Your First 10 Customers Won't Come From Marketing

In the early days of Airbnb, the founders did something that sounds almost silly for a tech company. They went to New York and went door to door. They signed up new hosts. They helped existing hosts improve their listings. They took professional-looking photos of hosts’ apartments themselves.

Paul Graham tells this story in his July 2013 essay Do Things That Don’t Scale. He tells another one about Stripe. Most founders, when someone shows interest, ask politely: “Will you try our beta?” Stripe’s founders, Patrick and John Collison, didn’t wait. If you agreed to try Stripe, they would say, in effect, “Right then, give me your laptop,” and set you up on the spot. At Y Combinator, Graham writes, this became known as the “Collison installation.”

Neither story is about marketing. Nobody bought an ad. Nobody wrote a blog post optimised for search. The first users came because the founders went and got them, one at a time, by hand.

That is the subject of this post. If you’re starting something and you have zero customers, your first 10 will almost certainly not come from marketing. They’ll come from conversations, from people you can reach directly, and from you doing work that feels too small to matter. It isn’t too small. It’s the job.

Why marketing doesn’t work at zero

Marketing is an amplifier. Ads, search engine optimisation (SEO, the work of getting your pages to rank on Google), and social media all take something that already works and spread it to more people.

At zero, there is nothing yet to amplify. Three things are missing.

1. No proof. A stranger who sees your ad has no reason to trust you. There are no reviews, no customer names, no “used by 200 hosts.” Every first purchase is a risk for the buyer, and you’ve given them nothing to lower it.

2. No message. You don’t yet know which words make the right person say “that’s me.” You might describe your product as “smart scheduling,” while your real customers would describe their pain as “my cleaner showed up on the wrong day again.” You only learn their words by hearing them, many times.

3. No funnel. A funnel is the path from “first heard of you” to “paid.” At zero, you don’t know your conversion rates, your price, or where people drop off. Paying for traffic into an untested funnel is like pouring water into a bucket before checking it for holes.

There’s also a simple math problem. Suppose an ad gets one click for every 100 people who see it, and one sale for every 100 clicks. To get 10 customers, you need 1,000 clicks and 100,000 views. You can pay for that. But if your page, price or message is wrong, you’ll get almost nothing back, and you won’t know why. A conversation tells you why in ten minutes.

At zero, you don’t need reach. You need to learn. Conversations are the fastest way to learn.

The example we’ll follow

To keep this concrete, I’ll use one fictional business through the whole post.

Fictional example: TurnBoard. A founder, call her Mai, wants to build a simple tool for short-term rental hosts. It takes each host’s bookings and turns them into a cleaning schedule, then sends each cleaner their jobs for the day. Her guess: hosts with 2 to 10 units waste time every week coordinating cleaners on chat apps, and mistakes cost them bad reviews.

Mai has no product, no customers and a small budget. Her goal: 10 paying customers in 4 weeks. Everything below is what she should do, step by step.

(TurnBoard is invented. The numbers in this post are illustrations, not data from any real company, including mine.)

Where first customers actually come from

There are only a few places. All of them are slow, manual and personal. That’s the point.

1. Your own network, and one step beyond it

Start with people you already know, and people they know. Not to sell to your friends out of pity, but because warm introductions get replies and honest time.

Mai makes a list of everyone she knows who hosts, manages properties, or cleans for hosts. Then she asks each of them one question: “Who’s the most stressed host you know?” The second-degree contacts are the gold. They don’t know her, so they won’t be polite for her sake, but they’ll still take the call because a friend asked.

2. Communities where the problem is discussed

Your customers already gather somewhere. For hosts, it’s Facebook groups, local host meetups, forums, and the comments under hosting videos. For tutoring centres, it’s parent groups and teacher associations.

The rule is: join, help, then ask. Answer questions. Share what you know. Only after people recognise your name do you post something like: “I’m researching how hosts coordinate cleaners. If you manage more than two units, I’d love 15 minutes to hear how you do it. No pitch.” Read each group’s rules first. Many ban promotion, and they should.

