How do you get your first 10 customers?

Your first 10 customers will not arrive because your launch post is clever or your website looks finished. You get them by identifying a narrow group with an expensive or irritating problem, speaking to those people one at a time, and asking them to pay for a specific result. At this stage, sales and customer research are the same job.

That changes the goal. You are not trying to build a repeatable acquisition machine yet. You are trying to find ten independent pieces of evidence: ten people or companies that feel the problem, understand your offer, can decide, and exchange money for it. Personal contacts, communities, cold outreach, and partnerships are four routes into those conversations. None can rescue a vague customer or a weak offer.

Ten customers is a sales problem, not a launch

The fastest route to ten paying customers is manual sales led by a founder. A broad launch can create attention, but attention hides the questions you need answered: who cares enough to act, what words make the problem recognizable, what blocks a purchase, and whether the buyer will pay now.

Paul Graham's essay "Do Things That Don't Scale" says founders usually have to recruit early users manually. The useful part of that advice is often softened into "be scrappy." I take it more literally. Find the person, write the message, run the setup, watch the first use, and ask for the payment yourself. You need contact with the whole buying process, including the uncomfortable pause after you name the price.

Patrick McKenzie's Stripe Atlas guide, "Your first 10 customers," makes a related distinction: early customers should have bought something similar before. Someone facing a problem for the first time may sound enthusiastic while having no budget, buying process, or realistic sense of value. An early adopter is not merely curious about new products. She has already spent money or staff time trying to solve the problem you address.

Count a customer only when money changes hands. Free testers, friendly advisers, newsletter subscribers, and verbal commitments can help, but they test different things. A design partner who pays a reduced fee counts if the work and terms are explicit. A person who promises to "definitely use it when it is ready" does not.

You also do not need a polished product before selling. You need enough truth to state what exists today, enough capability to deliver the promised result, and a safe way to support the customer. If part of the service is manual behind the scenes, say so. Passing off a manual service as finished automation creates the wrong evidence and an avoidable trust problem.

Set a working window of four to six weeks. That is long enough to test several messages and follow up, but short enough to expose avoidance. If you spend that window changing fonts, posting general content, or adding features without live buyer conversations, you are not working on the problem of the first ten.

Define the customer narrowly enough to find her

A usable early customer profile lets you name actual prospects today. "Small businesses" is not a segment. "Independent physical therapy clinics in California with two to ten locations that still schedule staff in spreadsheets" gives you a list, a likely buyer, a visible workflow, and a reason the problem may hurt.

Write the hypothesis in one sentence with five parts: the type of organization or person, the specific situation, the painful task, the current workaround, and the person who can approve a purchase. For example: "Operations leads at seed stage US software companies that hire internationally spend several hours each week reconciling contractor invoices in spreadsheets and can approve a monthly operations tool."

That sentence is a prospecting filter, not positioning copy. Test every name against it. If you cannot tell why a prospect fits before contacting her, the profile is still too broad.

Look for evidence of pain rather than demographic resemblance. A recent job posting, a public request for recommendations, a new location, a compliance deadline, a spreadsheet template, or a complaint about an existing method can all show that the work exists. Do not invent personal detail to manufacture intimacy. Reference only the business evidence that connects directly to your offer.

Build a simple list of 40 prospects to start. Give each one a reason for inclusion and a likely route to contact. Forty is large enough to reveal whether a message works and small enough that you can research each person. Add more only after you have read every rejection and revised the weak parts.

Use four columns beyond the contact details:

Do not score prospects with a complicated formula. Rank them A, B, or C based on pain, ability to buy, and access. Start with A prospects. If nobody qualifies as A, revisit the customer hypothesis before increasing outreach volume.

Start with people who can tell you the truth

Your personal network is the best first channel when you use it for relevance and trust, not obligation. Friends should not buy a product they do not need. Former colleagues, clients, vendors, classmates, advisers, and friends of friends can become good early customers when they match the profile and can judge the work honestly.

