How to Market a Two-Sided Marketplace: Solving the Chicken-and-Egg Problem

  1. Blog
  2. Demand & Channels

By Saroj Jha · October 5, 2026 · 11 min read

A marketplace has to win two groups at once, so do not try to win everywhere at once. Recruit the harder side first, then concentrate both sides in one place until most requests find a match.

In three numbers

  • 91% of traffic to Airbnb came through direct or unpaid channels in the nine months to September 30, 2020[1]
  • 50% of Airbnb’s new listings in 2019 were booked within 4 days of becoming available; 75% within 16 days[1]
  • 1.4 million active sellers and 19.8 million active buyers on Etsy at the end of 2014, by its IPO filing[2]

A marketplace sells access to the other side. Clinics come for equipment; refurbishers come for clinics. Economists Jean-Charles Rochet and Jean Tirole put it this way: “Many if not most markets with network externalities are two-sided,” and platforms “must ‘get both sides of the market on board’”[3]. That is the chicken-and-egg problem: neither side joins until the other is there.

In my view, most marketplace marketing fails because it treats the two sides as one funnel and spreads thin. Below is the order I would work through it with a founder: understand the loop, choose the first side, pick one place, measure matches, then let both sides market for you.

Running case: Example Co. Fictional

This Example Co. is a different fictional company from the diagnostics Example Co. in my other posts. It is a seed-stage online marketplace that connects outpatient clinics (the buyers) with certified refurbishers of used medical equipment (the sellers), for items such as exam tables, patient monitors and ultrasound units. Sellers list each item with its service history and warranty. Example Co. takes an assumed 8% commission from the seller on each sale: on an illustrative $4,000 order, it keeps $320 and the refurbisher receives $3,680. It launches in one state first. All its numbers are illustrative.

Step 1: Understand the loop you are trying to start

In a two-sided market, each side’s value depends on the size of the other side. Economists call this a cross-side network effect. More refurbishers mean more stock and more choice for clinics. More clinics mean more orders for refurbishers. When it works, it feeds itself. When it does not, each side leaves because the other is missing.

That is why marketplace marketing is different: a single-sided product grows one customer at a time, but a marketplace has to grow two groups in step, in the same place.

Example Co., step 1 Fictional

A clinic fitting out a new exam room needs to find a used exam table, listed with its service history and warranty, this week. A refurbisher who lists a dozen items needs an order within weeks, or stops updating its listings. Both experiences depend on the other side being there. All fictional.

Figure 1 Each side is the reason the other side shows up
  • Sellers — Certified refurbishers — List exam tables, patient monitors and ultrasound units, each with its service history and warranty — Needs An order within weeks, or it stops updating its listings
  • More refurbishers: more stock and more choice for clinics Example Co. The marketplace in the middle More clinics: more orders for refurbishers
  • Buyers — Outpatient clinics — Buy used equipment, for example to fit out a new exam room — Needs A used exam table, with its service history and warranty, this week

One illustrative $4,000 order Fictional

$3,680

$3,680 to the refurbisher $320 to Example Co., an assumed 8% paid by the seller

When the loop fails, a clinic that finds nothing to buy may not come back, and a refurbisher without orders stops updating its listings.

Show the numbers as a table
Illustrative $4,000 order Amount Share
To the refurbisher $3,680 92%
To Example Co. (assumed commission, paid by the seller) $320 8%

The loop is my framing of the cross-side network effect in two-sided markets (Rochet and Tirole, 2003). Example Co., its assumed 8% commission and the $4,000 order are fictional; the order bar is drawn to scale.

What the filings show

Three well-known marketplaces describe this loop in their IPO prospectuses. These are what the filings say, not an account of how the companies made their choices internally.

