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

by Andrew Chen

Big Bang Launches can be great at landing, but often fail at expanding—and as we discussed, many networks with low density and low engagement will fail.

Economic network effect, which is how a business model—including profitability and unit economics—improves over time as a network grows. This is sometimes driven by what are called data network effects—the ability to better understand the value and costs of a customer as a network gets larger. This helps in driving higher efficiency when promotions, incentives, and subsidies into a network. The Economic Effect also can grow revenue by increasing conversion rates, by building features for the network as opposed to tools for the tool.

The Economic network effect, alongside its siblings in acquisition and engagement, provide a strong defense against potential upstarts. By acquiring and engaging large networks of users, a new competitor has to be significantly better than the status quo.

It might seem that premium pricing is a bad thing, but for many networks like marketplace companies, cryptocurrencies, and payment networks, the users of the network actually win as well. If eBay becomes the trusted, primary place to trade collectibles, then higher conversion rates and higher prices will benefit the sellers. They will make more money, and build their own businesses. When startups like Patreon and Substack create the ability for creators to earn a living by creating content on YouTube or via premium email newsletters, all parties benefit.

When something’s not growing on the Internet, it’s basically on the brink of declining, precipitously.

Rocketship Growth Rates for Marketplaces

In the marketplace case, you could look at publicly traded marketplace companies and see them trading at roughly five times net revenue. This means to achieve $1 billion in valuation, a company must hit $200 million in net revenue. And you want it to happen in year 10.

Next, you want to set goals for the intermediate years. In year 1 and 2, you might argue there’s probably zero revenue since the team will be focused on product development. Then year 3 becomes about solving the Cold Start Problem, and only in year 4 does it get to meaningful revenue. The company might hit $1 million/year in revenue that year. Then to extrapolate out, the marketplace product needs to grow from $1 million to $200 million over year 4 to 10—in other words, 266x over a year period.

starting from $1 million a year, you’d need to grow at an average rate of 2.4x over 6 years to hit $200 million. This is an average growth rate, and typically products grow fastest in the early years when revenues are small. A trajectory like 5x 4x 3x 2x 1.5x 1.5x would work, as would 4x 3.5x 3x 2x 2x 1.2x. Usually by the time a company is doing this analysis, they already have a year or two of data, and they can use this equation to figure out what the rest must look like. Or they can take empirical data from companies that are further along to extrapolate the first few years.

this is why the most valuable products in the world—most of the apps and platforms with a billion users—are typically networked products. When they work, they usually continue to work for a long time.

What looks like an exponential growth curve is often, in reality, a series of lines layered quickly on top of each other.

Evan Spiegel reported on the diminishing returns of connections: Your top friend in a given week contributes 25% of Snap send volume. By the time you get to 18 friends, each incremental friend contributes less than 1% of total Snap send volume each.

Our success was anchored on what I now call The Adjacent User Theory. The Adjacent Users are aware of a product and possibly tried using it, but are not able to successfully become an engaged user. This is typically because the current product positioning or experience has too many barriers to adoption for them.

While Instagram had product-market fit for 400+ million people, we discovered new groups of billions of users who didn’t quite understand Instagram and how it fit into their lives.

success in proliferating collaboration tools within one company will make it more likely that a close partner will also adopt the tools, because those networks are so interconnected. If all of an accounting firm’s clients use Dropbox, it’s likely that they will eventually try it, too.

The types of teams that have been used to incrementally build large, multi-year franchises may never have encountered the Cold Start Problem, so they bring the wrong background and tools to the table.

Another way to think of the risk of new initiatives inside of a company: if a new product’s success rate inside a company is similar to that of the venture capital industry as a whole, the success rate will be 50/50 at most. Exceptional outcomes might happen 1 in 20 times, if this pattern mirrors the startup world as well.

over a nearly decade time span, email marketing clickthrough rates dropped from 30 percent to under half that—13 percent.

In studies where people are shown web pages and their eyes are tracked to see exactly where they are looking, they show incredible skill in ignoring ads and just focusing on the content. This was recognized as early as 1998 by usability researchers (Benway/Lane, Rice University) and termed “banner blindness.”

The solution to the Law of Shitty Clickthroughs is to embrace its inevitability. When new products launch, there are usually one or two acquisition channels that work—but they might not scale.

The best practice is for products—whether they have network effects or not—to constantly layer on new channels.

if the Law of Shitty Clickthroughs says that marketing channels decline over time, the other strategy is to embrace new marketing ideas early.

The Law of Shitty Clickthroughs is best countered through improving network effects, not by spending more on marketing.

twenty apps drive 15 percent of all app downloads!

Slack’s S-1 filing showed less than 1 percent of Slack’s total customers accounted for 40 percent of the revenue, and Zoom’s indicated that 30 percent of revenue came from just 344 accounts, again less than 1 percent of their customer base.

The hard side of the network is the most difficult and expensive to scale. As the market saturates, it eventually becomes more important to “scale up” than to continue to acquire new members of the hard side.

In the earliest days, there was very little content to organize. Getting to the first 1,000 videos was the hardest part of YouTube’s life, and we were just focused on that. Organizing the videos was an afterthought—we just had a list of recent videos that had been uploaded, and you could just browse through those.

