When you launch a new product or service, or put a creative work like a song or a book out in the world, you often face a “cold start problem.” It’s harder to get momentum when there are no existing users, just like it’s harder to start a vehicle’s engine when it’s cold.
Chen’s book has been eagerly anticipated in startup and venture capital circles, given his prominence as a general partner at Andreessen Horowitz and board membership at startups including Substack and Clubhouse—and the book’s promise to explain how to start, grow, and defend successful businesses.
Chen, who earlier led the rider growth teams at Uber, is focused on the role that networks of users play for the most successful tech companies. A “network effect describes what happens when products get more valuable as more people use them,” he writes. A simple example is that a telephone is useless unless someone else you know has one, and becomes more valuable the more people you know who have a phone.
Contemporary products that benefit from such network effects include Facebook, Uber, Airbnb, Dropbox, Slack, Instagram, Reddit, TikTok, YouTube, and Twitter. “The ability to attract new users, or to become stickier, or to monetize, become even stronger as its network grows larger,” Chen writes. (p. 23)
As he acknowledges, network effects are sometimes touted as a semi-magical business dynamic, but they’re poorly understood. And they are particularly vulnerable to the cold start problem, since even good products can fail if there’s no one else to message with, or rent a room or get a ride from, or share photos or videos with, for example.
One central takeaway of The Cold Start Problem is that many dominant businesses got their early start by serving niche markets—what Chen calls “atomic networks.” Rather than introducing it more broadly, for example, Bank of America launched the first credit card just in Fresno, California, in 1958, mailing 60,000 residents a BankAmericard. Tinder and Facebook first took off in specific college communities (USC and Harvard), and Slack was built initially for individual teams at startups or within larger companies. “Your product’s first atomic network is probably smaller and more specific than you think,” Chen advises. (p. 77) In its earliest days, he writes, Uber’s focus was on serving users in “narrow, ephemeral moments—more like ‘5pm at the Caltrain station at 5th and King St.’” in San Francisco. Google’s failed Google+ social network, which had a large-scale launch but failed to establish successful atomic networks, according to Chen, is a cautionary tale.
To get things started, Chen advises focusing on the “hard side” of the network, generally the small number of users whose participation creates disproportionate value (vs. the “easy side” of more freeloading or less-committed people.) Examples include the 4,000 volunteer Wikipedia editors who contribute a large portion of the edits, Uber’s “power drivers” who handle 60% of trips, and the attractive people, especially women, in dating apps.
Making the product compelling for the “hard side” is thus a critical early priority, which in the case of Uber involved extensive bonuses and financial incentives for drivers. Tinder’s founders threw big parties for college fraternity and sorority members to get them to use the app. Chen suggests looking at hobbies and side hustles to find ways that people can be engaged to participate in the “hard side.” He writes, “What people are doing on their nights and weekends represents all the underutilized time and energy in the world.” (p. 96)
The product also needs to be compelling, but Chen argues against initially focusing on elaborate features. “The product idea itself should be as simple as possible—easily understandable by anyone as soon as they encounter it,” Chen writes. “And at the same time, it should simultaneously bring together a rich, complex, infinite network of users that is impossible to copy by competitors.” (p. 106) An example is Zoom, with its simple interface and freemium model, which means you don’t have to pay for basic usage.
Once a product has succeeded in an atomic market, its creators can extend it into other adjacent markets—other cities, demographics, sectors, for example—until a tipping point when the momentum shifts and entire markets start coming on board.
Chen proposes three principal components of network effects:
- The engagement effect. Products don’t just get more valuable to you as there are more users, but more use cases emerge and your engagement goes up. While Twitter initially was largely about friends messaging each other, today many people spend hours following news and celebrities and other public figures on the platform.
- The acquisition effect. Many successful network products have mechanisms for users to help spread them to others, such as when a user sends money via PayPal and prompts a friend to sign up to collect it. This reduces reliance on paid marketing, which generally gets more expensive and less effective over time.