  • Airbnb (S-1, 2020). “This demand encourages new hosts to join, which in turn attracts even more guests. It is a virtuous cycle — guests attract hosts, and hosts attract guests”[1].
  • Etsy (S-1, 2015). “Our seller-aligned business model creates network effects. The more we invest in our platform, the more we enable Etsy sellers to pursue their craft and grow their businesses and the easier we make it for Etsy buyers to find unique goods.” By 31 December 2014 it had “connected 54.0 million members, including 1.4 million active sellers and 19.8 million active buyers”[2].
  • Uber (S-1, 2019). “Our strategy is to create the largest network in each market so that we can have the greatest liquidity network effect, which we believe leads to a margin advantage”[4].
Figure 2 Three IPO filings describe the same loop: each side brings in the other
  • Airbnb Form S-1, filed 16 November 2020 — In the same filing — 91% of traffic to Airbnb came through direct or unpaid channels in the nine months to September 30, 2020
  • Etsy Form S-1, filed 4 March 2015 — In the same filing — Members connected at December 31, 2014: 54.0 million, including — Active buyers 19.8 million — Active sellers 1.4 million
  • Uber Form S-1, filed 11 April 2019 — In the same filing — The cost of paying both sides — Incentives for drivers and consumers “can result in a negative margin until we reach sufficient scale to reduce incentives.”
Show the numbers as a table
Filing Number Value
Airbnb S-1 (2020) Traffic through direct or unpaid channels, nine months to September 30, 2020 About 91%
Etsy S-1 (2015) Members, December 31, 2014 54.0 million
Etsy S-1 (2015) Active buyers, December 31, 2014 19.8 million
Etsy S-1 (2015) Active sellers, December 31, 2014 1.4 million

Airbnb, Etsy and Uber registration statements on Form S-1, quoted exactly. What the filings say, not how the companies made their choices internally. Bars drawn to scale.

Do this

Write one sentence for each side: “They come because the other side gives them ___.” If you cannot fill both blanks, you do not have a marketplace yet.

Step 2: Choose which side to recruit first

You cannot recruit both sides equally on day one, so choose. Platform economics gives a useful frame. Geoffrey Parker and Marshall Van Alstyne asked “How can firms profitably give away free products?” and showed, for markets of content providers and end consumers, “that either can be a candidate for a free good”[5]. In other words, which side pays, and which side you court, is a design choice, not a law of nature.

In practice, I would recruit first the side that is scarcer and slower to join, and whose absence ruins the other side’s first try. That is often supply: the hosts, sellers or refurbishers. A clinic that searches and finds nothing to buy may not come back; a refurbisher who lists early can wait a few weeks if you are honest about the timeline.

Used medical equipment adds one more supply-side question: what the seller actually did to the item. The FDA says it decides what counts as remanufacturing by the tasks performed, “not according to an entity's self-identified designation as a 'remanufacturer' or a 'servicer'”[6]. In my view the lesson for a marketplace is to ask each seller what work it did, and to show the service history on the listing, rather than rely on a label. This is not regulatory advice; a regulatory adviser should set your seller requirements.

If AI is part of your marketplace

In my view, AI changes this step in three ways. It can lower the work of joining for the scarce side: a tool that drafts a listing from a seller's own records makes listing faster. It adds a trust question: an AI-written listing must say no more than the seller actually did, and a person on the seller's side should check it. And it can move the sides: AI can create demand for some kinds of supply while replacing others, so recheck which side is scarce more often than once a year.

Upwork, a public work marketplace, describes both movements in its annual report for 2025. It calls the Upwork Marketplace "the world's human and AI-powered work marketplace that connects businesses with on-demand access to highly skilled independent talent worldwide", with a "proprietary AI assistant, Uma, which enhances productivity for both talent and clients". It reports that "GSV from AI-related work performed on the Upwork Marketplace has increased as demand for AI talent has grown in recent periods" (GSV, as the filing defines it, is the total dollar value transacted through all Upwork platforms and other workforce solutions). Its risk factors also describe competing companies that use AI and machine learning to "provide automated alternatives to the talent available through our platforms and other workforce solutions", and warn that "AI outputs may be, or alleged to be, deficient, inaccurate, inappropriate, biased, or infringing on the intellectual property rights of third parties"[7]. This is what the filing says, not how the company decided internally.

