How to validate a marketplace idea before building
How to validate a marketplace idea before building
To validate a marketplace idea before building, run 30 supply-side interviews and 30 demand-side interviews, then operate a manual concierge marketplace in one zip code for two to four weeks. If you can hand-match 10 paid transactions without writing code, the idea is real. If you cannot, no amount of engineering fixes it.
That is the playbook. The rest of this post explains why each step exists, what to use, and where most marketplace founders waste six months and $40,000 before learning the answer.
Why marketplace validation is harder than SaaS validation
A SaaS startup has one ICP. A marketplace has at least two: the supply side and the demand side. You have to convince both, on the same week, in the same geography, that your platform beats the status quo. That is not 2x harder. It is closer to 5x.
If your vertical is regulated (mobility, food, healthcare, finance, childcare, alcohol, firearms), there is a third party: the agency that decides whether you exist. Skipping the regulator call is the single most common way validated marketplaces die in month four.
Liquidity is binary. A SaaS tool with one customer is a real product. A marketplace with sellers and zero buyers is dead air. Half-empty is worse than empty, because the empty platform still has hope and the half-empty one has churned suppliers telling their friends it does not work.
The supply-first default (and when to break it)
Modern marketplace wisdom has converged on a default: build supply first. Outdoorsy did it with RV owners. Airbnb did it by manually photographing early NYC apartments. DoorDash co-founders hand-delivered the first 100 orders themselves, using their own cars. Supply usually carries a higher per-person economic incentive (a host earns $200 a night; a guest saves $50), so suppliers tolerate a janky early experience that buyers will not.
The exception: you already own the demand audience. Catalant could lead with demand because the founders had Fortune 500 procurement contacts. Substack led with writers because Hamish McKenzie had a press list. If you cannot name 100 demand-side users you can text today, default to supply.
A useful tie-breaker question: which side has the bigger incentive to put up with a manual, ugly experience for the first three months? Whoever wins, recruit them first.
Step 1: 30 supply interviews + 30 demand interviews
The reason 30 per side is the threshold: at 20, you hear individual stories. At 30, the patterns start to repeat. By interview 25 on each side you should be able to predict the next answer before it comes.
Sixty conversations is roughly 20 hours of founder time spread over two weeks. Cost: $0 if you do them yourself (and you should; a hired interviewer cannot pivot the next interview based on what they just heard).
What to ask supply
Five questions, every interview, in order:
- How are you delivering this service or selling this thing today?
- What channel brings the most reliable revenue, and what does it take from you (commission, time, hassle)?
- If a new platform took 15% but brought you a steady stream of pre-vetted buyers, would you list?
- Walk me through the last bad customer experience you had. What did the existing channel fail to do?
- What would have to be true for you to make this your primary channel?
You are listening for: take-rate tolerance, pain with the current channel, and whether they would actually switch (not whether they say they like the idea).
What to ask demand
- Last time you needed this, where did you go and what did it cost you (in dollars, hours, frustration)?
- What does the current option fail at?
- If a platform pre-vetted suppliers and matched you in under an hour, would you pay a 10% premium for it?
- How often do you need this, weekly, monthly, yearly?
- Who else in your life needs this, and how would they find out?
You are listening for: transaction frequency, switching cost, and word-of-mouth potential.
How to recruit interviewees for free
Supply is easier to find than founders expect. Search the niche on Reddit, Facebook groups, LinkedIn, niche subforums. Cold-DM 100 people with a one-sentence ask: "I am researching the [X] industry and would value 20 minutes; can I send you a Calendly link?" You will get a 15 to 25% response rate if the message is honest and specific.
Demand is harder, but a single Reddit post in the right subreddit, a tweet from your personal account, or a flyer at a relevant in-person venue (gym, coworking space, farmers market) usually fills 10 slots inside 48 hours.
Step 2: The single-zip concierge run
After 60 interviews, you have a hypothesis. Now you test it the only way that counts: by running the marketplace by hand.
Pick one zip code, or a 5-mile radius. Not a city. Not a metro area. One zip. Geographic narrowness is the single biggest predictor of whether a concierge run produces transactions, because density solves the matching problem you have not yet built software for.
You are the matchmaker. Your stack is roughly:
| Component | Tool | Cost |
|---|---|---|
| Listings + match queue | Notion or Google Sheets | $0 |
| Payments | Stripe Checkout payment links | 2.9% + 30¢ |
| Communication | WhatsApp or SMS via Twilio | <$10/mo |
| Booking | Cal.com or Calendly | $0-15/mo |
| Demand landing page | Carrd or Framer | $0-19/mo |
| Total stack | ~$50/month |
Goal: 10 paid transactions in two weeks. Not 10 signups. Ten transactions where money moved and both sides reported a positive experience. If you cannot hand-match 10, the bottleneck is not software; it is supply quality, demand desire, or matching accuracy. Software will not fix any of those.
