A Mental Model for “Cold Launching” Start-ups (Four Launch Modes)

Disclaimer: I cannot claim originality to the framework used here but owe it to Cao Zheng, a former core member of Baidu’s business analysis group, 4399.com chief architect, and a current gaming entrepreneur from China. He is an avid thinker/ blogger on the internet business, from whom I have learned much (link to Cao Zheng’s public account on WeChat “caoz的梦呓”)

After more than half a year in incubation, a friend recently launched the beta version of his marketplace app for people to rent or buy things from people nearby, yes, the old “neighborhood sharing economy” concept again (aka. using app to borrow a power drill from your neighbor). I have served as an adviser to him prior to the launch. During our conversations, one counterfactual I constantly used to push the discussion was all the neighborhood sharing apps that have struggled to gain traction before, which is well documented by this article on FastCompany – The “Sharing Economy” is Dead, and We Killed It.

“Everything made sense except that nobody gives a sh**. They go buy [a drill]. Or they just bang a screwdriver through the wall.”

Despite featuring a super sleek interface (he is a great product designer), I think he has yet to address the very essence of the business problem faced by his predecessors and grasp how hard and crucial it is to distinguish between “Real Needs” vs. “Fake Needs” a priori, which deserves another blog post by itself. For my friend’s sake, I hope I am wrong on this.However, when the cost of usage (time, money, effort to learn) drops to a sufficiently low level and social proof for adoption catches on, “fake needs” could turn into “real needs.” Examples include cell phones and the internet, 20 years ago vast majority of people did not consider them essential parts of life.

However, I would like to use this blog to lay out a mental model for thinking through the difficulty and cost of “cold launching” a start-up, using four different launch modes, which has particular relevance to some of my friend’s challenges at hand and applicable to many aspiring entrepreneurs.

The four modes are:

  1. Single-point launch
  2. Single-sided launch
  3. Double-sided launch
  4. Multi-sided launch

From 1 to 4, the cost and difficulty of launch increase sequentially prior to a start-up reaching scale, and after the start-up reached scale, the barriers to entry established is also sequentially higher, by and large. Aspiring entrepreneurs need to clearly think through the problems and strategies for each launch mode during business and financial planning.

  1. Single-Point Launch

Core attribute: single-player product where a customer’s usage and experience is independent of other users or suppliers

Good examples are utility apps such as Evernote for individual users and Salesforce for corporate customers, and single-player games such as Angry Birds. As long as users find it useful, there is no need for a critical mass to launch. Of course, having a critical mass helps with word of mouth and viral marketing.

With single point launch, entrepreneurs could take the time to find product market fit without worrying about losing customers due to lack of network effects. Entrepreneurs could gradually reconfigure and improve their products based on feedback from new users until the product matures before making a push into the broader market. The associated expenses and marketing costs also tend to be lower at the outset.

Before moving on, another good example of single-point launch, Google’s search engine, which enjoys enormous data network effects.

  1. One-Sided Launch

Core attribute: the launch and adequate user experience require a batch of similar users to be onboard, otherwise users will churn quickly.

Classic examples are social networks, messenger apps, dating apps, and multi-player online games, and recently Uber Pool (due to its Peer-to-Peer nature)

This type of product requires acquiring and on-boarding users by the batches and reaching a certain level of user density in a short time. Without enough density, acquiring users one by one often times means losing users one by one. No one wants to step into an empty stadium.

Thus one-sided launch faces the chicken and egg issue and is harder than single-point launch. However, start-ups that could crack this initial issue could drive network effects and benefit from the spinning fly wheel later on, where new users beget more users.

Examples of growth hacks used by companies to crack the chicken and egg issue include:

  • Facebook: during his Harvard days, Mark Zuckerberg web-crawled university house directories (not hacked) to pre-populate Facebook with everyone’s profiles so that the site does not appear empty when the first user showed up. They also followed a classic cluster by cluster strategy by saturating a colleague campus to ~80% before moving on to the next one, and before opening up the social network to the general public
  • Gaming Firms: a lesser known and dark side fact is that some gaming companies use bots to trick players into thinking there are many other players online to get the initial traction going. Other tricks include buying Amazon product reviews, App Store reviews.


