Getting your first ten customers
The first customers come from doing things that do not scale: a list of named people, personal outreach, conversations with every user, and asking for the money. You are finding out whether anyone pays before you build a machine.
37 steps, about 42 hours of work · 6 plays · 15 checked claims
The first ten customers of a SaaS come from work that does not scale: a list of a hundred named people, personal outreach to the warm ones first, cold messages that reference something real, and a conversation with every person who signs up. Marketing channels come after, once you know what made those ten buy.
The concept, explained in full: ICP.
Do it
Every step in the free launch plan that this topic needs, in the order you need them. Each keeps its plan number.
- Who are the first hundred people to contact?8 steps
- How do you reach them?12 steps
- What do you learn from the first users?5 steps
- How do you make the first session succeed?7 steps
- How do you turn them into proof?5 steps
Who are the first hundred people to contact?
Named humans, not a segment. Your network, the communities your buyer posts in, and unhappy competitor users.
2.1.1List 25 people from your own network
The list you are embarrassed to use is the list that converts first.
Done when: List the 25 names, each with one line on how you know them. Twenty-five distinct people, none of them a company.
2.1.2List 25 from communities where your ICP posts
Pull them from people who described your problem in their own words in the last 90 days.
1h
2.1.3List 25 from competitor reviews
G2 and Capterra 2- and 3-star reviews are a list of people actively unhappy with the incumbent.
1h
2.1.4List 25 with a live trigger event
Hiring for the role, a new person in the seat that buys, a funding round, a migration announcement, a public complaint. Free places to watch: the Greenhouse, Lever and Ashby job feeds, SEC Form D filings, and the ad libraries. If what you sell replaces the role, one still open after two months is the strongest of all. A trigger from a year ago is not a trigger — date every one and drop the stale.
2.1.5Qualify each one in a sentence
Problem, trigger, and whether you can actually reach them. Cut anyone failing the third.
Done when: Quote the one-sentence qualification for any five people on the list. Each sentence names a problem that person has, not only a job title.
1h
2.1.6List everyone who ever signed up to anything you built before
Dead waitlists, a shut-down side project, the product that did not work. They already trusted you once, which is the expensive half, and nobody else can email them.
45m
2.1.7List 25 who commented on a competitor founder’s posts this month
Pick the three accounts that already own your audience and read their comment sections: people mid-comparison, in public, for free. The signal decides who you write to — never what you write. “Saw you liked their post” reads as surveillance.
1h
2.1.8List 25 from the rosters your buyers’ industry already publishesB2B
Association member lists, state licence lookups, or a Google Maps search by category and postcode, where the review count is a free size estimate. A licensed business is real and operating, and almost nobody else is writing to it from that list. Maps gives the phone number too: an owner who runs the business from a phone answers a call, not an email. In the UK a sole trader counts as an individual, who must agree before you email them, so call or write instead.
How do you reach them?
Warm contacts first, then specific cold outreach, then the communities they already use.
2.2.1.1Email 25 warm contacts personally
No template, no BCC. Ask for a reaction, not a purchase.
2h
2.2.1.2Ask each for two introductions
Name the two people you want. “Anyone come to mind?” gets nothing.
2.2.2.1Write a four-email sequence
Trigger → one-line pitch → proof → break-up. Under 75 words each. Every email to US recipients carries your postal address and a plain way to opt out: CAN-SPAM makes no exception for business email.
2.2.2.2Send 25 a day for five days
Slow and warmed. A cold domain that blasts 250 in one morning is a dead domain.
2.2.2.3DM 25 on the platform they actually use
Reference something they posted. Anything else reads as automation, because it is. On LinkedIn a free account can add a note to only a few invitations a month, so connect without a note and send the specific message once they accept.
2.2.2.4Answer every reply the same day, and give the inbox one owner
A reply is a person who is interested right now, and that decays by the hour. The same replies produce different revenue depending only on whether someone answered in minutes or tomorrow.
2.2.2.5Verify every address before the first send goes out
A guessed or scraped list carries dead addresses, and a high bounce rate on a new domain is read as spam by the people you are trying to reach. Drop the unverifiable ones rather than hoping — the domain you are protecting is the one your invoices come from.
2.2.3.1Join the five communities your ICP lives in
Read the rules page of each first: traffic is not permission. The big ones route every product mention to one weekly thread and remove the rest, so check where promotion is allowed before you pick the five. Then rank by activity, not size — comments per thousand members on last month’s top posts. A five-million-member hobby sub can be quieter than a niche one, and a big group can be one admin posting to nobody.
30mhow to do it
2.2.3.2Answer ten real questions with no link
This buys the standing that makes the eleventh post, with a link, acceptable.