Example Co., step 2 Fictional

Certified refurbishers with stock are the scarce side, so Example Co. recruits them first. They list for free; Example Co. earns only when an item sells. Its goal is 20 signed refurbishers, with about 300 listings, before inviting any clinics. Every listing must show its service history and warranty, and the team asks each refurbisher what work it does rather than accepting a label. If it later adds a tool that drafts listing text from a refurbisher's service records, the refurbisher still checks and approves each listing before it goes live. All fictional.

Figure 3 Three questions point to the same answer: recruit refurbishers first
Question Example Co.’s answer So

  1. 1 Which side is scarcer, or slower to join?

    Refurbishers. Certified refurbishers with stock are the scarce side, and every listing needs its service history and warranty.

    Recruit refurbishers first

  2. 2 Which side leaves after one failed search?

    Clinics. A clinic that searches and finds nothing to buy may not come back; a refurbisher who lists early can wait a few weeks.

    Have the stock in place before clinics arrive

  3. 3 Which side pays, and which joins free?

    Refurbishers list free and pay an assumed 8% only when an item sells; clinics pay for the equipment.

    Let the side you court first join free

Example Co.’s choice Fictional

Sign 20 refurbishers, with about 300 listings, before inviting any clinics

20 refurbishers, one dot each

My three questions, from Step 2; the point that either side can be the one that joins free follows Parker and Van Alstyne (2005). Example Co.’s answers and its target are fictional.

Do this

For each side, write down how many you need before the other side can have a good first experience. Recruit the side with the harder number first.

Step 3: Start in one place and get dense

Liquidity is local, even online. Uber’s prospectus describes its strategy as creating “the largest network in each market”, market by market, not worldwide at once. It also shows the cost of building both sides with money: “We can choose to use incentives, such as promotions for Drivers and consumers, to attract platform users on both sides of our network, which can result in a negative margin until we reach sufficient scale to reduce incentives”[4]. A seed-stage startup cannot pay for that everywhere.

Airbnb’s own origin story, as its prospectus tells it, starts with one city and one event: “The year was 2007. Brian and Joe — two of our founders and friends from design school — were looking for a way to cover the cost of their San Francisco apartment.” An “international design conference was coming to town, and every hotel was sold out”[1]. One place, one moment of unmet demand.

For Example Co., “local” still matters. In my view, equipment is bulky and needs delivery, and clinics in one state hear about new sellers from the same associations and from each other. Density means a clinic sees enough listings in the categories it needs, and a refurbisher sees enough requests to move its stock.

Example Co., step 3 Fictional

It launches in one state. Spread across ten states, its 20 refurbishers would leave two in each, with about 30 listings, too few for a clinic to find what it needs. Concentrated in one state, the same 20 can list about 300 items across most of the categories clinics ask for, and clinics hear about Example Co. from several peers rather than once. The numbers are illustrative.

Figure 4 In one state, the same 20 refurbishers offer a clinic about 300 listings, not 30

One refurbisher Illustrative numbers, fictional

Spread across ten states 2 refurbishers and about 30 listings in each

State 1 State 2 State 3 State 4 State 5 State 6 State 7 State 8 State 9 State 10
  • Too few listings for a clinic to find what it needs
  • Clinics hear about Example Co. once

Focused on one state All 20 refurbishers and about 300 listings

State 1
  • Listings across most of the categories clinics ask for
  • Clinics hear about it from several peers

Listings a clinic can browse in its own state, drawn to scale

Spread About 30 Focused About 300 Show the numbers as a table
Launch option Refurbishers in each state Listings in each state
Spread across ten states About 30
Focused on one state About 300

Example Co.’s two launch options from Step 3 of the running case; all numbers are illustrative and fictional. Listing bars drawn to scale.

