founders, the question is whether to optimize for breadth (acquire many users with shallow engagement) or depth (acquire fewer users with deep engagement). The right answer depends on the product's revenue model and competitive dynamics.

Breadth-focused products usually monetize through advertising (where audience size matters), through indirect network effects (where the presence of users attracts other categories of value), or through commoditized transactions. Depth-focused products usually monetize through subscriptions, through high-value transactions, or through enterprise sales.

The strategic clarity about whether breadth or depth matters more shapes growth investment, content moderation, feature priorities, and metric selection. Conflating the two produces unfocused strategy.

On the difference between viral growth and network effects

A common confusion the book clarifies: viral growth and network effects are different things. Viral growth is a mechanism by which users bring in other users (referrals, invitations, content sharing). Network effects are a quality of the product whereby additional users increase value for existing users.

A product can be viral without network effects (a meme app that spreads but where each user's experience is independent). A product can have network effects without virality (a B2B platform whose value grows with users but which acquires users through sales rather than referrals).

The strongest network-effect products usually have both — viral growth that brings in users plus network effects that make the network more valuable as it grows. The two reinforce each other; viral acquisition feeds the network, and the growing network makes the viral mechanism more attractive (more reasons to invite friends).

For founders building network-effect products, deliberately designing for both viral growth and network effects is the right move. Many products end up with one or the other; the strongest end up with both.

On the difference between hard side and soft side in detail

A topic worth expanding from the cold start chapter: the hard side and soft side dynamic is central to multi-sided network strategy. Examples across categories: in ride-sharing, drivers are hard and riders are soft (drivers must commit time and equipment; riders just download an app). In online dating, the underrepresented gender (typically women on most apps) is the hard side; the overrepresented gender is the soft side. In marketplaces for craft goods (Etsy), makers are hard and shoppers are soft. In professional networking (LinkedIn), recruiters and active job-changers are arguably the hard side.

Identifying the hard side correctly is a strategic first move. Founders sometimes mis-identify the hard side because they project their own preferences (the founder is a buyer-type and assumes buyers are the hard side); the resulting strategy fails because the actual hard side is under-served.

The book recommends specific tests for identifying the hard side: which side has higher acquisition cost? Which side has lower retention if the other side is sparse? Which side requires more handholding to onboard? Which side has more competing alternatives? The side that scores hardest on these questions is the hard side; cold start strategy should focus there.

Once identified, hard-side strategies usually include: subsidies to compensate for early imbalance, white-glove onboarding to overcome friction, guarantees that mitigate risk (Uber's hourly minimums for drivers, Airbnb's host insurance, marketplace seller protection), and outreach that explicitly recognizes the hard side's higher-effort participation.

On geographic vs vertical atomic networks

A practical decision in cold start strategy: should the atomic network be defined geographically (a specific city or neighborhood) or vertically (a specific industry or community)? The book covers both but the choice has implications worth thinking through.

Geographic atomic networks work best for location-dependent services (delivery, ride-sharing, local marketplaces, dating apps in some configurations). The density of users in a single geography produces a usable network even when total user base is small.

Vertical atomic networks work best for non-location-dependent services (B2B SaaS with network effects, professional communities, creator platforms). The density of users in a single industry or interest area produces relevant interactions even when geographic distribution is wide.

Some products benefit from combined geographic and vertical concentration. Tinder launched at specific colleges (geographic + vertical: the college student segment in a specific location). Substack launched with specific writers (vertical + brand: the audiences of named writers in a niche). The combined concentration produces deeper density than either dimension alone.

For founders choosing an atomic network, the question is which dimension produces the tightest functional density for the specific product. Pick the dimension that matters most; then identify the smallest unit within that dimension that can be dominated.

Final word

For any founder or PM working on a network-effect product, The Cold Start Problem is essential reading. The lifecycle framework, the cold start strategies, and the loop-compounding patterns are not available with equivalent depth anywhere else in the literature. Read it, apply it, and let the framework shape your strategic decisions across years.

Network effects produce some of the most extraordinary outcomes in business — winner-take-most categories, generational moats, multi-billion-dollar companies built on user base alone. The path to those outcomes is specific and learnable. This book is the path's most useful guide.