https://madhavajay.com/the-cold-start-problem-using-network-effects-to-scale-your-product/ Madhava's Blog X in The Cold Start Problem: Using Network Effects to Scale Your Product - A Review Madhava Apr 16, 2025 So you wanna build a Network? [?][?] Time to read: 12 minutes I have to say this is one of the best books I have ever read. Author @andrewchen (Andrew Chen) from a16z goes into depth from personal experience and research on what it takes to build a networked product and platform and more importantly what critical mistakes that people make and how to recover from them with actionable strategies and tactics. The book is long but worth it, but since I know you are busy (hopefully trying to build a network), let me give you the tasty bits. It's a Paradox Andrew writes an amazing insight at the end of the book that sums up the whole shebang: But here's the paradox: To build a massive successful network effect, I argue that you must start with a smaller, atomic network. And use the success in the first set of networks to tip over the next set of small networks. I'm not convinced this step can be avoided. Pay attention to that last bit. Chen has spent a lot of time thinking about this and concludes there is no one-shot way to skip the "a big network" is built from "the smallest possible network". Define: Atomic Network The smallest possible network of users that can stand on its own and provide value to its members, thereby becoming stable and self-sustaining. So to Chen's suggestion, we must focus on the MVN (Minimum Viable Network) the smallest number of users for which they will continue to generate actions in the system. Each action that one user does generates actions from the others and they inturn continue like a perpetual motion machine. They do this because the network gives them value for participating and participating generates value for the others. Define: Hard Side The group of users who create disproportionate value for the network, often by providing the core content, service, product listings, or liquidity that attracts the "easy side" (e.g., consumers, viewers, buyers). Key characteristics often include: * They are essential for the network's core value proposition to function. * They are typically smaller in number compared to the easy side. * They are usually harder, more complex, or more expensive to acquire and retain. * Solving their specific problems or pain points is often the key to overcoming the Cold Start Problem. * Examples from the book include Uber drivers, Airbnb hosts, YouTube creators, and eBay sellers. Successfully attracting and retaining the hard side is crucial for bootstrapping and scaling a network effect business. The Who's Who Okay let's take a look at a list of the players mentioned in the book, how they started, what their hard side customer was, the approximate size of their atomic networks; and the one or more strategies they employed to beat the Cold Start problem and achieve network effects. Company Beginning Hard Side Atomic Strategy Size Solving a Hard Problem (unreliable Limo/Black ~15-20 taxis), Paying Up Uber Car Service Drivers cars for (driver incentives), in SF <3 min ETA Hustle (local recruitment/events), Competing over the Hard Side (vs. Lyft) Hustle (founder Renting Air photography), Mattresses Hosts ~300 Solving Hard Problem Airbnb during (Property listings (trust, unique Conference Lister) stays), Building the Moat (defending vs. clones like Wimdu) Solving a Hard Problem (social friction in dating), Matchbox Initially Whole Atomic Network Tinder (Initial Women (to Party (college campuses, Name) attract men) sororities/ fraternities), Hustle (party outreach) Flintstoning Initial Initial (founders seeding concept Content ~1000 content), Community Reddit pitched to Y Creators / Subreddit Moderation (managing Combinator Submitters Eternal September), Atomic Network (niche subreddits) Atomic Network Glitch 3 people (built a tool their Slack (Failed Initial Teams (>2000 game company Online Game) / Power Users messages) needed), Killer Product (superior team communication) Video Conferencing at WebEx Meeting Hosts Killer Product Zoom (Founder's (who invite 2 people (reliability, ease previous participants) of use) company experience) Burbn Content Come for the Tool, Instagram (Check-in Creators 1 (tool Stay for the Network app with (Photo first) (hipstamatic filter photos) Sharers) clone) Come for the Tool, Stay for the Network (resume builder) Professional Initially 1 (tool Invite-Only LinkedIn Network Influential first) (exclusivity, Concept Professionals leveraging social proof), Atomic Network (focus on professional niche) Come for the Tool, Stay for the Network (buy on eBay) Paying Money Sellers / Up (referral Transfer via Merchants 1 (tool bonuses), PayPal Palm Pilots (esp. on first) Acquisition Effect (Confinity) eBay) (eBay integration), Economic Effect (network value driving transactions) Come for the Tool, Less distinct Stay for the Network File Syncing early on; (free storage) Tool potentially 1 (tool Killer Product (easy Dropbox (Personal users first) file sync/backup Need) initiating utility), shared Acquisition Effect folders (viral referral program) Come for the Tool, Stay for the Network (embed videos on Video your website), YouTube Sharing Site Video 1 (tool Solving Hard Problem Concept Creators first) (easy video sharing), Managing Overcrowding (competition among creators) Solving Hard Problem (niche focus on game Justin.tv 1 streamer streaming/ Twitch (General Streamers + 1 viewer community), Managing