The first ten customers, one channel at a time
The first ten customers are two jobs pretending to be one. Three of them come from people you can already name, and no channel is involved. The other seven have to come from somewhere repeatable, and that means picking one channel and running it properly for four weeks. This page has which channel by what you sell, the numbers to expect at every step so you can tell a broken funnel from a small one, and four weeks of a real launch with the counts.
Almost every guide to the first ten customers is really a guide to the first ten thousand, shrunk. It lists eleven channels, tells you to build an audience, and leaves you running a bit of everything at a tenth of the effort each. At the end of a month you have six customers and no idea which thing produced them, which is the same as having none, because you can't do it again.
The honest shape is this. Your first two or three customers will not come from a channel at all. They'll come from the twenty people you can already name — an ex-colleague, a neighbour, someone in the group you're already in — and they'll come quickly, and they'll teach you almost nothing about acquisition, because those people were always going to be reachable. Take them. Then notice that the list is finite, and that customer four is the first real one.
Customers four to ten are the actual exercise, and the job isn't to get them. It's to get them from one place, in a countable way, so that at the end you can say a sentence like: a customer from this costs me $75 and forty minutes, and is worth $180. Once you can say that, the business has a lever. Until then every good month is luck and every bad month is a mystery.
So: one channel, four weeks, four numbers written down each Friday. What follows is how to pick the channel, what the numbers should look like at each step, and how to tell whether a disappointing week means the funnel is broken or just too small to read yet — which is the question that actually costs founders their first quarter.
Pick the channel by what you sell
Channel choice is mostly determined by how your buyer is findable. If they're defined by where they live, buy geography. If they're a list of named businesses, write to the list. If they buy your judgement, they have to see you think first. Everything else is taste.
Pick the row that matches what you sell and ignore the others for four weeks. The last column is the kill number — write it down before you spend anything, because a number chosen in advance is the only one you'll believe in week three.
| What you sell | The one channel | Why this one | Four-week test | Stop if, after four weeks |
|---|---|---|---|---|
| A service to households in one town — repairs, cleaning, lawns, dog walking | Paid social in a tight radius: Facebook and Instagram, three to five miles | Your buyer is defined by geography, which is the one thing ad targeting is genuinely good at, and you need volume before word of mouth exists. | $250–$400 | Under 3 paid bookings from 200+ visits, after the checkout has been tested on a phone |
| A service to local businesses — bookkeeping, commercial cleaning, IT, trades on contract | A named list of 40–200 businesses, emailed slowly and followed up | Your buyers are findable, countable and reachable by name — and small lists reply at 15–20% against 8% for lists over 500. | An hour a day, plus a domain | Under 5 replies from 100 well-chosen emails |
| Judgement — consulting, design, coaching, advisory | The people you've worked with, plus one place you publish weekly | Nobody buys judgement from a stranger's ad. They buy it from someone they've watched think about their problem, which takes weeks, not clicks. | Your time, twice a week | No inbound conversation at all after six weeks of publishing and forty personal messages |
| A physical product you'd have to make in a batch | Pre-orders to one community you're already in | Physical-product waitlists convert under 5% on wide release, so the list has to be large or the community tight. Money up front also funds the batch. | $0–$200 | Under 20 pre-orders, or under 2% of the community you asked |
| A shop, café or anywhere people walk in | Google Business Profile, reviews, and the pavement outside | Someone searching for a café near them has already decided to buy; an ad interrupts someone who hasn't. And running ads to a place with two reviews spends money teaching people to distrust it. | $0, plus a week of asking | Under 10 reviews after four weeks of asking every customer |
| Software or anything online with no geography | One community where your users already argue about the problem | You have no radius to buy and no list to write to. What you have is a place where the problem is discussed, and a reason to be useful in it every day. | An hour a day | Under 30 sign-ups, or under 3 paid, after four weeks of daily presence |
Two things that look like channels and aren't, at this stage. Search is real and it's the best long-run channel most small businesses will ever have, but it pays out in months and your question is what happens in four weeks — start it in the background, don't count on it. And “posting on social media” is an outcome, not a channel: posting into an audience of eleven people isn't distribution, it's practice. Post anyway, because the account has to start somewhere and it makes you findable when someone checks you out. Just don't put it in the column where you're expecting customers.
The mechanics of each of these live elsewhere so this page can stay about the choice: pulling a real list of local businesses out of Google Maps and qualifying it, the DNS records and send caps that decide whether cold email lands at all, and how much your cash can actually stand on ads.
