How to Validate a Marketplace Idea by Testing the Supply Side

To validate a marketplace idea, validate only the constrained side: the side that is scarce or hard to recruit, which is usually supply. Prove you can sign that side up and get it to complete real jobs. Then broker the first 10 to 20 transactions yourself with a spreadsheet, a phone and a payment link, before building any software.

Why is surveying both sides a waste of time?

The usual approach is two surveys. You ask buyers whether they would use a platform for X, and you ask sellers whether they would list on it. Both sides say yes, because saying yes costs them nothing. Buyers picture a full catalog and sellers picture free customers. Neither answer tells you whether the marketplace can work. A marketplace is only worth something when it has liquidity, which a16z defines as the likelihood that a seller finds a buyer, or that a buyer finds what they're looking for.

So validate the side that decides whether liquidity is possible. Lenny Rachitsky interviewed early team members at successful marketplaces, and 14 of the 17 companies put almost all of their early resources into growing supply. In those cases, the supply either brought in its own demand or word of mouth took care of it. Get the constrained side right and the other side mostly follows. Get it wrong and buyer enthusiasm won't save you.

Bill Gurley argues that a true marketplace needs natural pull on both the consumer and supplier side. He's right, but that's a test for scaling, not for validation. At the start, one side is the bottleneck. Your first job is to find out whether you can clear it.

How do you tell which side is constrained?

The constrained side is the one where each extra participant is hard to get and the marketplace can't run without them. Ask these questions about your idea and see which column wins.

Question	Points to supply constrained	Points to demand constrained
What is participation to this side?	Their main living, needing a license, a skill or inventory	Spare time, a spare room, low skill
Are current providers booked up?	Waitlists, "not taking new clients"	Idle capacity, providers advertising for work
Does the buyer care which provider they get?	Yes: quality is subjective and trust matters	Any competent provider will do
Is the buyer already searching?	Loud demand signals: complaints about not finding anyone, forum threads asking for recommendations	Buyers don't know the service exists
How many options does a buyer want before choosing?	Several	One is enough

The first row is where the standard advice goes wrong. The a16z glossary says it's typically easier to jumpstart supply than demand because suppliers are economically motivated. That's true for sign-ups. A plumber will happily fill in a free profile. Whether that plumber answers your lead within an hour on a Saturday, at your price, after your cut, is a separate question, and that's the one that decides liquidity. Count suppliers who complete a job, not suppliers who fill in a form.

Demand-constrained marketplaces do exist, and they have a clear signature: the supply is spare capacity that people are glad to earn money from. In Lenny's research, only Rover, TaskRabbit, and Zillow were demand constrained. Rover's team said supply was easy because dog lovers who work from home were happy to earn an extra $50. If your supply side looks like that, reverse the rest of this article and validate demand instead.

A quick field test settles most cases. Contact 10 providers and ask how soon they could take a new job. If most say "next month," supply is constrained, and what you offer buyers is access. If most say "tomorrow," supply is easy, and what you offer sellers is customers.

Why is supply usually the answer?

Buyers come for the supply. A buyer who opens your app and finds nothing leaves and doesn't come back. Lyft's team described the company as always supply constrained: there were no cars on Friday nights, and they eventually had to build a waitlist for supply. They found the turning point was a 3 minute ETA. Above that, riders shopped around, took the bus or walked. Nobody had to persuade riders to want a ride. They needed a car to show up.

There's also a gap in what you can research. Demand usually leaves public demand signals before you start: search volume, complaints, and reviews that say it took three weeks to find someone. You can measure all of that from your desk. Whether suppliers will accept your terms leaves no public trace. You only find out by asking them and watching what they do. The side you can't research is the side you have to test.

How do you validate the supply side?