Lifecasting) Ceiling (streamer support, content moderation) Talkshow ~1000s Invite-Only (Initial Interesting (extended (creating scarcity/ Clubhouse Audio App Speakers / friends of buzz), Magic Moments Concept) Moderators founders) (experiencing live conversations) AuctionWeb Atomic Network (Personal Niche (Not (focus on specific eBay Side Sellers mentioned) categories like Project) Beanie Babies), Managing Saturation Email Simplicity, Broad Newsletter Sellers / Utility (vulnerable Craigslist for Local Posters N/A to Cherry Picking by Events niche competitors, Airbnb) Platform Bundling (IE with Microsoft Operating Existing ecosystem Windows), Building (IE/ System (DOS) Distribution (Different Moat (network effect Windows) dynamic) around Windows developers/apps) Cold to Warm We can split these strategies into two phases. The 0 - 1 Cold Start where there are no atomic networks or users or they are very small, and 1 - N (Warm) where companies fought to grow beyond their initial niches and take on competition to win network effects. 0 - 1 Cold At a glance this tells us that most of these networks leveraged a few key strategies in their earliest phases: * Solving a Hard Problem * Come for the Tool * Atomic Network * Killer Product Many of these could be argued to share similar meanings, for example, a useful tool for one can be the killer product itself, a killer product might solve a hard problem (but it doesn't have to) however behind all of them is value being generated for a primary user, and then as an extension atomic networks forming as that value draws in secondary users. While the book doesn't mention it, I would argue that many of these had magic moments built into them as well like the first time you get an Uber and watch the car on the screen as it gets closer to you. 1 - N Warm Interestingly, as these network effects begin to take hold the ways to grow, prevent anti-network effects and defend against competition seem to vary greatly by the specific product or company. I think the lesson here is best summed up by @reidhoffman (Reid Hoffman) in the book Blitzscaling: "The strategy that gets you to your first hundred users is not the strategy that gets you to your first thousand--or your first million." Stage 0 - 1 (Cold) If you haven't got any Atomic Networks yet, it might be worth double checking the anti pattern list from the book to make sure you haven't approached the problem wrong. Anti-Pattern / Common Suggested Approach / Treatment (from the Mistake book) Counterintuitively, building a massive network requires starting with the smallest possible, viable, dense Atomic Network. Launching Too Broadly / Dominate a niche first. This initial focus Targeting Large TAM from might seem like a "toy" market (e.g., Day 1 Airbnb's air mattresses, eBay's Beanie Babies, Tinder's frat parties, Instagram's filters), but this hyper-focus is essential to build the critical first stable network before expanding. (Ref: Chapters 4, 6) Ignoring Network Density Prioritize achieving high density within the (Focusing on total users Atomic Network. Ensure sufficient over connections / interaction/value exchange between initial interactions) users before scaling. (Ref: Chapters 4, 6, 11) Focusing Only on the Identify and prioritize solving the problem "Easy Side" (e.g., for and acquiring the "Hard Side" first, as attracting buyers before they often create the value that attracts sellers) the easy side. (Ref: Chapter 7) Use "Flintstoning" - manually create or fake the hard side's contributions initially. Relying too much on Hard Create the first posts, listings, or content Side users (like yourself to make the network look active and creators or sellers) to valuable (like the Reddit founders did). join, causing the This can attract the "easy side" and network to fail before eventually draw in the real "hard side" once it gains value. there's an audience. Don't be afraid to "fake it 'til you make it" to overcome the initial emptiness. (Ref: Chapter 15) Understand the core definition: network Confusing Virality with effects mean the product inherently becomes Network Effects more valuable with more users. Focus on building this value, not just on tactics for rapid viral spread. (Ref: Chapter 1) Intentionally design onboarding and early interactions to quickly deliver the core network value proposition (the "Magic Moment") to hook new users. For Clubhouse Neglecting the Early during its peak hype, the "Magic Moment" was User Experience / Lack often described as the experience of a new of "Magic Moments" user opening the app and almost immediately dropping into a live, engaging, and perhaps even serendipitous audio conversation featuring interesting people discussing relevant topics. (Ref: Chapter 10). Build atomic networks organically first; Attempting a "Big Bang" solve the Cold Start Problem through focused Launch for a New Network effort before attempting large-scale rollouts. Avoid the "Big Bang Failure" pattern like Google+. (Ref: Chapter 32) Proactively design mechanisms for Curation, Moderation, and Quality Control from the start to mitigate potential issues like spam, fraud, and noise. When new users Not Anticipating or encounter these problems, they are likely to Managing Anti-Network have a negative first impression, fail to Effects Early see the platform's value, and ultimately leave (churn). You cannot achieve viral growth (like a K-factor > 0.7) if new users don't stay long enough to use it. (Ref: Chapter 5, 27) Use Paying Up / Subsidies strategically and