  1. Two-Sided Launch (Two-Sided Marketplace)

Core attribute: successful launch requires both supply and demand sides to reach a certain level of density. Without balanced participation from both sides, users and suppliers will churn quickly.

Classic examples include eBay, Taobao, Seamless, Opentable, Uber X. E-commerce sites that actually source products from suppliers such as Amazon or Jet.com are actually examples of single-point or single-sided launch. Amazon started out only selling text books, then expanded into CDs, DVDs, and gradually into other categories.

In practice, the supply side is typically the end that needs to be addressed first to bootstrap a network, as a marketplace with zero suppliers has zero value to consumers. Different businesses require different tactics, one way is to create an initial hook to get suppliers on board. Companies like OpenTable built tools for restaurants first then got them to participate in a network. Chris Dixon penned the frame “come for the tool, stay for the network.” For Uber, it means getting a few cars and drive around San Fran to pick up customers yourself to build the initial mass. Other times, the road to building a critical mass on one-side is a long grind, taking Airbnb for example.

Two-sided launch typically takes more money, more time and effort. Word of mouth alone is not enough, just take a look at all the subsidies that have been given out by ride sharing companies. However, once established, the barrier to entry is also very high due to the upfront spend required. Here I would like to make a distinction between True Network Effects vs. Supply-Side Economies of Scale. While eBay, Taobao, Opentable clearly have the former, what Uber X and Uber Taxi (not Uber Pool, due to its Peer-to-Peer nature) have is more supply-side scale effect, thus more prone to competition. One sign to distinguish the two is recent market entries, after Uber, Lyft, Gett, and most recently Juno all entered the NYC ride sharing market. Any meaningful eBay competitors (in its core marketplace) in the past decade?

  1. Multi-Sided Launch

Core features: besides the usual supply and demand sides, also need participation from a third party or fourth party to launch. Without enough participation from all these sides, growth would not be sustainable and users will churn.

One good recent example is Uber Eats, which requires participation from consumers, delivery drivers, and now restaurants. In fact, we could see that it is an evolution of Uber’s two-sided launch. Another example is Alibaba’s Taobao, only after having clearly established its two-sided marketplace did it start integrating logistics service providers, which were left to sellers to solve initially. Today, with Ant Financial, Alibaba also integrated a slew of financial services into its marketplace. Cold launch all the sides in a multi-sided business simultaneously is really really hard, with added complexity around operations, customer services, quality control, even for Uber and Alibaba. However, once launched and having reached scale, the competitive moat is also stronger.

Of course, if one has practically unlimited cash and does not have to answer to any investors, this would be less of a problem.

Strategic Intuition: moving from single-point >> single-sided >> double-sided >> multi-sided to drive network effect, scale, and build defensive moat

Products or services that use the single-point launch mode is the easiest to launch cold and with relatively low cost and no existing influence, but also vulnerable to competition. To build up defensive moat, entrepreneurs could transition the product from single-point to single-sided or double-sided. As we saw before, some companies strategically used single-point launch (single-player tools) as a hook to build up critical resources on one side of the market to eventually launch a double-sided business, such as OpenTable. Excel started out as a single-player utility app but later becomes a universal single-sided tool yet to be displaced. Instagram provided beautiful photo filters as the hook for people to participate in its network. Once having controlled important resources on both sides, to raise the bar even higher, consider moving to multi-sided. Even if competitors could match your product features, they could not match your number of users or consumers/suppliers to mount any credible threat.

In conclusion, during business and financial planning, aspiring entrepreneurs would need to clearly distinguish between True Needs vs. Fake Needs, as well as clearly assessing the resources required for launching their start-ups. This mental model is not a catch-all but a useful one.

“Hard work in tactics cannot compensate for laziness in strategy.”


Other helpful readings related to the topic:

Version One Ventures: A Guide to Marketplaces


A16z Podcast: Getting Network Effects


Matt Turck: The Power of Data Network Effects


Matt Turck: Can the Bloomberg Terminal be “Toppled”?


Punchcard Research: Reconsidering OnDeck & Data Network Effects


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s