Make two lists. The first contains possible buyers you know directly. The second contains people who understand the market and could make one relevant introduction. Search old email threads, calendars, professional groups, and contacts from prior jobs. Memory alone favors your closest friends, who are often the least useful prospects.

Send a direct note to a possible buyer:

Hi Maya, I am working on a service for operations leads who lose time reconciling international contractor invoices. It turns the monthly files into one review queue, and I am looking for three teams willing to use it on a real cycle. You mentioned this process at your last company. Is it still a problem where you are now? If yes, would you spend 20 minutes showing me how you handle it today?

This message names the audience, problem, current state, and small request. It does not ask for vague "feedback," which invites polite opinions. It asks whether the problem exists and requests access to the current workflow.

For an introduction, make the connector's work easy:

I am looking for an operations lead at a US startup that pays international contractors and still reconciles invoices manually. I help turn that monthly process into a review queue. Does one person come to mind? If so, I can send a forwardable note in two sentences. No worries if nobody fits.

Never ask someone to "share this with your network." That hands her an unpaid marketing job and produces weak introductions. Ask for one person who meets a specific description. Give the connector an honest way to decline.

Warmth gets you a reply, not a sale. Run the same discovery and pricing conversation you would run with a stranger. If a friend receives special terms, put the scope, price, discount, and end date in writing. Otherwise you may count a favor as demand and discover later that nobody outside your circle will buy.

For women building in the Bay Area, Sisters can make this work less solitary: members can ask peers for candid feedback, advisers, and relevant introductions. Membership requires an invitation and costs nothing after a successful application, so treat the community as a place to build real working relationships, not a directory to mine.

Earn your place in communities before you pitch

Communities produce customers when people can see your judgment before they see your offer. Choose two or three places where your narrow customer already discusses the painful work: a professional association, a specialized online group, a local meetup, an alumni channel, or an operator community. A large founder forum may be less useful than a small group for clinic managers if clinic managers are your buyers.

Read the rules and the recent conversation before posting. Some groups ban promotion or require it in a dedicated channel. Even where pitching is allowed, a launch announcement from a stranger rarely earns trust. Answer questions with enough detail that a reader can act without buying anything. Mention your work only when it explains why you know the problem.

Keep a small community routine:

1. Find five recurring questions connected to the workflow you solve.
2. Answer two with concrete examples from your own work.
3. Ask one informed question about how members handle a disputed part of the process.
4. Invite relevant respondents to a short conversation only after they engage.
5. Return with what you learned, without exposing private details.

The transition from discussion to a sales conversation should be explicit. Try: "Your comment about rechecking every invoice matches the process I am working on. I have a rough version that handles that review. Would you be open to testing it on one monthly cycle? I charge $300 for the pilot because I want to test the buying decision as well as the workflow."

That price may be wrong for your product; the structure is the point. Name the observed pain, the limited use case, the commitment, and the fact that payment is part of the test.

Do not send unsolicited direct messages to everyone who reacts to a thread. Respond in context, then ask permission to continue privately. Trust drains quickly when members suspect that every useful conversation feeds a prospect list.

Events work the same way. Do not measure a meetup by business cards collected. Aim for two conversations with people who match the profile, then send a note within a day that refers to the actual discussion and proposes one small next action. A meeting without a dated next step is pleasant networking, not a customer process.

Cold outreach works when the evidence is specific

Cold outreach is useful after you can recognize a good prospect and describe her problem in her language. It fails when founders buy a giant list, write one generic message, and call the low response rate proof that buyers do not care.

Research one business reason to contact each prospect. Good triggers include a new role responsible for the workflow, expansion that increases the workload, a public complaint, a required process change, or evidence that the team uses the workaround you replace. Personal trivia is not personalization. Mentioning someone's college or hobby when it has nothing to do with the purchase feels invasive.

A plain first email can fit in five lines:

Subject: contractor invoice review at Acme

Hi Jordan, I saw that Acme is hiring in three more countries. Operations teams at that stage often reconcile contractor invoices across separate files. I built a review queue that flags missing fields before approval. Would it be useful if I sent a walkthrough that takes 90 seconds using a sample invoice set? If this is not yours, who owns contractor payments?

