Two-Sided Marketplace Launch Playbook: How to Go to Market in 2027
Launch a two-sided marketplace in this order: (1) name the hard side, (2) shrink the market until it is one city, niche or moment, (3) seed that hard side by hand and with incentives, (4) give each side a reason to show up before the other arrives, (5) measure liquidity in that small market, (6) set a take rate once matches are happening, (7) build trust and safety before you scale, and (8) expand by copying the working unit. The chicken-and-egg problem is solved by not trying to serve both sides everywhere at once.
The core idea comes from Andrew Chen's cold start work: start with the smallest network that can stand on its own. Lenny's Newsletter on the atomic network defines the atomic network as "the smallest network needed that can stand on its own", and says the first one is often on the order of "hundreds of people." NFX's James Currier argues you should get the hardest side first because, once you have enough of them, "the other side is 2-10X easier to bring onboard."
Published numbers on how long each step takes do not exist, and this page does not invent any. Where a figure is not public, it says so.
Solve the chicken-and-egg problem by shrinking it
A marketplace is useless to a buyer with no sellers and to a seller with no buyers. The fix in every well-documented case is the same: stop launching to "the market" and launch to a very small group where both sides can actually meet. Andrew Chen's framing, summarized by Food On Demand, is that an atomic network is a small, dense group that can sustain itself. Chen's examples include Zoom needing two to three people, Slack five to ten and Tinder about 500 people on a campus.
The risk of going broad is real. One summary of Chen's lessons says launching too broadly risks anti-network effects, where quality suffers because each user finds too few relevant connections.
Pick the hard side first
Every marketplace has a side that is harder to recruit, usually supply. The same summary of Chen says that side needs disproportionate focus and incentives. NFX makes it tactic number one of its 19 chicken-or-egg tactics and gives an example: Outdoorsy got RV owners first, and demand came 5X faster and cheaper.
Pick by asking which side is scarcer and which side the other will follow. If you cannot say, interview both sides before you build anything. NFX also lists "pay cash to the most valuable side of the market" as a tactic, citing Uber paying drivers and ClassPass paying gyms upfront.
Constrain the market: one city, one niche, one moment
NFX lists geography, time and demand as separate constraints. For geography it cites Lyft, Yelp and Craigslist. For time, Tophatter allowed bidding only 8-9pm PT at launch. For demand, Groupon limited supply to one deal a day and Fiverr launched with everything at five dollars.
Uber is the clearest example of a narrow start. According to Lenny's Newsletter on the atomic network, early efforts focused on a narrow moment, "5pm at the Caltrain station at 5th and King St." The network then grew to the city's 7x7 square miles, then to the East Bay and Silicon Valley. NFX also points to a niche version of the same move: eBay started with Beanie Babies and Poshmark with urban female professionals.
Seed supply by hand before you automate
The early launches of the big marketplaces were manual and local. Each of these is reported by Lenny's Newsletter:
- DoorDash. The first version was a website with PDF menus for Palo Alto restaurants, and the team tested demand with flyers charging $6 for delivery around Stanford. The DoorDash S-1 confirms the start: founders launched a website displaying menus from local restaurants in Palo Alto, and the company was incorporated in 2013 as Palo Alto Delivery Inc.
- Etsy. It attended craft fairs across the United States to recruit sellers, reasoning that sellers would bring their own buyers.
- Uber. Street teams were used heavily early on, and Travis Kalanick personally handed out referral codes at Twitter HQ, per Andrew Chen.
- Airbnb. The 2008 Democratic National Convention in Denver was the turning point, when demand for rooms outpaced supply. The founders pitched the smallest bloggers first, bigger outlets such as NBC and CBS followed, and cereal boxes made for tech bloggers later sold for $40 per box.
NFX also describes doing the matching manually: Zappos staff bought and shipped shoes by hand. It warns against faking activity when you make supply look bigger, and notes that exclusive access "almost never works".
Give each side single-player utility
Single-player utility means a user gets value even when the other side is thin. NFX gives several routes: build software for one side (its tactic seven, with OpenTable, Honeybook and StyleSeat as examples, because "it helps lock them in so you have time to attract the other side"), or change the behavior of only one side, as Square did.
The Airbnb founding story fits too. The Airbnb S-1 says that in 2007 two founders created AirBedandBreakfast.com to rent airbeds to conference attendees, a single event with a built-in crowd, and the company was incorporated on June 27, 2008.
Measure liquidity in the small market
Liquidity is commonly framed as the probability that a listing sells or that a searcher finds what they want, a definition eChai Ventures credits to the NFX approach. I did not find a published numeric target in any source for this page, so set your own in your first market: for example the share of listings that get a response, and the share of searches that end in a match.