The numbers to expect at each step
Every channel is the same four steps — leads, replies, conversations, paid — and the useful thing is knowing roughly what each step should hand to the next, because that's what turns a bad week into a diagnosis. Here is what a real unit of effort becomes in the three channels most first-ten founders end up in.
| Step | The people you can name | Local paid social | Cold email to local businesses |
|---|---|---|---|
| One unit of effort | 40 names, messaged individually over two weeks | $300 over three weeks in a five-mile radius | 200 emails over four weeks, 40 a day, four follow-ups |
| Leads — people who see it | 40 messages, near enough all delivered | About 200 visits. Meta's average cost per click on a traffic campaign is $0.70; a tight local radius runs above that, so budget $1.00–$1.50 a visit | About 195 delivered — a verified list bounces around 1.5%, an unverified one 2.6% |
| Replies — anyone who responds at all | 20–25. They know your name, so half is normal and a quarter is a sign the offer is confusing | 12–15 form starts, on the 6.6% median across 41,000 landing pages. Under about 2% means the visitors were wrong, not the page | 20–30 if the list is under 200 and genuinely well-chosen (15–20% reply). About 7 if you built a big list instead (3.4–3.7% average) |
| Conversations — a real back and forth | 10–12 | 6–8. The rest are price-checking or out of area | 8–12. Between a quarter and a half of replies are actually interested; the rest are “not now” and unsubscribes |
| Paid — a card actually charged | 3–4 | 3–5, so roughly 2% of visits. Northline's four cost $75 each | 2–4 from 200. Top campaigns book 2–3 meetings per 100 emails |
| So ten customers is | All forty names, and then you have no more names | About $750 at that $75 — but $75 is this example's rate, not a benchmark. Home services pay $47 to $149 for a lead alone, before anyone books | About 600–1,000 emails and five to eight weeks at 40 a day |
Three of those numbers are worth stopping on.
Small lists beat big ones, badly. Across 53.1 million cold emails in 60,000 sequences, campaigns aimed at fewer than 200 prospects replied at 15–20%, against 8% for lists of 500–1,000 and a 3.7% average overall. That is the opposite of the instinct, and it's the single best piece of news for someone with no sales team: your disadvantage — you can only find forty relevant businesses — is the thing that makes the numbers work. If you're getting 3.7%, the fix is almost never the writing. It's that you built a big list instead of a good one.
The warm list is finite and that's the point. Forty names produces three or four customers and then it's empty, which is exactly why it can't be your channel. Use it in week one — it gets you paying customers, testimonials and the objections you'll need for the ad copy — and then treat it as spent.
Speed is worth more than any wording change you'll make this month. In home-services data compiled from Invoca and ServiceTitan, 78% of homeowners hire the first contractor who responds, a five-minute response converts many times better than a thirty-minute one, and 85% of callers won't leave a voicemail. Before you rewrite the ad, answer faster.
Broken, or just small?
This is the question that costs founders their first quarter. Forty emails, no replies — is the message wrong, or have you simply not sent enough emails for a reply to be likely yet? Most people answer it by feel, rewrite everything, and destroy the only comparison they had.
It's answerable with arithmetic. If a step converts at 4%, the chance of seeing zero from 25 tries is 0.96 to the power of 25 — about 36%. So a zero at that volume happens by luck alone more than a third of the time, and it means nothing. At 100 tries the chance of zero drops to 1.7%: now a zero is evidence. The table below is that calculation for the four steps, at the rates in the section above.
| Step, and the rate to expect | Zero here is ordinary — happens ~1 time in 3 by luck | Zero here is a verdict — under 1 time in 30 | If it really is broken, change this first |
|---|---|---|---|
| Cold email → any reply (4% on a mixed list) | 25 emails sent | 100 emails sent | The list, then the first line. Never the template first — a good email to the wrong forty businesses is still zero. |
| Ad visit → form start (6.6% median) | 15 visits | 60 visits | The audience and the offer on the page above the fold. A page that converts under 2% is usually getting the wrong visitors. |
| Ad visit → paid booking (about 2%) | 50 visits | 200 visits | The checkout, on a phone, on mobile data — then the price, then the risk: add a date, a named service area, a refund promise. |
| Conversation → paid (about 25%) | 4 conversations | 12 conversations | The price and the ask. Most founders never actually ask; they describe, and wait to be asked back. |
Two consequences of that table are worth acting on directly. The first is that you cannot learn anything from a week in most of these channels — 40 visits and 55 emails is inside the noise on every row — so committing to four weeks isn't discipline for its own sake, it's the minimum volume at which the experiment has an answer. The second is that when a step is genuinely broken, it's nearly always the earliest broken step that matters: fixing the checkout when the ads are reaching the wrong town changes nothing, and you will have spent a week doing it.