The message shows why you chose the company, states the problem without exaggeration, explains the mechanism, and asks for a small response. It does not demand that a stranger spend thirty minutes on a call. Do not attach a deck or bury the request under your founding story.

Follow up two or three times across roughly two weeks, each time adding something useful: a clearer example, an answer to a likely objection, or a relevant observation. "Just bumping this" consumes attention without earning it. Stop after the sequence, record the result, and move on. A later trigger can justify a new message.

Use replies to diagnose the funnel. If people do not reply, your list, subject, or relevance is weak. If they reply but refuse a call, the problem may lack urgency or the request may be too large. If calls happen but nobody tests, your proposed result is unclear. If people test but will not pay, examine value, trust, implementation work, and price separately. Cutting the price before you know which one failed destroys information.

US founders also need to treat commercial email as regulated communication. The Federal Trade Commission's CAN-SPAM compliance guide says the law covers B2B commercial email as well as bulk campaigns. Use accurate sender information and subject lines, include a valid postal address and a clear opt-out method where the rules require them, honor opt-outs, and do not assume an outreach vendor carries the legal responsibility for you. Get legal advice for your exact situation and every jurisdiction you contact.

Partnerships borrow trust, not demand

A good early partnership gives you credible access to customers who already have the problem. It does not mean placing two logos on a webinar page and hoping leads appear. The partner needs a reason to introduce you, a defined audience, and a limited offer that makes the partner look useful.

Look for businesses that serve the same buyer before or after your part of the workflow without competing with you. An accountant may meet a payroll tool's buyer. A commercial interior designer may meet an office inventory service's buyer. A specialist recruiter may meet a manager who needs an onboarding product. The connection should be visible in the customer's actual sequence of work.

Start with a customer need, not a partnership label:

Several of your clients probably start paying international contractors before they have a clean invoice review process. I run a paid setup for a single cycle that organizes those approvals. If two clients have that problem now, I will handle the setup personally and send you a short written account of what changed. Would that be useful to them?

Offer a bounded pilot before negotiating referral percentages, joint marketing plans, or integrations. You are testing whether the partner recognizes demand and will make an introduction. One qualified introduction is more informative than a memorandum about future cooperation.

Protect the relationship. Tell the partner who is a fit, what you will say, how you handle customer data, and when you will report back. Never add the partner's contacts to a campaign. Ask before using the partner's name in sales material. If a referral fee applies, disclose it where law or professional rules require and write down who pays, when it is earned, and what happens on a refund.

Reject partners who offer reach without relevance. A newsletter with 50,000 general readers can produce less than a consultant with five active clients in your niche. Early distribution is about concentrated trust. You can chase audience size after you know which promise converts.

Price the first deal to test value

Early pricing should test whether the result matters enough to buy while covering the unusual work you must do to deliver it. Free is often popular because founders fear rejection and buyers enjoy experiments with little risk. It is usually the wrong default for the first ten: it removes the buying decision you most need to observe and attracts people with weak urgency.

Start from the customer's economics. Estimate the cost of the current process, the cost of errors or delay, the revenue your offer may enable, and the budget owner who benefits. Do not claim a precise return you cannot support. Use the estimate to set a price that is materially smaller than a credible result and large enough that approval means something.

Then check your own delivery floor. The US Small Business Administration expresses unit break-even as fixed costs divided by price minus variable cost. Your first ten will not establish a stable break-even point, but the formula forces a useful question: does each sale contribute toward the business, or does support make every new customer deepen the loss? Include founder time, onboarding, payment fees, materials, and any custom work when you inspect variable cost.

For an early B2B offer, one paid pilot is often cleaner than a permanent discount. Define:

• one business result and what is outside scope;
• a fixed time window or usage limit;
• the work each side must provide;
• the pilot fee and payment date;
• the price or decision process if the customer continues.

