Marketplace business ideas that make money
Two-sided platforms that connect buyers and sellers and take a cut — the model, the money moment, and the take rate for each real company. GMV is labelled where it is GMV; revenue is what the platform kept.
40 real companies · every number links to the page that reports it · corrections published
- 1
Holds no inventory. 23M users list and buy; Mercari keeps 10% per sale, then adds payments and ads on top.
Seller lists an item free in under a minute → Buyer pays in-app; Mercari escrows the cash → On sale, Mercari keeps 10% plus shipping and payment fees
The catch GMV growth has stalled in a maturing Japan.
Read the full teardown → - 2
Taxes buyers, not sellers. Free listing grows supply, supply grows demand — liquidity fee-charging rivals can't match.
Seller lists an item — free, unlimited, no seller fee → Buyer browses a deep, liquid catalog and checks out → Buyer pays a Buyer Protection fee on top — Vinted's main revenue
The catch Most revenue is one fee buyers love to hate.
Read the full teardown → - 3
A flat fee per order plus free equity delivery undercut percentage brokers; options volume at scale yields 55% margins.
Trader opens a free account; equity delivery costs nothing → Options cost a flat Rs. 20 per executed order; intraday and futures cost 0.03% or Rs. 20, whichever is lower (about $0.24) → Millions of daily orders yield $1B revenue at 55%+ margin
The catch Most revenue is retail F&O; SEBI now curbs it.
Read the full teardown → - 4
AI trained on millions of orders prices any custom part in seconds; rivals still quote by email.
Buyer uploads a CAD file; AI quotes a price and lead time in seconds → Xometry routes the job to a matched shop, which makes and ships the part → Buyer pays the quoted price; Xometry pays the shop less, keeping a 34.7% margin
The catch Thin 34.7% cut; still a net loss.
Read the full teardown → - 5
Feeds, follows and Posh Parties turn one-time resale into a social habit; buyer retention topped 104%.
Seller lists closet items free → Buyers discover via feed and Posh Parties → Sale closes - Poshmark keeps 20%
The catch Never profitable; heavy marketing spend.
Read the full teardown → - 6
It out-executed same-era rival Fotocasa on listing quality and trust — now valued about 20x higher.
Buyers browse and private sellers list, free → 40,000+ agencies pay monthly to list and win leads → Agency subscriptions = about 75% of revenue
The catch Liquidity is local; each market restarts.
Read the full teardown → - 7
Launched the marketplace, then built Tuts+ — the web's biggest design-tutorial network — to feed it buyers.
Authors list digital assets; Envato brings the buyers via Tuts+ → Buyers pay per item, or $16.50/mo for unlimited Elements downloads → Envato keeps about half of every sale plus subscription margin
The catch Its subscription now cannibalizes its authors
Read the full teardown → - 8
Every shift is a direct hire, not temp dispatch — no staffing license, no interview, worker paid same day.
Business posts an open shift; a worker taps 'work' — no resume, no interview → Worker shows up, does the shift, and is paid the same day → Business pays the wage plus a 30% service fee to Timee, billed monthly
The catch Clients can hire workers direct, fee-free.
Read the full teardown → - 9
Instead of maximizing supply like Upwork, Toptal rejects 97% of applicants and sells scarcity at a premium.
Screen applicants: only about 3% pass → Client hires; pays one blended hourly rate → Toptal pays talent less, keeps the markup
The catch Big markup underpays talent; they can defect.
Read the full teardown → - 10
Owns the bidding rails thousands of auction houses depend on, and the bidder demand they can't aggregate alone.
Auction houses list lots on ATG's ten marketplaces → Global bidders bid online and win lots → ATG takes commission on each lot won, plus payments and shipping fees
The catch GMV tracks cyclical art and surplus spend.
Read the full teardown → - 11
Cannabis sellers are banned from Google, Meta and Yelp ads, so Weedmaps became the one place buyers and sellers meet.
A dispensary can't run ads on Google, Meta or Yelp → It lists on Weedmaps and buys placement plus storefront tools → It pays Weedmaps every month to stay visible to nearby buyers
The catch Revenue has fallen two years straight.
Read the full teardown → - 12
Only sellers pay: 217K of 7M+ suppliers fund the whole marketplace while 200M+ buyers browse free.
200M+ buyers browse free → Inquiries flow to sellers → Sellers prepay yearly subs
The catch Paying base flat; growth from price hikes.
Read the full teardown → - 13
Plugged into Discord groups already selling access by hand — became the paid checkout, so demand pre-existed.
Creator lists paid access: Discord, course, bot, software → Buyer checks out; Whop grants access and runs billing → Whop keeps about 5.5% blended on every payment
The catch Much GMV is faddish get-rich-quick creators
Read the full teardown → - 14
Free to start pulled in millions of tiny sellers Rakuten priced out; BASE earns a cut of every sale.