temporarily, primarily to bootstrap the hard side or overcome initial friction. Focus on Over-relying on building inherent, long-term network value. Unsustainable Subsidies/ You should generally avoid deploying Incentives significant spending on subsidies until you have validated that the core product provides enough fundamental value to retain users. (Ref: Chapter 14). Believing the Software Recognize that the value and defensibility Alone Creates the (The Moat) lie in the network of users and Network their interactions, not just the technology itself. (Ref: Chapter 29) Clearly identify the fundamental reason Not Defining the Core users connect or interact within the network Interaction or Value (the "atomic" unit of value) and build the Unit initial experience around facilitating that. (Implied in Atomic Network concept) Understand that the Cold Start phase is Giving Up Before inherently difficult and slow. Persist with Reaching the Tipping focused Hustle and iterative strategies to Point build the Atomic Network until it can grow organically. (Ref: Chapter 16) Things to Consider * To be able to infer what the root cause might be you should be looking at data where possible, of who is hitting your pages, what they do when they are there and how long they stay. * Consider what the lowest friction actions are that users can do to generate value for another user. For example in youtube that was likes and comments and make sure that this back and forth activity is optimized for. * You will want to try to classify behavior / actions that you believe produce value for your network and try to see why they aren't materializing. * Where ever possible, watch or survey your real users to understand what their friction points are and if they are encountering any anti-network effects. Every user you convert into an action producing member brings your k-factor up. * If you have a good idea on the atomic network size needed ensure you are trying to seed that network by bringing together those users even if you have to hustle or flintstone their counter parts. * Make sure the experience is magical, if there's no wow factor, either you failed to show why its interesting or its just not interesting. * Be careful to curate an experience that appears more full than it is by refining how the listings and interactions are displayed. You may need to down scope and only invite individuals that match that scope. Solving a Hard Problem Perhaps you have not really solved a hard problem yet, as @paulg (Paul Graham) says: A startup's initial problem is usually what the product should be and how to get the first users. It may sound like those are two problems, but they're not. Ideally you solve them simultaneously, and the process of trying to get users shows you what the product should be. Single Player As per the above, if you make something useful for one person (or entity) then you can continue to follow the gradient of value to discover what they need and how they can start to interact as a network. As co-founder of Youtube Chad Hurley said: "The idea is to initially attract users with a single-player tool and then, over time, get them to participate in a network. The tool helps get to initial critical mass. The network creates the long term value for users, and defensibility for the company... starting a network from scratch is very hard. Think of single-player tools as kindling." Gradient of User Value This dove-tails nicely into a tweet from the legendary @ID_AA_Carmack (John Carmack) Feedback beats planning. My plea at Meta was "No grand plans, follow the gradient of user value". To which @jimkxa (Jim Keller) piled-on: "Follow the gradient" That's great. We are learning so much from open source, shipping, screwing up and figuring it out. Plans are good, reality is better Cherry Picking Finally, if you are having trouble getting that initial thread of user gradient to follow you might need to look at what the competition is doing. Since users have limited eye-ball time, consider if your initial users might be poorly served elsewhere and try the Cherry Picking (with some Flintstoning) strategy to serve their needs better and try to switch them over. The book talks about how Craigslist was an initial online directory and collection of mini communities which was unable to defend against Cherry Picking resulting in the following companies which unbundled their customers: * Indeed: Targeted the Jobs section. * StubHub: Targeted the Tickets section. * Etsy: Targeted the Arts and Crafts for sale section. * Airbnb: Targeted the Room Rentals/Short-term Stays category. * Tinder: Targeted the Personals/Dating category. * Zillow/Trulia: Targeted the Housing/Real Estate listings. Stage 1 - N (Warm) I imagine after you have this initial user value gradient, one or more self sustaining (for now) atomic networks, and potentially even some reasonable <1.0 k-factor growth you could consider yourself no longer requiring a "Cold Start" and should focus your attention on the parts of the book relating to: * The Tipping Point (Part III, Chapters 11-16): This is the immediate next phase. Focus here is on scaling beyond the first atomic network(s) by finding repeatable growth strategies (like Invite-Only, Come for the Tool, Paying Up, Flintstoning, or Hustle) to capture adjacent networks and accelerate growth across the broader market. * Escape Velocity (Part IV, Chapters 17-21): Aiming for rapid, self-sustaining growth. This involves understanding and optimizing the Trio of Forces: + The Acquisition Effect (viral