So read the funnel from the top and stop at the first step that is a verdict rather than noise. And when nothing is a verdict, the answer is not to change anything. It's to get the volume up.
- The step that's broken most often
- Leads → replies, and the cause is usually the list or the audience rather than the words. Forty businesses you chose one by one out-reply four hundred you exported.
- The step founders blame most often
- The copy. It's the one thing you can change without leaving the house, which is exactly why it gets rewritten while the real problem sits upstream of it.
- The step nobody checks
- Paid — specifically the checkout, on a phone, on mobile data, from the link as a customer actually receives it. Pay yourself and refund it. About half of “my ads don't convert” is a broken mobile checkout nobody has tried.
- The step that fixes itself with speed
- Replies → conversations. Answering in five minutes instead of five hours moves this more than any rewrite, and it costs nothing but a notification you actually respond to.
The four weeks, in order
Every step here is something you could do this afternoon. The order matters more than the effort: the page has to take money before the traffic arrives, and the kill number has to be written down before you're emotionally invested in the answer.
Before week one · twenty names and one page
Write down twenty people you could message today without it being strange. Not personas — names, with a way to reach each one. Then put up one page on its own address with the offer, the price and a working checkout, and pay yourself on a phone on mobile data to prove it works. Refund it.
Then write the kill number down somewhere you'll see it: what result after four weeks would make you change channel. For a local service that's usually “fewer than three paid bookings from 200 visits”. Choosing it now costs nothing; choosing it in week three is negotiation with yourself.
Week one · spend the warm list, write the ads
Message all twenty individually — not a broadcast, and not a paragraph. What you're doing, what it costs, and a direct ask: would you like one, or do you know someone who would. Expect about twelve replies and three or four customers. Book them in.
In the same week, set up the channel but don't launch it: write four ads on four different angles (price, speed, convenience, the thing your competitors are bad at), or build the list of forty businesses one at a time. Launching on Monday with everything ready beats launching on Wednesday with half of it.
Weeks two to four · run one channel, change one thing a week
Launch, and then leave it alone for seven days. At the end of each week look at the four numbers and change at most one thing — switch off the two weakest ad angles, or change who the emails go to. One change a week means you can still attribute the difference; three means you can't.
Answer every enquiry within five minutes during your working day and within the hour otherwise. Put the notification somewhere you can't miss it. This is the highest-return habit in the whole four weeks and it needs no budget.
Every Friday · four numbers, in a notes app
Leads, replies, conversations, paid. Ten seconds to write, and they're the only way to tell a broken step from a small one. Add the spend and you can compute cost per customer, which is the sentence the whole exercise exists to produce.
If a step looks bad, check it against the noise table above before you touch anything. Most week-one and week-two disappointments are volume, not quality.
At ten customers · ask, fix, raise
Ask all ten the same question: what nearly stopped you buying? Not “how did we do” — the objection, specifically. The answer that comes up more than twice is worth more than the next month of ads, and it's usually free to fix. Northline's was “does he even come to my street?”, and the fix was a service-area map on the page.
Then put the price up on new work. The first price is almost always too low, you now have ten people's worth of evidence about why, and a rise on new bookings is reversible in a way that almost nothing else is. Then decide whether to keep the channel or change it — against what a month actually costs you.
Four weeks, counted
Northline Bike Repair is the example business that runs through these pages: one founder, a van, mobile bike repair in Denver, $60 for a 40-minute tune-up in your driveway. Below is the four-week channel test exactly as it ran, days 31 to 58. These are this example's own figures, not benchmarks — what's worth copying is that every cell was written down on a Friday.
| Week | What ran | Leads | Replies | Conversations | Paid | Spent |
|---|---|---|---|---|---|---|
| Days 31–37 | Twenty individual messages to people he could name. Four ad angles written and published paused. | 20 sent | 12 | 7 | 3 | $0 |
| Days 38–44 | Ads live, four angles, five-mile radius, $14 a day. | 71 visits | 9 booking starts | 3 | 1 | $100 |
| Days 45–51 | Two weakest angles switched off. Nothing else changed. | 88 visits | 12 booking starts | 5 | 2 | $100 |
| Days 52–58 | Budget moved onto the surviving angle. Service-area line added to the page. | 55 visits | 7 booking starts | 3 | 1 | $100 |
| Four weeks | One channel, one page, one change a week. | 214 visits + 20 messages | 40 | 18 | 7 | $300 |
Four of those seven came from the ads, at $75 a customer. Then the part that doesn't show up in any funnel: two of the four ad customers referred a neighbour within a fortnight, and one of the people who'd paid a $20 deposit back in week two booked again. Counting everyone since day one, customer number ten arrived on day 47.