Done when: List the URLs of ten answers you posted in communities. Ten distinct URLs, none of them containing a link to your own product.
2.2.3.3Give away one genuinely useful resource
No signup wall, no email gate. The goodwill is the point.
2.2.3.4Set alerts for the request phrasings and answer them the same day
Someone writing “is there an app that…” is a buyer describing the need out loud. Where a platform’s own search is useless, ask Google: site:facebook.com/groups or site:reddit.com plus intitle:"looking for" and your category, filtered to the past week. Answer with the real options including rivals, and disclose that one is yours.
45mhow to do it
2.2.4Count the five outbound numbers, not just the reply rate
Replies, then the ones that are actually positive, then meetings booked, then meetings that happened, then closes. Each rung fails for its own reason, and a 10% reply rate made of polite noes is a copy problem wearing a success. Split it by list segment too — one or two segments usually carry everything, and the average hides which.
Done when: Quote the five outbound numbers for the last send: sent, delivered, replied, positive, booked. All five present, for the same period.
45m
What do you learn from the first users?
Every call and every cancellation tells you what to fix and what to write on the homepage.
2.3.1Offer every signup a 15-minute call
Maybe one in five accepts. Those five are worth more than a hundred analytics sessions.
2.3.2Run ten problem interviews and record them
Ask what they did before you, and what it cost. Never pitch during one.
Done when: List the ten people interviewed: first name or role, the date, and the one thing each did before your product. Ten distinct entries; a sales call does not count.
2.3.3Interview every user who leaves
At ten customers, one churn interview outweighs any dashboard you could build.
1h
2.3.4Write the top-five objections list
Verbatim, in their words. This document becomes your homepage, FAQ and ad copy.
Done when: Quote the five objections in the words the customers used. Five entries, each attributed to a named interview.
45m
2.3.5Fix the number one objection
In the product or in the copy. Fixing it in copy is legitimate and much faster.
4h
How do you make the first session succeed?
A signup that never reaches the activation moment is a lost customer you already paid for.
2.4.1Instrument time-to-activation
Median minutes from signup to the activation event. Then halve it.
Done when: Report the median minutes from signup to the activation event across the last 20 signups, and how many of the 20 could be measured.
45m
2.4.2Delete one step from signup
Every removed field is measurable conversion. Start with the one you “might need later”.
Done when: Quote the signup steps before and after the cut. The after list is one step shorter than the before list.
1h
2.4.3Make the empty state do the work
Sample data, a template, or a one-click demo. Never a blank canvas and good luck.
2h
2.4.4Send a three-email first-week sequence
Day 0 get started, day 2 the one feature that matters, day 5 ask what got in the way.
2h
2.4.5Personally message everyone who stalls
Signed up, never activated. One sentence, from you, asking what stopped them.
30m
2.4.6Put one real result in front of them before the priceB2C
Not a tour of the feature — one actual output, on their own input. The ask lands differently once they have already seen the thing work.
3h
2.4.7Treat the first-run sequence as a permanent test surfaceB2C
Superwall’s case study of Cal AI reports 424 variants across 123 experiments, 160 of them paywall designs, for a 31% relative lift in trial-to-paid over 12 months. It is a growth engine, not a screen you finish.
How do you turn them into proof?
Testimonials, a case study and future reviewers, collected while you still know everyone by name.
2.6.1Ask your three happiest users for a testimonial
Send them a draft to edit. “Write whatever you like” gets you nothing for three weeks.
Done when: List the testimonials received: who said it, their role, and the sentence. Three entries, each with a named person; an anonymous quote does not count.
30m
2.6.2Put names or logos on the homepage
Ask permission first. Three real names beat a wall of unfamiliar logos.
30m
2.6.3Write case study number one
One customer, one before number, one after number, one quote. Two pages maximum.
Done when: Open the published case study. It names the customer, one number from before and one from after. The URL is the evidence.
2.6.4Record a two-minute customer video
A shaky call recording with real specifics beats a produced testimonial with none.
1h
2.6.5Line up 20 users willing to review you later
G2’s own figure for review requests is an 11% conversion, so 20 asks will not land 10 reviews. Start with 20, keep adding names, and ask permission now to cash it in in Phase 2.
More steps on this topic
Ask for the money and close
- Ask for payment on the first call, not the third30m
A call that ends with “send me something” is a no. Name the price, then stop talking.
- Send a payment link within the hour of a yes30m
Intent decays by the hour. A yes that waits for an invoice turns into a maybe.
- Offer the hesitant a paid pilot, never a free one45m
A free pilot tests politeness. A paid one, even at $100, tests whether the problem hurts.