If you are a non-technical founder running the concierge phase, Cursor used as a non-technical founder can stand up the landing page and a basic intake form inside a Saturday, no engineer required.
Step 3: Fake-door the take rate
Most marketplaces die on take rate, not on idea. Suppliers say yes when you describe the platform abstractly. They say no when they see a 25% deduction on their first invoice.
Run a fake-door pricing test during the concierge phase. Tell three cohorts of suppliers three different rates: 10%, 15%, 20%. Watch the acceptance rate. Survey responses lie; behavior does not.
Industry benchmarks for healthy take rates:
| Vertical | Healthy take rate |
|---|---|
| Services (consulting, home, beauty) | 15-30% |
| Physical goods | 5-15% |
| B2B / high-ticket | 5-10% |
| Digital goods | 10-30% |
If suppliers reject the take rate at all three tiers, you have a unit-economics problem, not a product problem. Marketplaces with sub-10% take rates need scale that takes years; marketplaces with 25%+ take rates need a clear value-add suppliers cannot replicate themselves.
Step 4: Regulator and insurance pre-checks
This is the step nobody else writing about marketplace validation mentions, and it is where validated marketplaces die quietly.
If your vertical is mobility, food delivery, alcohol, financial services, healthcare, childcare, or short-term rental, you have a regulator. Call them before week one. Ask: "I am planning to operate a platform that connects [supply] with [demand] in [zip]. What licenses, bonds, or insurance do I need?" Get the answer in writing, even if it is an email confirmation of a phone call.
Then check insurance. General liability for a regulated marketplace can run $5,000 to $50,000 a year minimum, with deductibles that wipe out a year of revenue. Decide explicitly: who carries the risk, the platform, the supply side, or the demand side? Uber spent $20M+ in early-state legal fees because it answered "neither, we are just software." Airbnb fought NYC for a decade. Do not assume the regulator is asleep.
A 30-minute phone call now saves you a $500,000 cease-and-desist letter at month nine.
Step 5: Decide whether to build
After two to four weeks of interviews + concierge + take-rate tests + regulator calls, you have a decision. Use these criteria, in writing:
Pass:
- 10 manual paid transactions in two weeks
- At least 3 buyers AND 3 sellers came back for a second transaction
- Take rate accepted by at least 60% of supply at your target tier
- Regulator green light or a clear, affordable path to compliance
- Repeat-usage rate of 30%+ within 30 days
Fail:
- Cannot hit 10 transactions even with you doing all the work
- Single-side liquidity stalls (great supply, no demand, or vice versa)
- Take rate rejected at all three tiers
- Regulator says no, or path to compliance costs more than your seed round
- Buyers and sellers transact once and never come back
Pivots are normal. Most marketplaces pivot the vertical, the geography, or the take-rate model once during validation. The pivot is a feature of doing real validation, not a sign you wasted four weeks.
What to build first if you pass validation
You passed. Now you write code, but only enough to replace the most painful manual step. The order of usual operations:
- Matching (your spreadsheet is the bottleneck first)
- Payments (Stripe Connect, not Checkout, once you have multiple suppliers)
- Messaging (replace WhatsApp once volume hits 50+ matches/week)
- Reviews and ratings (only after 100 completed transactions)
- Scheduling automation (last; Cal.com handles you to ~500 bookings)
A v1 of this stack takes 4 to 6 weeks for one mid-tier engineer working full-time. If you are scoping the budget before you commit, our breakdown of the actual cost to build a marketplace walks the math at each tier. For a B2B marketplace specifically, scope a B2B SaaS MVP and validate a B2B SaaS idea cover the SaaS-flavored variant of this same playbook.
If you are bringing in an engineer for the post-validation build, the fastest path is to book a mid engineer for 4 weeks, ship the matching layer, then evaluate weekly. Cadence engineers start in 48 hours and the median time to first commit across the platform is 27 hours. Every Cadence engineer is AI-native by default, vetted on Cursor, Claude, and Copilot fluency before they unlock bookings. The 12,800-engineer pool means you usually get 4 shortlists inside 2 minutes of posting the spec. You can book a mid engineer here and use the 48-hour free trial to confirm fit before any money moves.