Measure it inside the atomic network, not across the whole company. A healthy small market is the signal that it is time to copy the unit, not a blended average over many weak ones.
Set the take rate after matches happen
Take rate is revenue as a share of the value that flows through the marketplace. Public filings show it moves with the business. The DoorDash S-1 reports 2019 Marketplace GOV of $8.0 billion and 2019 revenue of $885 million, which works out to roughly 11 percent by my own arithmetic (the filing excerpt I read does not state a take rate). The filing says revenue grew faster than Marketplace GOV primarily because of higher merchant commission rates and growth in Drive.
Etsy shows the same lever from the other direction. The Etsy 10-K states that "On April 11, 2022, we increased the Etsy marketplace seller transaction fee." Its Q4 2023 results release put consolidated take rate (consolidated revenue divided by consolidated GMS) at 21% for the quarter. The Airbnb S-1 says the company has "changed, and may in the future reduce, our service fees for strategic or competitive reasons." Take rate is a setting you tune, not a one-time decision.
Build trust and safety before you scale
Trust is part of the product. The Etsy 10-K says "The trustworthiness of our marketplaces and the connections among people in our community are cornerstones of our business."
The Airbnb S-1 lists what that looks like at scale: over 68% of guests left reviews of their stays in 2019, and hosts and guests had written more than 430 million cumulative reviews as of September 30, 2020. It also describes a guest refund policy under which Airbnb will rebook or refund a guest if a listing does not meet hosting standards, host protections including property damage protection and liability coverage, and secure messaging, risk scoring and secure payments. You will not have all of that on day one, but a review system and a clear dispute path belong in the first launch.
When to expand
Expand when the first atomic network works on its own, then copy it. Lenny's Newsletter on the atomic network says that once one network works you replicate it, and describes Uber's path from a train-station corner to the city to the wider region. The DoorDash S-1 shows where repeating that can lead: over 390,000 merchants and over 1 million Dashers as of September 30, 2020.
No source publishes a rule for when a single market is "ready." Use your own liquidity numbers from the previous section, and expand to an adjacent city or niche rather than a distant one.
Questions people ask
What is the first step in launching a two-sided marketplace?
Choose the hard side, usually supply, and launch to it in a very small market. NFX says that once you have enough of the hardest side, "the other side is 2-10X easier to bring onboard." See NFX.
What is an atomic network?
The smallest network that can stand on its own, per Lenny's Newsletter. The first one is often on the order of hundreds of people.
How did Airbnb, Uber, DoorDash and Etsy get their first users?
By hand and locally: DoorDash with flyers around Stanford, Etsy at craft fairs, Uber with street teams and referral codes, and Airbnb through bloggers around the 2008 Democratic National Convention. Details are in Lenny's Newsletter.
What take rate should a new marketplace charge?
No source here publishes a right answer. Public examples differ: Etsy reported a consolidated take rate of 21% in Q4 2023 in its results release, and DoorDash's 2019 revenue of $885 million on $8.0 billion of Marketplace GOV is roughly 11 percent by simple division. Test it after matches happen.
What is a good liquidity number?
There is no published universal benchmark in the sources used here. Define liquidity as the chance a listing sells or a search finds a match, and track it inside your first small market.
When should I expand to a second city?
When the first market works without constant manual effort. The big marketplaces then copied the working unit, as described in Lenny's Newsletter. No source gives a numeric trigger.
Sources
- NFX: 19 Tactics to Solve the Chicken-or-Egg Problem · James Currier, July 2019
- Lenny's Newsletter: atomic network · smallest network that can stand on its own; Uber example
- Lenny's Newsletter: how the biggest consumer apps got their first users · DoorDash, Uber, Etsy, Airbnb launches
- Food On Demand: Andrew Chen on the cold start problem · atomic network sizes
- Lessons from Andrew Chen (Antoine Buteau) · hard side and anti-network effects
- Airbnb Form S-1 (2020) · founding, reviews, trust features
- DoorDash Form S-1 (2020) · launch, Marketplace GOV, revenue
- Etsy Form 10-K (FY2023) · trust, seller fee change
- Etsy Q4 and full year 2023 results · consolidated take rate
- eChai Ventures: validating a marketplace idea · liquidity definition, narrowing the market
Statements are drawn from the cited pages as read in October 2026. The DoorDash take rate is my own division of two figures in its S-1. This is general information, not financial advice.
Top photo: USDA / Public domain via Wikimedia Commons