- The economic effect. The business model for some products improves as the network grows, allowing it to reduce costs or accelerate revenue. The more people use Slack in a company, for example, the more likely they are to upgrade to a premium plan.
The book—a hefty 387 pages—covers five stages of businesses that Chen identifies as part of his “Cold Start Theory,” including the cold start problem, tipping point, escape velocity, hitting the ceiling, and the moat.
When network products reach scale, they can hit a ceiling and growth can stall because of negative forces like fraud and oversaturation of the market. “In the end, only new products and innovation will kick off the next big growth curve, which is what encourages startups to grow from single products into multi-product companies,” Chen writes (p. 242). EBay moving beyond straight auctions to introduce “Buy It Now” is an example. He argues that networks are less defensible against competitors than many people think, and uses the example of Airbnb’s ill-fated European rival Wimdu to show how networks can collapse—in that case because it neglected to build a quality “hard side” of the network in the form of a community of hosts.
To be sure:
- Chen doesn’t sufficiently acknowledge the downsides of some of the companies and technologies he’s extolling. While he discusses Uber at length, he doesn’t mention the many controversies during former CEO Travis Kalanick’s tenure, including reports of a toxic corporate culture that tolerated sexual harassment and gender bias. His discussion of YouTube’s algorithms for promoting content doesn’t note that they’ve had a problem with driving people to extremist and misleading videos.
- As Chen notes, network effects only go so far. Uber exited China and Southeast Asia in the face of competition. And “being dominant in New York did not help the company succeed in San Diego, as its network effects were localized primarily to each individual city,” Chen writes. “This was always the critique of Uber’s business, and the root cause of the vicious trench warfare that needed to be fought city by city.” (p. 323)
- Chen writes that “crypto looks to be one of the most important new technologies emerging, and has networks at its core,” but he doesn’t discuss in any detail the implications of Web3 for the arguments he’s making. Perhaps that’s his sequel?
Memorable anecdotes and facts:
- Once teams have exchanged 2,000 messages on Slack, they’re unlikely to switch away from the product. Airbnb found that 300 listings, including 100 reviewed listings, was a ‘magic number’ to see growth take off in a market. Uber focused on getting wait times for cars under three minutes on average.
- A San Francisco nonprofit called Homobiles pioneered the model of having normal people as drivers that Lyft and Uber later copied.
- Uber’s app was initially built in Mexico so company engineers sometimes had to be given Spanish-English dictionaries to understand comments in the source code.
- Tinder first gained momentum in 2012 after its founders threw a birthday party for a hyperconnected friend at USC and required attendees to download their app to get in. It was “500 of the right people,” one co-founder later explained.
- “Flintstoning” is when manual human effort stands in for product functionality that hasn’t been built yet or for user activity that hasn’t happened yet. For example, Reddit’s co-founders populated its homepage with links until other people started using it.
- At the peak, Uber offered over $50 million per week in incentives to drivers in the US. It also did the same in China.
- “The term ‘network effect’ has almost become a cliche. It’s a punch line to difficult questions, like ‘What if your competition comes after you?’ Network effects. ‘Why will this keep growing as quickly as it has?’ Network effects. ‘Why fund this instead of company X?’ Network effects.” (p. 10)
- “The technology ecosystem is downright hostile to new products—competition is fierce, copycats abound, and marketing channels are ineffective.” (p. 24)
- “The next big thing will start out looking like it’s for a niche network.” (p. 76)
- “The Law of Shitty Clickthroughs says every marketing channel degrades over time. This means lower clickthrough, engagement, and conversion rates, regardless of if you’re talking about email, paid marketing, social media, or video. This is a core reason why products hit growth ceilings—when marketing channels stop performing, the growth curve starts to bend downward.” (p. 265)
- “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 to the next set of small networks. I’m not convinced this step can be avoided.” (p. 345)
The bottom line is that The Cold Start Problem provides a clear analysis of how one should approach building a network product, with ample examples from the tech industry. It’s relevant to people interested in what distinguishes some of the most successful companies of our time from competitors that collapsed—and is likely of special interest to readers in product, startup, and investing roles.