3. Direct outreach

This means writing to specific people, one by one. Not a mass email. A short message that shows you looked at them.

For Mai, a host with five listings in one area is easy to find on booking sites. A message might read: “Hi, I saw you run five places near the beach. I’m talking to hosts about how they schedule cleaners between check-outs. Would you have 15 minutes this week? I’m not selling anything yet.”

Most won’t reply. That’s normal. Reply rates vary a lot by market, but if almost nobody answers, your message or your target list probably needs work.

4. Go where the problem physically happens

Airbnb went to New York because that’s where the hosts were. Mai can go where turnovers happen: talk to cleaners at a laundry service, visit a property manager’s office, attend a local host meetup. In person, you see the whiteboard of cleaning dates and the three phones on the desk.

5. Partnerships with people who already serve your customer

Someone already has your customers’ attention. For hosts: cleaning companies, laundry services, property managers, the local accountant who does hosts’ taxes. A partner who says “you should talk to Mai” is worth a hundred ads, because the trust transfers.

Skip formal deals. Start with one person happy to introduce you to three of their clients.

How to run a customer conversation

Most founders ruin their first conversations by pitching. They describe their idea, the other person nods, says “that sounds great,” and the founder leaves happy with nothing useful.

Rob Fitzpatrick wrote a short book about exactly this problem, The Mom Test. The name comes from the idea that you should ask questions so good that even your mom couldn’t lie to you about your business, even though she loves you and wants you to succeed. His core rules, as summarised in many reviews of the book:

1. Talk about their life, not your idea. Once you describe your idea, people get polite. Ask about their problems, habits and goals instead.
2. Ask about specifics in the past, not opinions about the future. “Would you use this?” invites a guess. “When did this last happen, and what did you do?” gets a fact.
3. Talk less and listen more. If you’re talking most of the time, you’re pitching, not learning.

He also warns about three kinds of bad data: compliments (“great idea!”), fluff (generic or hypothetical statements like “I usually…” or “I would definitely…”), and ideas (feature requests, which you should dig under to find the real reason).

Here’s a short script Mai could use. It takes about 15 to 20 minutes.

OPEN (1 min)
"Thanks for the time. I'm researching how hosts handle cleaning
between guests. I'm not selling anything today. Just want to
hear how you do it."

CURRENT LIFE (5-8 min)
- Walk me through the last turnover you had. What happened, step by step?
- Who did the cleaning? How did they know when to come?
- What tool did you use to tell them? Chat, sheet, calendar, something else?
- How long did coordinating it take you that week?

PAIN, IN THE PAST (5-8 min)
- When did a cleaning last go wrong? What happened?
- What did it cost you? A refund, a bad review, your evening?
- What have you already tried to fix it?
- Have you paid for anything to solve it? What, and why did you stop?

CLOSE (2 min)
- Who else do you know who deals with this?
- Can I come back in a week or two and show you something rough?

And the questions that sound helpful but produce bad data:

One more tip: take notes in their words. If three hosts say “my cleaner didn’t know the guest checked out early,” that phrase is your future headline. That’s how you build the message you didn’t have at zero.

What a real signal looks like

At the end of a good conversation, you’ll feel encouraged. Be careful. Feeling encouraged is not data.

The Mom Test suggests a cleaner test: did the person commit something they value? Reviews of the book describe three kinds of commitment:

• Time. They agree to a next meeting with a clear goal, or to try a rough version for a week.
• Reputation. They introduce you to their business partner, their boss, or three other hosts. They’re putting their name next to yours.
• Money. They pay a deposit, sign a letter of intent (a short written promise to buy if you deliver), or pre-order.

The stronger the commitment, the stronger the signal.

Fitzpatrick’s point is that a meeting with no commitment and no next step is a meeting that didn’t go anywhere, no matter how nice it felt. So always ask for something at the end. A next meeting, an introduction, or a small payment. A “no” is fine. A “no” is information. A vague “maybe later” is the thing to watch out for.