Seller opens a free BASE shop in minutes → Seller lists products; buyers check out → BASE takes about 6.6% of each sale
The catch Tiny sellers churn; GMV per shop stays low.
Read the full teardown → - 15
Brokers built it in 1998 and fed it every listing. Private equity later bought control and made the sellers pay.
Broker uploads a home to Hemnet → Seller buys reach: base fee plus premium tiers → Hemnet books the fee; ARPL rises yearly
The catch One country, one housing cycle.
Read the full teardown → - 16
Paid experts vet every lot, so buyers trust it enough to pay a 21.5% combined take in a market full of fakes.
Seller submits a special object to auction → An in-house expert vets, prices, and lists it → It sells - Catawiki takes 12.5% + a fixed fee from the seller, plus a 9% + €3 Buyer Protection fee from the buyer
The catch Experts scale linearly; margins stay thin.
Read the full teardown → - 17
Free built-in escrow lets strangers wire $50K for a used watch across 150 countries without fear of fakes.
Sellers list watches free; buyers browse 500,000+ of them → Chrono24 holds the money in escrow until the watch arrives — free → 6.5% commission on a private sale; dealers pay rent plus an undisclosed cut
The catch Dealers could bypass; GMV swings with prices.
Read the full teardown → - 18
Eleven years of startup raises built a 1.79M-investor list. Selling that list pre-IPO stakes now pays 79% of revenue.
1.79M people sign up free to browse startup raises → StartEngine buys shares of late-stage private companies → Sells stakes in those funds to accredited users: $86.6M
The catch 79% of revenue sits in one product line.
Read the full teardown → - 19
Free to list, so supply floods in; buyer and seller each pay about 3% on a sale for what used tech lacks: trust.
Seller lists a device free; Swappa staff hand-review it → Buyer pays the listed price (fee baked in) via PayPal → Swappa keeps about 3% from each side on every sale
The catch Scale Low capped
Read the full teardown → - 20
Invite-only curation plus authentication let strangers trust five-figure buys that open marketplaces can't safely host.
Vetted dealers list rare design, art, and watches → Designers and wealthy buyers order on-platform → 1stDibs takes a 5-50% commission on each sale
The catch Rare, cyclical buys cap repeat and growth.
Read the full teardown → - 21
Pros pay upfront for lead credits, so Bark earns before any job happens or any deal closes off-platform.
Customer posts a free request → Bark matches nearby pros → Pros buy credits to reply
The catch Earns whether the pro wins the job or not
Read the full teardown → - 22
The product is distribution, not deals: a 1.5M-buyer list makers hand 70-80% of a Select deal to reach.
Software maker lists a lifetime deal at a steep discount → AppSumo emails it to 1.5M deal-hungry subscribers → Sellers report AppSumo keeps 70-80% on Select deals (AppSumo disputes it and gives no number); maker gets buyers
The catch One-time cash, no recurring revenue.
Read the full teardown → - 23
Global ENS firms sold pricey retainers to megafunds in English. VisasQ made expert calls self-serve, in Japanese.
A fund or consultancy needs a niche expert - fast → VisasQ matches a vetted expert; books a 1-hour call → Expert paid hourly; VisasQ keeps about 30% (VQ Lite)
The catch Demand swings with deal & consulting cycles.
Read the full teardown → - 24
Sells what Fiverr won't — anonymous fortune-telling and worry-chats are flagship categories, not novelties.
Anyone lists a skill: design, code, or a fortune reading → Buyers pay in-app via chat, video, or an anonymous phone call → Coconala rakes 22-60% of every completed deal
The catch Core GMV up 4.6%; growth leans on staffing.
Read the full teardown → - 25
Paid matchmaking: brides and grooms subscribe to contact verified matches, split across 300+ caste and language sites.
Free profiles fill each community site with brides and grooms → Contacting a match requires a paid subscription → About 1M members pay for multi-month subscription plans
The catch Young urban India is leaving for dating apps
Read the full teardown → - 26
Every online gun sale must end at a licensed dealer. GunBroker wired 32,000 of them in and bills the sale.
Sellers list guns and gear, paying for listing options → Buyer wins; the item ships to a licensed FFL dealer → GunBroker bills a final value fee plus a buyer service fee
The catch Revenue still sits below its FY22 peak.
Read the full teardown → - 27
Useme becomes the legal seller of the work, so someone with no registered business can still issue a real invoice.
Freelancer with no registered business posts a deal → Useme contracts both sides and invoices the client → Client pays Useme; it keeps 4.99% and pays the freelancer
The catch Thin take: 4.99% of a pass-through flow.