Suppose a comparable solution costs $500 to $1,500 per month, your buyer spends about $2,000 a month on manual work, and your early version requires personal setup. Do not automatically charge $49 because the product has fewer features. You might offer a $750 pilot lasting four weeks with setup included, then propose $500 per month if the agreed workflow succeeds. Those numbers are an example, not a benchmark. Change them to match the buyer's economics and your delivery cost.

Stripe Atlas's SaaS pricing guide argues that early software companies often overestimate the advantage of a low price. I agree for considered business purchases: a tiny price can signal that the product is a side project and leave no room for support. I would qualify the advice for consumer products and buyers with little cash, where a higher price can simply exclude the people with the problem. The test is not whether the price feels bold. It is whether the right customer understands the exchange and can approve it.

Call discounts by their name. Write "regular price $500 per month, early customer price $300 for three months" instead of quietly anchoring the product at $300 forever. Ask for something concrete in return, such as weekly feedback sessions, permission to measure the workflow, or a reference if the customer is satisfied. Never make a positive testimonial a condition of the discount.

Run every conversation toward a paid decision

A first sales call should uncover a real current process and end with a decision or a dated next action. It is not a product tour. Spend most of the call on what happened the last time the buyer faced the problem, because specific past behavior predicts a purchase better than opinions about what she might want.

Ask for the sequence: "Walk me through the last contractor invoice cycle." Then ask who did the work, how long it took, what went wrong, what they tried, what the failure cost, and who approves a new tool. Avoid asking, "Would you use a product that makes this easier?" Most considerate people will say yes, and you will learn almost nothing.

When the pain and authority are real, propose the smallest paid engagement that can produce evidence. State the result, scope, timing, responsibilities, and price. Then stop talking. A founder who answers her own price with an immediate discount teaches the buyer to wait.

Objections need classification. "Too expensive" can mean the value is unclear, the wrong buyer is on the call, there is no current budget, the implementation looks risky, or the price truly exceeds the result. Ask, "Compared with what you expected?" and "Which part is hard to justify internally?" The answer tells you whether to change the offer, involve another person, reduce scope, or walk away.

After the call, send a short written recap with the customer's current situation, the promised result, open questions, owner, and next date. Keep a decision log. Exact language matters: if four buyers describe the same pain with the same phrase, that phrase belongs in your sales copy. If several stall at security review or procurement, the next product work may be operational documentation rather than another feature.

Ask for referrals after you have delivered a result, not immediately after someone signs. A useful request names the fit: "Who else runs contractor operations across several countries and still reviews invoices in separate files?" Give the customer a forwardable note in two sentences, and ask permission before naming her or her company.

Walk away from customers whose requested work pulls you outside the chosen problem, whose buying process will outlast your runway, or whose behavior makes delivery unsafe. Revenue from a custom project can look like traction while turning your company into an agency you never meant to build.

Use ten customers to choose what scales

Ten customers should leave you with a sharper market and a body of evidence, not a victory photo. Review the set customer by customer: what triggered the search, which channel created trust, how long the decision took, what they paid, what work delivery required, which promise mattered, and whether they continued or referred someone.

Separate channel performance from customer quality. Your network may close quickly but contain poor lasting fits. Cold outreach may take more attempts but reveal a repeatable buyer. A partnership may deliver only two customers, both excellent. With such a small sample, do not crown a winner from conversion percentages alone. Read the cases.

Keep a weekly scoreboard with counts for qualified prospects added, first conversations, paid proposals, wins, losses, and active customers who reached the promised result. Followers, impressions, and website visits belong elsewhere. They do not explain whether you can find and serve a buyer.

After ten, choose one primary channel for the next 25 customers and keep one secondary channel alive. Pick based on customer fit, speed, delivery burden, and how much control you have over access. Document the message and steps only after you have repeated them successfully. Automation applied before that point makes an unproven process fail faster.

Do not pretend ten customers prove a vast market. They can prove that a narrow group will buy a defined result through a known conversation. That is enough to make the next decision with evidence. Preserve the manual contact until you can explain, in the buyer's words, why customer ten paid and why prospect forty did not.