Pre-orders and deposits

Asking for money before you’ve built anything feels awkward. Do it anyway. It’s the clearest test there is.

Mai might say: “I’m going to run this by hand for five hosts next month. I’ll build your cleaning schedule every week and send your cleaners their jobs. It’s $30 a month. If it’s not useful after the first month, I’ll refund you.”

If nobody says yes, she has learned something important before writing code. If five say yes, she has five customers and a reason to build.

Do the work by hand first

Here’s the part that feels most backwards to engineers like me: you can deliver the product before you build it.

There are two common ways to do this.

Concierge MVP. MVP means minimum viable product, the smallest thing that delivers real value. In a concierge MVP, you deliver the service manually, and the customer knows a person is doing it. You’re like a hotel concierge: personal, hands-on, not automated.

Wizard of Oz MVP. The customer sees something that looks automated, like a form or a simple website, but a person does the work behind the scenes, like the wizard behind the curtain in the film. The customer may not know it’s manual.

Paul Graham’s essay has an example of the second kind. Stripe offered merchants “instant” merchant accounts, but in the early days, the founders were signing those users up for traditional merchant accounts by hand, behind the scenes.

He also describes another manual move from his own company, Viaweb. When some merchants hesitated to use the software, Viaweb offered to build their online stores for them. Doing it themselves taught them how their software felt to use.

For TurnBoard, the concierge version is simple:

1. Each host shares their booking calendar with Mai.
2. Every evening, Mai reads the calendars and writes the next day’s cleaning plan in a spreadsheet.
3. She sends each cleaner a message with their jobs, addresses and check-out times.
4. If a booking changes, she updates the plan and tells the cleaner.

No app. No code. Maybe one shared Google Sheet. It will take her an hour or two a day for five hosts. That’s fine. It won’t scale, and it isn’t supposed to. It tells her which steps are painful, which mistakes happen, and what the hosts actually care about. When she does build, she’ll build the parts she has already done 200 times by hand.

A note on honesty. If you run a Wizard of Oz test, be careful. Never mislead people in ways that would hurt them if they knew, and be especially careful with their data and money. I’d lean toward telling customers early that “for now, a person handles this.” Most first customers like that. It feels like service.

Founder-led sales

At this stage, you are the sales team. Not a hired salesperson, not an agency. You.

That’s because selling at zero isn’t really selling. It’s research with a price attached. Every “no” tells you something about your target, your offer or your price. If you hand it to someone else, the lesson goes to them.

Founder-led sales in practice:

• You write the outreach messages yourself, and change them every week based on what got replies.
• You run every call, and write notes after each one.
• You handle every objection personally. “Too expensive” from three hosts in a row is a pricing lesson. “We already use a spreadsheet” is a competitor lesson. (Often your real competitor is a spreadsheet.)
• You onboard every customer yourself, Collison-style. Don’t send a sign-up link and hope. Sit with them, or get on a call, and set it up together.

Graham also mentions that Wufoo sent each new user a hand-written thank-you note. At zero, personal touches like that are your edge: bigger companies can’t care that much about one customer. You can.

A 4-week plan: from 0 to 10

Here’s a realistic plan for Mai. The targets are illustrative, but the arithmetic is checked.

Her working assumptions:

• About 1 in 3 people she contacts agree to a conversation (warm intros convert better, cold messages worse).
• About 40% of conversations confirm a real, recent problem.
• About half of those agree to try the manual service.
• About half of the trial users pay after the first week or two.

Ten customers at $30 a month is $300 a month. That’s not a business yet. But it is proof: ten people who pay, a message that works, and a list of what to build.

Now spread the work over four weeks:

Check: 60 + 90 + 90 + 60 = 300 contacts. At 1 in 3, that’s 20 + 30 + 30 + 20 = 100 conversations, which matches the cumulative column (20, 50, 80, 100).