Read the full teardown → - 28
Takes 100% of a tutor's first lesson, then a commission that falls from 33% to 18% the more they teach.
Learner picks a tutor and pays per lesson → Tutor teaches 1-on-1 over video → Preply takes 18-33% - and 100% of lesson one
The catch Tutors resent the 100% first-lesson grab.
Read the full teardown → - 29
No inventory, no imports: 244k personal shoppers abroad are the supply, and BUYMA underwrites every sale as authentic.
Shoppers abroad list luxury, zero inventory → Buyer orders; escrow + authenticity guard the deal → BUYMA takes about 5% buyer + 5-7% shopper fee
The catch Discretionary luxury; weak yen bites GMV.
Read the full teardown → - 30
A marketplace's real bottleneck is demand-side awareness — Airtasker bought it with equity instead of burning cash.
Poster lists a task with a budget, like flat-pack assembly or a house clean → Taskers bid; poster picks one; payment is held in escrow until done → Airtasker releases the funds and keeps a 12.5-20% service fee, plus tax
The catch New markets restart the liquidity problem.
Read the full teardown → - 31
It only tests games — so its 1.5M-player panel earns trust generic UX tools can't win for a launch.
Studio uploads an unreleased build and defines its target players → Matched gamers from the 1.5M panel play it, recording video + think-aloud → Studio pays per test or subscribes for recordings, transcripts & AI analysis in 48h
The catch warn Scale mid
Read the full teardown →
human cost each test - 32
A flat monthly subscription where agencies take 30-35% of first-year salary — cheap enough for cash-poor startups.
Startups post roles by mission — no salary, no resume → Candidates click 'Visit' for a casual meeting → Companies pay a flat monthly subscription to keep hiring
The catch Japan-only; casual-visit hiring is local.
Read the full teardown → - 33
Spent 3 years nailing one city's liquidity before scaling — density that VC copycats never earned.
Homeowner posts a job; nearby pros send bids → Homeowner picks a pro and pays in-app → GreenPal skims a small fee off every completed job
The catch Thin cut on a low-ticket, seasonal chore.
Read the full teardown → - 34
he lived the pain as a server kid, then became the default payment rail before rivals showed up.
Minecraft servers sell ranks & perks to players → Tebex hosts the webstore + auto-delivers in-game → Takes 5% of every sale (15% on FiveM)
The catch ⚠ Control Low
Read the full teardown →
rides games it can't own - 35
Free to list, so captains flooded in. FishingBooker takes 10-30% only when a trip books. No VC, profitable in 4 mo.
Captain lists boat, free → Angler books the trip online → FishingBooker keeps the commission the captain sets, 10-30%
The catch Captains resent fees, nudge 'book direct'.
Read the full teardown → - 36
Beat incumbents with SEO and free author tools, then vetted freelancers so hard Reedsy became the default.
Indie author needs an editor, designer or marketer → Reedsy matches vetted pros; the two agree a fixed quote → Reedsy keeps 10% from the author and 10% from the pro
The catch Almost all growth comes from Google SEO.
Read the full teardown → - 37
he hiked the fee and cut to zero staff — a shrinking company that's wildly profitable.
Creator uploads a product, gets a checkout link → Buyer purchases; Gumroad runs checkout, payment, and delivery → Gumroad keeps 10% + $0.50 on your own links, and 30% on sales it sends you via Discover
The catch GMV plateaued · flat fee invites churn
Read the full teardown → - 38
Aaron sold surety bonds by hand; cousin Chris underwrote them. They automated the exact friction they'd lived.
A contractor needs a surety bond, asks their agent → Agent issues it on Propeller: apply, approve, pay by card in minutes → Propeller earns commission on every bond issued
The catch Carriers could build this and go direct.
Read the full teardown → - 39
MTurk runs on bots and pennies. Prolific ID-vets and fairly pays humans, so data runs 4x cleaner — now AI-grade.
Researcher or AI lab funds a study or annotation task → Vetted participants complete it, paid a fair wage (about 70%) → Prolific keeps a 33-43% platform fee on top
The catch High take rate; big AI labs may in-source.
Read the full teardown → - 40
Licenses amateur-shot photos to Japanese buyers by credit or subscription; keeps 58-78% of each sale.
Amateur and pro photographers upload local imagery → Buyers license by credit pack or subscription → PIXTA keeps about 47-78%, paying creators 22-53% depending on asset type
The catch AI now generates what buyers once licensed
Read the full teardown →
This list is a slice of Kaeda's library. Every figure is filed, stated by the company, or our estimate with the arithmetic shown — and when a source turns out not to say what we said, we change the number and publish the correction.