Read the middle two rows against the noise table and something useful appears. Week one of the ads produced 71 visits and one booking — at a 2% expected rate, a single booking from 71 visits is completely ordinary, and so would zero have been. There was nothing to learn and nothing to fix, and the discipline that week was doing nothing. It took the full 214 visits before 4 paid bookings meant anything at all, and what it meant was: $75 a customer, against an average order worth about $186. That's the sentence the four weeks were for.
By the end of it the site was live and taking real cards, $1,286 had been collected that month, the $300 of ads had brought back $1,486 counted from Stripe charges rather than from what Meta reported, and the Money page showed 8.4 months of runway. Nothing about the channel was clever. It was one channel, four weeks, and a Friday habit.
After ten
Ten customers is not a business yet — it's a working measurement. What you have that you didn't have five weeks ago is a cost per customer, a conversion rate at every step, one objection that came up more than twice, and a price you now know was too low. That's enough to make the next decision with evidence instead of hope, which is the only thing the exercise was ever for.
Three moves, in this order. Do more of the thing that worked — the same channel, more budget or more hours, before you touch anything else, because a channel you understand at $300 is worth more than two you don't at $150 each. Fix the objection, which is usually a sentence on the page. Then, and only then, add the second channel, and run it the same way: four weeks, a kill number written down first, four numbers on a Friday.
And if the four weeks failed — under three customers from a channel you ran properly — that is a result about the channel, not about the business. A bookkeeper who sends 200 hand-picked emails and gets four replies and no clients has found out that the small businesses in her town don't buy bookkeeping from a stranger's inbox. That cost four weeks and a domain, and it points straight at the two channels left: the accountants who already have those clients, and the people she's worked with. Change the channel; don't add to it. The order the rest of it goes in is on the ninety-day sequence, and if you're not yet sure the offer itself holds, that question comes first: validate it with money before you buy traffic for it.
Common questions
How long should the first ten customers take?
Six to ten weeks for a local service run properly — three or four from your own list in the first fortnight, the rest from one channel over four to six weeks. Longer for anything sold to businesses, where a single decision can take a month on its own. If it's taking more than three months, the problem is usually that three channels are being run at a third of the effort each.
Should I discount to get the first customers?
Offer something other than money if you can — a faster slot, extra work included, a guarantee. A discount teaches the first ten what your price is and they're the people most likely to refer, so the discount travels. If you must, make it visibly temporary and tied to something real: a launch week, a founding-customer place, ten slots.
How many people should I message on my warm list?
All of them, individually, over about two weeks — twenty is a normal list and forty is a good one. Individually is the whole thing: a broadcast to twenty gets you two replies, twenty separate messages gets you a dozen. Expect around half to reply and three or four to buy, and expect the list to be spent afterwards.
What conversion rate should I expect from my landing page?
The median across 41,000 landing pages and 464 million visits is 6.6% for a form fill, and professional services sit at 6.1%. A page asking for a card rather than an email converts far lower — around 2% of visits ending in a paid booking is a healthy number for a local service. Under 2% of visitors doing anything at all usually means the traffic is wrong, not the page.
How many cold emails do I need to send to get a customer?
On a small, hand-picked list of under 200 local businesses, reply rates run 15–20%, and something like 2–4 customers out of 200 emails is a realistic outcome. On a big exported list you'll see the 3.4–3.7% average and need three times the volume for the same result. Ten customers is roughly 600 to 1,000 emails either way — five to eight weeks at 40 a day.
My ads got clicks but nobody bought. What's wrong?
Check in this order: the checkout on a phone on mobile data (pay yourself, then refund it), then whether the visitors were the right people, then the price, then the risk — a named service area, a date, a refund promise. And check the volume first: 50 visits with no booking happens about a third of the time even when everything works. At 200 visits with none, something is genuinely broken.
Do I need a website before I start?
You need one page that takes a payment. Not five pages, not a brand. For the warm list you can get by with a payment link and a phone number, but the moment you send strangers anywhere, they need somewhere to land that says what it costs and lets them pay without speaking to you.
When should I add a second channel?
When the first one has produced ten customers and you can say what one costs. Adding a second before that halves the volume going into a channel that hasn't reached the point of being readable yet, and you end up with two experiments that both sit inside the noise. Do more of what worked first — that's usually the highest-return move available.
One channel, run for you, counted from real payments.
Velofound puts the site live taking card payments, builds the lead list, writes the four ads and publishes them paused for your approval, then tells you what came back from Stripe charges rather than what the ad platform claims. Free to start, no card.
Start free →