Do this

Pick one state (or one tight niche) where you can reach both sides cheaply, and say no to demand from anywhere else until it works there.

Free tool

Price incentives against what an order is worth

Before you pay either side to join, the free Unit Economics Calculator works out lifetime value, acquisition cost and payback, so you can check what an order is worth after your commission.

Open the calculator

Step 4: Measure liquidity, not sign-ups

Sign-ups on either side tell you little. What matters is liquidity: the chance that someone who shows up finds a match quickly. Airbnb’s S-1 reports a supply-side version of this: “for active listings in 2019 that were new to our platform, 50% received a booking within 4 days of becoming available, and 75% received a booking within 16 days of becoming available”[1].

Figure 5 Half of Airbnb’s new listings in 2019 had a booking within 4 days, and three in four within 16

Active listings in 2019 that were new to Airbnb: share that had received a booking

Within 4 days of becoming available 50% Within 16 days of becoming available 75%

Booked Not yet booked Each square is one in 20 new listings

The supply side: how fast a new listing finds a guest. This is the version Airbnb reports.

The demand side: how often a request turns into an order within a set time. Example Co.’s match rate, in the next figure, measures this.

Show the numbers as a table
Time after the listing became available Share booked
Within 4 days 50%
Within 16 days 75%

Airbnb, Inc. Form S-1, filed 16 November 2020: active listings in 2019 that were new to the platform. The filing reports these two points only.

I would track three numbers each week, one for each side and one for the match itself:

  • Match rate: the share of clinic requests that turn into an order within a set time.
  • Seller sales: items sold per refurbisher each month.
  • Repeat rate: the share of clinics who order again.

The weakest number tells you which side to recruit next. A low match rate means you need more of the right stock; low seller sales mean you need more clinics.

Example Co., step 4 Fictional

In week 10, 55% of clinic requests turn into an order within 14 days against a target of 80%. Refurbishers sell an average of 4 items a month against a target of 5, and 52% of clinics order again within 30 days against a target of 60%. Requests cluster on exam tables and patient monitors, but most listings are ultrasound units, so Example Co. recruits refurbishers with exam tables and monitors before spending on ads to clinics. All numbers are fictional.

Figure 6 Match rate is furthest from its target, so Example Co. recruits the missing stock first
  • Week 10 Now, as a share of the target
  • Match rate The match Clinic requests that turn into an order within 14 days — 69% of target Weakest — Seller sales Refurbishers Items sold per refurbisher a month — 80% of target — Repeat rate Clinics Clinics that order again within 30 days — 87% of target
  • Why the match rate is low — Clinic requests cluster on Exam tables and patient monitors — Most listings are Ultrasound units — So Example Co. Recruits refurbishers with exam tables and monitors before spending on ads to clinics — The rule: a low match rate means you need more of the right stock; low seller sales mean you need more clinics.
Show the numbers as a table
Number What it counts Week 10 Target Of target
Match rate Clinic requests that turn into an order within 14 days 55% 80% 69%
Seller sales Items sold per refurbisher a month 80%
Repeat rate Clinics that order again within 30 days 52% 60% 87%

Example Co.’s check in week 10, from Step 4 of the running case; all numbers are fictional. Each bar shows the number as a share of its target, so the three can be compared.

Do this

Define your match rate in one sentence and put it at the top of your weekly report, above sign-ups.

Step 5: Turn both sides into marketers, then expand

Once a state is liquid, the people on each side are your cheapest channel. Etsy’s prospectus says: “Etsy sellers and Etsy buyers have been our best marketers, sharing their positive experiences with their own communities.” It adds: “We believe that much of the growth in our member base to date has originated from word-of-mouth referrals and other organic means, as our historical marketing efforts and expenditures have been relatively limited”[2]. Airbnb reported “approximately 91% of all traffic to Airbnb coming through direct or unpaid channels during the nine months ended September 30, 2020”[1].