This is where Google’s expertise in dealing with massive amounts of data were crucial in developing two key features for YouTube in the ensuing years: Search and Related videos. Both helped users quickly navigate to videos they cared about, and because they were algorithmically driven, didn’t require the company to manually edit or curate the content.

On TikTok, the For You feed reflects preferences unique to each user. The system recommends content by ranking videos based on a combination of factors—starting from interests you express as a new user and adjusting for things you indicate you’re not interested in, too—to form your personalized For You feed. Recommendations are based on a number of factors, including things like: User interactions such as the videos you like or share, accounts you follow, comments you post, and content you create. Video information, which might include details like captions, sounds, and hashtags. Device and account settings like your language preference, country setting, and device type. These factors are included to make sure the system is optimized for performance, but they receive lower weight in the recommendation system relative to other data points we measure since users don’t actively express these as preferences.

“Data network effects” are often invoked as a path for networks to solve relevance and overcrowding issues that emerge over time. The signals are a combination of individual actions but are also based on algorithmic models built on the combined behaviors of hundreds of millions of users. More users means more behavioral data, which in turn allows for more fine-grained content recommendations—a

In the early days at Airbnb, we would always talk about creating a positive “Expectations Gap.” In the early days, when we were new, guests go in with low expectations, but then would be blown away by the experience. You need this high NPS to get people to tell their friends, and it makes hosts more likely to join too. Our competitors who took shortcuts couldn’t deliver here.

a theme that appears again and again in network-based competition: it’s asymmetric. The smaller player and bigger player use different strategies. And the most intense competition tends to happen as networks compete over the most valuable users of one network to another—this is “Competing over the Hard Side.”

The moat is about a successful network that defends its turf, using network effects in a perpetual battle against smaller networks trying to enter the market.

The key to investing is not assessing how much an industry is going to affect society, or how much it will grow, but rather determining the competitive advantage of any given company and, above all, the durability of that advantage. The products or services that have wide, sustainable moats around them are the ones that deliver rewards to investors.

replicating the complete functionality of a Slack or Airbnb might take time, but it is tractable. It’s the difficulty of cloning their network that makes these types of products highly defensible.

As the early Airbnb team described, the Cold Start Problem lies in the difficulty of launching a new city to a Tipping Point of over 300 listings with 100 reviews.

Platform dependence can be disastrous if not managed well. If you integrate too closely with a preexisting network, allowing them to control your distribution, engagement, and business model, you become just a feature of their network.

cherry picking is an enormously powerful move because it exposes the fundamental asymmetry between the David and Goliath dynamic of networks.

A new product can decide where to compete, focus on a single point, and build an atomic network—whereas a larger one finds it tough to defend every inch of its product experience. It’s one of the reasons why, particularly in consumer markets, it’s been so hard for “winner take all” to really happen in a literal way. The largest networks can take a lot, in many networks, but they remain vulnerable to any new upstart that uses cherry picking as a core strategy.

The Big Bang Launch is convenient for larger, more established companies as a method to launch new products because they often have distribution channels, huge engineering teams, and sales and marketing support. But counterintuitively, for networked products, this is often a trap. It’s exactly the wrong way to build a network, because a wide launch creates many, many weak networks that aren’t stable on their own. When companies don’t understand these nuances, it leads to disaster.

When networks are built bottom-up, they are more likely to be densely interconnected, and thus healthier and more engaged.

product’s first network is unlikely to be its last, when the team is working furiously to refine its network forces to conquer adjacent markets and networks. What might look like an airbed company eventually comes to disrupt the entire hotel industry. A chat product for small teams and startups eventually takes over the entire market as the de facto way for teams to communicate.

Focusing on the hard sides of the network, which are usually smaller in number, provides leverage in competitive moves. For a social network or video platform, it might make sense to pursue this side by giving content creators special economic incentives, or distribution for their content. For B2B products, it might be special features and pricing for enterprises. The core goal is the same regardless of the category—move the best and most important nodes from one network to another, and it will be a competitive win.

Bing is another example, when Microsoft wanted to get into search. It was the default search engine across the operating system, not just in Internet Explorer but also MSN and everywhere Microsoft could jam it. But it went nowhere. The distribution advantages don’t win when the product is inferior.

the product really matters, and bundling can provide a huge distribution advantage, but it can only go so far. It’s an echo of what we now see in the internet age, where Twitter might drive users to its now-defunct livestreaming platform Periscope, or Google might push everyone to use Google Meet. It can work, but only when the product is great.

Visual Basic was a key part of the flywheel for Windows. Every business, and especially small businesses, had all these programs that are part of their daily workflow. They weren’t super complicated programs, but necessary. VB made it simple. Companies could write them themselves without much prior programming experience. Or there were legions of resellers and small consultancy groups who wrote VB programs for clients. It was a whole ecosystem that really drove Windows forward. And it was only for Windows. There was never VB for OS/2 or Mac. You had to be part of the Windows ecosystem. It empowered people with little prior experience to be developers.

With Visual Basic, an infinite number of niche use cases, particularly within companies, could be automated. Thus the quote from early Microsoft execs, “For every copy of VB we sell, there are ten copies of Windows that go along with it.”