- Chase every open deal three times over two weeks45m
One follow-up is a nudge. Three that each carry something new, a result, a fix or a date, are a conversation. Stop at three and ask for a plain no.
- Ask every no for the reason and write it in one line30m
Ten no’s in the same words is your product spec. Ten different ones is a targeting problem.
- Count your paying customers against the ten you need15m
Ten different buyers, each charged in live mode at the real price. This is the phase exit.
- Write one sentence per customer on how they found you45m
The other half of the phase exit. If you cannot write it, you do not know which part of this phase to repeat.
Plays
Reusable patterns that explain how: backed by read sources first, then practitioner advice, then our own ideas.
- A specific customer win, framed as a lessonBacked · 1 source
Post one specific customer win — a real number, named outcome, and the before/after — framed as a lesson others can copy, with the customer’s permission. Concrete proof outperforms feature claims: it shows the result a buyer wants, makes your product the obvious path to it, and lets the prospect picture themselves there.
- When a prospect says "I will try it", set it up with them on the callPractitioner · 1 source
When an early prospect says they will try the product later, offer to set it up with them now: share a screen, use their real data, and stop when they see a first result. Interest cools within days, and a guided first result does what a trial link rarely does. Note what they stumble on; it is your onboarding backlog.
- Promote only in the designated promo threadsPractitioner · 1 source
Most large subreddits run designated promotion threads — the one surface where pitching is explicitly welcome. Inventory the recurring threads in your niche, show up every cycle with a tight pitch and a real update, and never promote outside them. The restraint itself builds the account trust that later makes your ordinary posts survive moderation.
- The disclosure-first founder answerPractitioner · 1 source
Open with the bias disclosure, then write the best answer in the thread — including when a competitor is the right pick. Disclosure converts skepticism into credibility: readers upvote founders who show their hand, and moderators leave disclosed expertise standing. Concealment is the pattern platform trust-scoring exists to catch; candor is the loophole that is also the rule.
- Drop a finished resource with no strings attachedPractitioner · 1 source
Share a genuinely finished resource — template pack, dataset, checklist, small tool — with zero gate: no email, no signup, hosted openly. Reddit rewards unconditional usefulness and punishes bait; the goodwill, profile visits, and "what else have you made?" comments convert later, on the community’s terms.
- Answer every "is there a tool that…?" thread in your nichePractitioner · 1 source
Monitor your niche’s communities for "is there a tool that…" and "looking for an app" threads, then answer fast and honestly: recommend genuine options including competitors, disclose your affiliation, and explain the fit. These threads are purchase intent in plain text, and they keep converting from search for years.
What the evidence says
Sources
- How to Get Your First 10 Customers
- Filing a Form D Notice
- Job Board API
- lever/postings-api (README)
- Electronic mail marketing (Guide to PECR)
- CAN-SPAM Act: A Compliance Guide for Business
- Google Workspace Admin Help: email sender guidelines
- Sender Best Practices
- Gmail sending limits in Google Workspace
- Invitation limit reached
- Connecting With Other Members - Best Practices
- The InsideSales.com/MIT Lead Response Management Study
- The Short Life of Online Sales Leads
- Strengthening Email Ecosystem: Outlook’s New Requirements for High-Volume Senders
- Reddiquette
- How to talk to users
- Startup School - How to talk to users
- What’s New, G2? June 2020
Founder questions
Where do the first SaaS customers come from?
Mostly from people the founder contacts personally: their own network, people who described the problem in a community, and unhappy users of a competitor. Ads and SEO are too slow and too expensive at this stage.
How many cold emails should a new domain send?
Start slowly, around 25 a day, on a domain with SPF, DKIM and DMARC set up. Sending hundreds on the first morning lands in spam and can burn the domain.
Should you talk to every early user?
Yes. Offer every signup a short call and interview everyone who leaves. At ten customers, one conversation tells you more than any dashboard.
Asked in public, and answered from the thread:
Related topics
Go further
Where it sits in the plan
- 2.1Build a list of 100 named humansin First 10 paying customers
- 2.2Do the outreach yourselfin First 10 paying customers
- 2.3Talk to every single userin First 10 paying customers
- 2.5Ask for the money and closein First 10 paying customers
- 2.6Harvest proof while you still know everyonein First 10 paying customers
Sources
- Paul Graham, How to Get Startup Ideas (2012)
- Clayton M. Christensen, Taddy Hall, Karen Dillon & David S. Duncan, Know Your Customers’ “Jobs to Be Done” (Harvard Business Review, 2016)
Compiled by The Marketinque Team to our editorial standards.