In practice, build a loop on each side: refurbishers share their listing links with clinics they already sell to, and clinics that have bought invite other clinics. Then expand only when the first state clears your liquidity bar for several weeks in a row, and reuse what worked.

Example Co., step 5 Fictional

Each refurbisher gets a shareable listing link, and each clinic gets a link to invite another clinic. Example Co. opens a second state only after the first holds an 80% match rate and a 60% repeat rate for eight weeks running. All fictional.

Figure 7 Example Co. opens a second state only after the first holds the bar for eight weeks running

State 1

  1. Recruit refurbishers 20 signed, about 300 listings
  2. Invite clinics Only once the stock is there

Hold the bar

Week Week Week Week Week Week Week Week

Every week: an 80% match rate and a 60% repeat rate. Eight weeks running, not eight weeks in total.

Then

Open state 2 Reuse what worked in state 1

Running underneath, on both sides

Refurbishers Each gets a shareable listing link to send to clinics it already sells to

Clinics Each gets a link to invite another clinic

Show the numbers as a table
Condition for opening state 2 Level
Match rate 80%
Repeat rate 60%
Held for 8 weeks running

Example Co.’s plan from Step 5 of the running case; fictional. The stages are in order, not drawn to time.

Do this

Write down the liquidity bar your first state must hold, and for how long, before you are allowed to open the second.

What changes by stage?

  • Pre-seed — Recruit the first side by hand, in one place or niche. Match requests yourself to learn what a good match looks like.
  • Seed — Concentrate. Track match rate, seller sales and repeat rate weekly, and fix the weakest. Add referral loops on both sides.
  • Series A–B — Expand state by state against a written liquidity bar. Watch incentives on both sides; as Uber’s filing shows, they can mean negative margins until scale.

What mistakes should I avoid?

  • Launching everywhere at once. Thin stock in many places means no liquidity anywhere.
  • Counting sign-ups. Count matches.
  • Recruiting the easy side first. Clinics that find no stock do not wait.
  • Buying demand before supply can serve it. Paid traffic into an empty marketplace burns money and reputation.
  • Expanding on a calendar, not a bar. Open the next state when the first one works, not when the plan says so.
  • Accepting a seller’s label. The FDA decides by the work done, so ask what was done.

Related on AAJ

Articles

Playbooks

Templates and tools

References

The superscript numbers in the article match this list.

  1. Airbnb, Inc. (2020). Registration statement on Form S-1, filed 16 November 2020. US Securities and Exchange Commission. sec.gov
  2. Etsy, Inc. (2015). Registration statement on Form S-1, filed 4 March 2015. US Securities and Exchange Commission. sec.gov
  3. Rochet, J.-C. and Tirole, J. (2003). “Platform Competition in Two-Sided Markets.” Journal of the European Economic Association, 1(4), 990–1029. doi:10.1162/154247603322493212
  4. Uber Technologies, Inc. (2019). Registration statement on Form S-1, filed 11 April 2019. US Securities and Exchange Commission. sec.gov
  5. Parker, G. G. and Van Alstyne, M. W. (2005). “Two-Sided Network Effects: A Theory of Information Product Design.” Management Science, 51(10), 1494–1504. doi:10.1287/mnsc.1050.0400
  6. US Food and Drug Administration (9 May 2024). “FDA Issues Final Guidance to Clarify ‘Remanufacturing’ of Devices that Need Maintenance or Repair.” Press announcement. fda.gov
  7. Upwork Inc. (2026). Annual report on Form 10-K for the fiscal year ended December 31, 2025, filed 13 February 2026. US Securities and Exchange Commission. sec.gov

More in Go-to-Market & Growth Planning

Part of the Go-to-Market & Growth Planning hub - see all 27 resources on this topic.

Also useful in Industry Playbooks.