1. Narrow to one category in one city, or one tight niche if you're online. Liquidity is local, so 20 providers in one city beat 200 spread across the country. If you're picking a city, Unthinkable's finder shows underserved business categories in any city, which is a sensible place to start a shortlist.
2. Write the supplier offer in one sentence: what they get, what you take, and when they get paid.
3. Recruit in person. At GrubHub, supply growth was all sales, door to door. The team took every excuse restaurants gave and removed it, until there was zero downside to sign up. Copy that: they pay only per completed job, there's no setup fee, and they can quit anytime.
4. Track three numbers: providers contacted, providers who agreed, and providers who completed a first job. Only the last number counts.
5. Send them the first jobs yourself. Don't wait for buyers to find you. You'll source them by hand in the next step.

How do you run the first transactions by hand?

DoorDash is the standard example. When it was still called Palo Alto Delivery, the only Dashers were the co-founders. They ran orders using Google Voice, the Find my Friends app and their own cars, and their marketing was fliers on dorm bulletin boards. There was no dispatch software. They learned what delivery actually involved before they wrote code for it.

Your version:

• A one-page site describing the service, with a request form. No catalog and no accounts.
• A spreadsheet with one row per request: date, provider matched, price, hours to match, outcome, buyer rating, and whether each side would rebook.
• Manual matching. You match every request yourself by phone or text.
• You hold the money. Take payment from the buyer through a payment link and pay the provider once the job is done. Holding the money is what makes you a marketplace rather than a directory. It also tests whether buyers will pay you instead of paying the provider directly.
• One question after every job, to both sides: would you do this again through me?

Aim for 20 completed transactions. With fewer than that, one bad week skews everything.

What does the math look like after 20 manual jobs?

Say you broker weekend mobile bike repair in one city. The average job is $120 and you take 15%.

Per job	Amount
Buyer pays	$120.00
Provider receives (85%)	$102.00
Your gross take	$18.00
Card processing at 2.9% + 30¢ on $120	$3.78
Your net	$14.22

Over 20 jobs you net $284.40. If each job took 45 minutes of coordination, that's 15 hours of your time, or about $18.96 an hour. Don't judge the idea on the hourly rate, because software should remove most of those 45 minutes. Judge it on two things the spreadsheet shows you.

First, the cost of recruiting supply. Say you contacted 40 mechanics, 12 agreed, and 7 completed at least one job. That's 7 ÷ 40 = 17.5% from first contact to active provider. To get 10 active mechanics in a second city, you'd need to contact about 10 ÷ 0.175 = 57. That number tells you what expansion really costs, and a buyer survey could never give it to you.

Second, how thin your take rate is. Card fees ate $3.78 of your $18, or 21%. Now run a $40 job. Your gross take is $6.00 and the fee is $1.46 (2.9% of $40 is $1.16, plus 30¢), so the fee eats 24%. The fixed 30¢ hurts more as tickets get smaller. If your category has small tickets, either raise the take rate now or accept that you need much higher volume per provider.

What results should make you stop?

Signal	Keep going	Stop or rethink
Agreed providers completing a first job	Most do	They agree, then go quiet
Providers asking for more work	Yes, without being asked	Only when it suits them
Buyers rebooking	Through you	Directly with the provider after job one
Your take rate	Accepted without pushback	Every provider haggles
Time to match a request	Falls as you add providers	Flat or rising

The strongest single signal is providers asking you for more work without being prompted. It means you've solved something the constrained side can't solve alone, and everything else can be fixed.

The worst signal is buyers going straight to the provider after the first job. If that happens, your marketplace only adds value at the first introduction, which makes it a lead-generation business. Either charge like one, with a fee per lead, or add something that keeps both sides coming back to you, such as payment protection, scheduling or a guarantee. Don't build a transaction platform for customers who will leave it after one transaction.

When should you start writing code?

Start when the spreadsheet breaks, and let the part that breaks decide what you build first. If matching is eating your evenings, build matching. If chasing payments is the problem, build checkout and payouts. If buyers keep asking whether a provider is free on Saturday, build availability. The manual phase is more than validation: it's the spec for your software, written from 20 real transactions instead of guesses.