growth). + The Engagement Effect (making the network stickier). + The Economic Effect (improving monetization and business model). * (Anticipating) Hitting the Ceiling (Part V, Chapters 22-28): While not the immediate focus, understanding the factors that eventually slow growth (saturation, degrading channels, network revolts, etc.) helps in making strategic decisions during the Escape Velocity phase to mitigate future slowdowns. To build something massive, you must start small. Nail the atomic network, follow the user value gradient, and let each success tip over the next. Plans are fine--reality wins. Listen. Hustle. Stack momentum. Regardless of if you are at the cold or warm stage of building a Network, I would urge you to read this excellent the book yourself. Good luck out there! Don't stay FROSTY. The Cold Start Problem by Andrew Chen The Cold Start Problem Book Cover Paperback | eBook Strategies Glossary from Cold Start * Acquisition Effect: One of the three core forces driving Escape Velocity; occurs when the network itself helps acquire new users organically (e.g., through invites, viral loops, embeds). * Atomic Network: The smallest stable, self-sustaining network of users that can provide value to its members and serve as the foundation for growth. Focuses on achieving density in a small niche before scaling. * Building the Moat: Strengthening the network effect to create a sustainable competitive advantage that is difficult for rivals to overcome due to the incumbent's established user base and value. * Bundling: Integrating a new network product with an existing successful product to leverage its distribution and user base, helping to overcome the Cold Start Problem (e.g., Microsoft bundling IE with Windows). * Cherry Picking: A competitive strategy where a startup targets a specific, high-value niche within a larger, established network, aiming to provide a superior, specialized experience (e.g., specialized sites competing with Craigslist categories). * Come for the Tool, Stay for the Network: Attracting initial users with a valuable single-player utility (the "tool"), then converting them into an active network as critical mass is achieved (e.g., Instagram filters leading to the social network). * Community Moderation: Implementing rules, tools, and processes to manage user behavior, maintain quality, and prevent anti-network effects (like spam or abuse) as the network scales. Crucial for managing issues like "Eternal September." * Competing over the Hard Side: When established networks primarily compete by attracting and retaining the most valuable and often scarcest participants (e.g., Uber vs. Lyft competing for drivers). * Economic Effect: One of the three core forces driving Escape Velocity; occurs when the scaled network enables stronger monetization, better unit economics (LTV/CAC), and reinvestment into further growth. * Engagement Effect: One of the three core forces driving Escape Velocity; occurs when the growing network becomes stickier, increasing user retention and activity due to more content, connections, and value. * Eternal September: A phenomenon where a rapid influx of new, often less-informed users can degrade the quality and culture of an established online community if not managed well through onboarding and moderation. * Flintstoning: Manually simulating network activity or value in the very early stages ("faking it 'til you make it") to make the product seem active and valuable to attract initial organic users (e.g., Reddit founders posting initial content). * Hustle: Employing intensive, often non-scalable, hands-on efforts (manual outreach, attending events, partnerships) during the early stages to acquire the first crucial users and build initial network density. * Invite-Only: Launching a network with access restricted to invitations, creating scarcity, social proof, and leveraging existing trusted relationships for growth, while also helping to curate the initial user base. * Killer Product: A product whose core utility or performance is so compelling, especially for one side of the network, that it drives adoption even before strong network effects kick in, effectively overcoming the Cold Start Problem through sheer value (e.g., Zoom's reliability). * Magic Moments: Designing the early user experience to quickly and reliably deliver the core value proposition of the network, ensuring new users understand the "magic" and are motivated to engage further. * Managing Ceiling: Strategies employed when network growth slows down, involving diagnosing the cause (saturation, competition, decay) and shifting focus to retention, new growth vectors, or optimization. * Managing Overcrowding: Addressing the challenge where excessive competition among the "hard side" (e.g., creators on YouTube) diminishes value and potentially leads to churn, requiring better discovery or differentiation tools. * Managing Saturation: Strategies for dealing with growth limits when a network has captured most of its target market, often involving expansion into new geographies, demographics, or product categories. * Paying Up / Subsidies: Using financial incentives (discounts, bonuses, payments) to attract users, particularly the "hard side" of a marketplace, to bootstrap the network and overcome the initial lack of value. * Solving a Hard Problem: Focusing the initial product on solving a significant, unmet need or pain point for a specific group of users, especially the "hard side," thereby attracting them and kickstarting the network. Share: X in f