Week 1: Learn. No offers yet. Talk to 20 people. Write their words down. By Friday, Mai should be able to describe the problem in one sentence the way hosts describe it.

Week 2: Offer. Keep talking. With people who clearly have the problem, offer the manual service. Ask for a commitment: a start date, an introduction, or a deposit.

Week 3: Deliver by hand. Run the concierge service for the first trial users. Do it well. After a week, ask the simple question: “Do you want to keep going at $30 a month?”

Week 4: Convert and ask for referrals. Turn trials into paying customers. Ask every happy customer: “Who else should I talk to?” Referrals from paying customers are the warmest leads you will ever get.

If the numbers miss, that’s information. If fewer than 1 in 10 reply, fix the message or the target list. If conversations don’t confirm the problem, the problem may be smaller than you thought. If trial users won’t pay, the price, the value or the customer is wrong. All three are cheap to learn in four weeks and very expensive to learn after six months of building.

The funnel, as a picture

Your real rates will be different. Some founders need 300 conversations; some close 10 customers from 30. What matters is that you measure each step, so you know which one to fix.

The mistakes

1. Pitching too early

The fastest way to ruin a conversation is to open with your idea. People get polite, you get compliments, and you learn nothing. Talk about their life for the first two-thirds of every call. If they ask what you’re building, give one sentence and turn it back: “But tell me how you handle it today.”

2. Counting compliments

“Ten people loved it!” Did any of them pay, book a next meeting, or introduce you to someone? If not, you have ten compliments and zero customers. Count commitments, not compliments. Keep a simple sheet: name, date, problem confirmed (yes/no), commitment (none / time / reputation / money).

3. Building before talking

This is the engineer’s favourite mistake. I understand it. Building is comfortable; talking to strangers is not. But a beautiful product nobody asked for is the most expensive way to find out you were wrong. Talk first. Deliver by hand second. Build third.

4. Talking to the wrong people

Your friend who rents out one room twice a year will be nice to Mai, but she isn’t the customer. A host with eight units and four cleaners is. Define who has the problem worst, and talk to them.

5. Spending on ads to “test demand”

Ads can test a headline. They can’t tell you why a host didn’t buy. At zero, spend the money on your own time: coffee with hosts, a day at a host meetup, a small gift card to thank someone for 20 minutes.

Why this is worth it

Getting your first 10 customers by hand leaves you with ten people who know you personally. They’ll tell you when something breaks, forgive early mistakes, and introduce you to the next ten. And you’ll understand their problem better than any competitor, because you solved it yourself, every day.

Marketing comes later. Once you have ten paying customers, you have proof (their names and stories), a message (their own words), and a funnel (your real conversion rates). Then an ad has something to amplify. Then SEO has something to say.

Your first 10 customers are not a marketing problem. They’re a conversation problem. Go have the conversations.

Key ideas

• At zero, marketing has nothing to amplify. No proof, no message, no funnel. Conversations fix all three.
• First customers come from people you can reach directly: your network and one step beyond it, communities, direct outreach, where the problem happens, and partners.
• Talk about their life, not your idea. Ask about specific past events, and listen more than you talk.
• Compliments are not signals. Real signals are commitments of time, reputation or money: a next meeting, an introduction, a deposit.
• Deliver by hand before you build. A concierge or Wizard of Oz version teaches you what to automate.
• Founder-led sales means you do every step yourself, from the outreach message to the onboarding, because every “no” is a lesson.
• Plan in numbers and measure each step. If a step misses, fix that step, not everything.

Sources: Paul Graham, Do Things That Don’t Scale (July 2013); Rob Fitzpatrick, The Mom Test; summaries of The Mom Test’s rules and commitment types by Sachin Rekhi, Shortform and Michael Lynch; concierge vs Wizard of Oz definitions from Mind the Product.

Related: Default Alive or Default Dead?, Unit Economics in Plain English, and Your Real Competitor Is a Spreadsheet.