Do not hire a senior engineer or a CTO before $20k MRR. Mid-tier ($1,000/week) handles the first six months of marketplace v1 work. Lead-tier ($2,000/week) only earns its keep when you are scaling matching algorithms or rebuilding payments under load.
Common founder mistakes during marketplace validation
The five mistakes that account for most failed validations:
- Building before 10 manual transactions clear. Almost every marketplace that dies in year one wrote code first. The concierge phase exists to make sure you are not one of them.
- Choosing a city instead of a zip. "Launch in Austin" sounds ambitious; in practice it dilutes density, kills matching, and produces zero transactions. Pick the smallest geography that still has supply and demand.
- Skipping the regulator call. This kills marketplaces six to nine months in, when the founder has already raised a seed and burned $400,000 on engineering.
- Hiring a senior engineer before code complexity demands it. A mid-tier engineer at $1,000 per week can ship matching, payments, and messaging in 4 to 6 weeks. Senior at $1,500 makes sense for scale, not for v1.
- Confusing landing-page signups with demand. Email signups are intent, not demand. Ten paid transactions are demand.
- Picking a vertical with low transaction frequency. Bill Gurley's rule: weekly use beats monthly use beats annual use. A marketplace for once-a-year purchases needs 50x the user base to feel alive.
If you are stuck deciding whether to write the first line of code, our non-technical founder's guide to managing developers in 2026 covers the operating cadence we use with founders during the first six weeks of a build.
If you have hit pass criteria and need an engineer to ship the matching layer, the cleanest path is to book a mid engineer on Cadence for 4 weeks. Weekly billing, 48-hour free trial, no recruiter loop. Most founders ship matching v1 inside the first paid week.
Steps
- Pick a side. Default to supply unless you already control demand. Write down which side you will recruit first and why, in one sentence.
- Run 30 supply interviews and 30 demand interviews over 2 weeks. Use the question scripts above.
- Pick one zip code or a 5-mile radius and build the concierge stack: Notion + Stripe Checkout + WhatsApp + Cal.com + a Carrd landing page. Total cost: ~$50.
- Hand-match 10 paid transactions inside 2 weeks. Track repeat usage by buyer and seller.
- Fake-door three take-rate tiers (10%, 15%, 20%) with three supplier cohorts. Watch acceptance, not opinion.
- Call the relevant regulator and insurance broker for your zip code. Get answers in writing.
- Apply pass/fail criteria. If you pass, scope a 4 to 6 week v1 with one mid engineer, starting with the matching layer.
- Ship matching, then payments, then messaging. Stay manual on everything else until 100+ transactions clear.
FAQ
Which side of a marketplace should I focus on first?
Default to supply unless you already control a demand audience. Supply usually has a higher per-person economic incentive to join early, and once you have liquid supply, demand acquisition gets cheaper. The exception: if you can name 100 demand-side users you can text today (existing audience, prior network, established channel), lead with demand.
How many customer interviews do I need before building a marketplace?
Thirty per side is the working minimum. Twenty per side surfaces individual stories; thirty surfaces patterns. Sixty total interviews, done by the founder, take about 20 hours spread over two weeks. Hiring a research firm to run them is a waste of money during validation; the founder needs to hear the answers directly to pivot the next interview.
What does it cost to validate a marketplace idea?
Under $1,000, often under $100 in real cash. The validation phase is interviews (free), a landing page ($0-19), Stripe fees (2.9% per transaction), and roughly $50 in tooling for the concierge run. The expensive money comes after validation: a v1 build runs $4,000 to $8,000 for one mid engineer at $1,000 per week over 4 to 8 weeks.
How long does marketplace validation take?
Two to four weeks for interviews plus the concierge run. Add another one to two weeks for fake-door pricing tests and regulator pre-checks. If validation is dragging past six weeks, the founder is usually avoiding the kill-decision rather than gathering more signal.
Do I need a technical co-founder to validate a marketplace?
No. Validation is all manual: spreadsheets, Stripe Checkout, WhatsApp, in-person matching, a Carrd landing page. You only need engineering after the concierge proves transactions repeat. At that point, book one mid engineer on Cadence for 4 to 6 weeks. You do not need a CTO until you have paying volume and unit economics worth defending.
What is the chicken-and-egg problem in marketplace validation?
It is the cold-start problem: suppliers will not list without buyers, buyers will not show without listings. The validation answer is to break it manually. You become the missing side: you call suppliers and pre-fill the inventory, you call buyers and hand-match them, you run the marketplace as a concierge until both sides see enough activity to come back without prompting.