Find the people who need it now.

Max Kolysh, a YC Visiting Partner, asked dozens of YC founders how they found their first ten customers. The first two or three came, “almost without exception,” from people the founders already knew or were one introduction away from. His reason: an early buyer is betting on a new company, so they need to trust the founder, and the people who already know you trust you most. ¹

The founder stories Lenny Rachitsky collected say the same. Gusto’s first ten “came from friends we knew who were just starting their businesses in California.” Amplitude’s first few were “connections through investors and other folks at YC.” ⁶ Prospecting tools, Kolysh says, “only start to matter once you have 10 to 20 quality customers.” ¹ Work through your network in this order:

1. People who know your work: former colleagues and managers, classmates, friends in the industry.
2. Your investors and the companies they’ve backed, and your accelerator batch.
3. Second-degree LinkedIn connections. Look at who your connections know and ask for introductions; one YC founder got introductions to about half the customers she closed during her batch this way. ¹
4. Tools that search your extended network in plain words, such as Happenstance, a YC company. ¹

Make each introduction easy to say yes to. Kolysh’s rule: tell the connector who you want to meet, why that person would care, and what to say in the forwarded email. ¹ Kazanjy suggests a subject like “Intro to <target contact’s name>?” and a note written so they can just hit forward. ⁹ Here’s the pair I’d send:

Pete Koomen, a YC General Partner and Optimizely’s co-founder, says to begin with a sales hypothesis: “customer X has problem Y, and our product will help them solve it.” A good one “makes prospecting easy by clarifying who you should be talking to.” ⁴ Optimizely’s was that marketers at small and medium tech, media and e-commerce companies wanted to run A/B tests but couldn’t without writing code.

So choose on urgency, not market size. Paul Graham’s question for startup ideas works for lists too: “who wants this right now?” ¹¹ YC’s partners put it in numbers: “recruiting 10 customers who have a burning problem is much better than 1000 customers who have a passing annoyance.” ¹⁰

Watch for the trap Koomen calls out: talking to whoever will take the call. It “selects for the people who are easiest to talk to, not the people who will be great customers.” ⁴ Other startups are the classic case. Gustaf Alströmer, a YC General Partner, calls them “the easiest category” to sell to ¹², while other partners warn that “just because you get someone early doesn’t mean they’ll stick.” ¹³ My test either way: does this group have the problem today?

Tom Blomfield, a YC General Partner and Monzo’s co-founder, tells founders to “disqualify customers who just aren’t ready, able, or willing to buy your product.” ¹⁴ Most of that can happen before the first call. Keep a group only if:

• They have the problem today, not someday.
• Fixing it is someone’s job, and that person has “the budget and decision-making authority to buy your product,” in Koomen’s words. ⁴
• They’re already trying to fix it with a spreadsheet, a hire or a tool they complain about.
• You can spot them from the outside, by a sign you can search for.
• You could find 50 of them this week.

A narrow group is small enough to list in full, and more of the people on it need you now. Written down, it’s your ICP, short for ideal customer profile: the kind of company, and the person in it, who need you most. Build it from things you can see from outside, so each step is a filter you can search on.

The best profiles combine size, team and a live sign of need. Pete Kazanjy, who founded TalentBin (later bought by Monster), describes a strong account as “a ~100-person organization with three recruiters and twenty open engineering requisitions.” ¹⁶ Christina Gilbert, a YC Visiting Partner and OneSchema co-founder, shows headcount working as a sign with compensation-planning software as her example: companies under about 100 people aren’t setting pay levels yet, companies over about 500 already have pay bands, so a company crossing 100 is the one to call. ³

Then choose the person. Write to the one whose job the problem is, not the one who’s easiest to reach.

Titles are where lists most often go wrong. Gilbert describes a founder selling infrastructure meant for senior engineering leaders whose list was full of developer-relations, go-to-market and customer success managers; hundreds of messages got no replies. ³ To find the right title, look at your deals: who pushed the last one to close, who signed, who paid. With no customers yet, ask who has the most to gain. Her example: a chief security officer could be fired over a breach, so device-monitoring software matters to them; an engineering manager would rather not be monitored.

Gilbert calls these intent signals: clues that a company needs what you make, like a job post or fast growth. ³ Kazanjy builds his whole list from them, “a set of characteristics that indicate potential demand for your solution.” ¹⁶ My rule: nobody gets a first message without a sign, the date you saw it and the link.

YC companies built their first cold lists from signs like these:

• Retool filtered Crunchbase by “when was the last time the company fundraised and what verticals they’re in,” kept operations-heavy startups, skipped SaaS companies because they “don’t build that many internal tools,” and emailed “the CTO and VP of operations.” ⁶
• Vanta “emailed anyone who’d ever said the word ‘compliance’ on Bookface in the prior decade.” Bookface is YC’s internal forum. ⁶
• Optimizely used BuiltWith to see whether prospects ran “analytics tools and JavaScript frameworks,” signs that a company cared about its website. ⁴

Kolysh’s first question, before any tool or channel: “Where does your target customer actually spend their time?” A sales leader lives in LinkedIn and email. A truck dispatcher, a property manager or an insurance agent mostly doesn’t. One YC founder spent months on email and LinkedIn selling into a legacy industry, then went to a trade show and “closed more in three days than he had in three months of sending cold emails.” ¹

For buyers at a desk, two tools do most of the work. Apollo “was the most common starting point” for the founders Kolysh asked ¹, and Koomen says many in YC’s Winter 2024 batch used Apollo and LinkedIn Sales Navigator. ⁴ Apollo is a database of people and companies that also finds emails; Sales Navigator is LinkedIn’s paid search, with no emails. Two searches to copy:

• In Apollo: job titles “VP Sales” and “Head of Sales,” seniority VP, 51 to 200 employees, United States, plus a tool your buyers use, such as Salesforce. Kazanjy’s similar search (Salesforce users with 100 to 200 staff in the San Francisco Bay Area) found about 300 accounts. ¹⁶ The free plan covers basic filters; advanced filters and signals need a paid plan. ²⁰
• In Sales Navigator: paste ("VP Sales" OR "Head of Sales" OR "VP of Sales") NOT (assistant OR intern) into the search bar, add headcount 51–200 and your region, then narrow to people who changed jobs in the last 90 days or posted in the last 30, or to companies with open jobs. ¹⁷

That’s Boolean search: exact phrases in quotes, AND, OR and NOT in capitals, parentheses to group, and no * wildcards, which LinkedIn doesn’t support. ²¹ A search shows at most 2,500 people, so split big ones by region or size. ²² Save each search, and LinkedIn emails you new matches every week. ²³

Search widely, but reveal contact details only for people you’ll actually write to: in Apollo an email costs 1 credit and a phone number 8. ²⁰ Check every company by hand before it goes on the list. Gilbert’s warning: search Sales Navigator for tech companies and you’ll often get IT consultancies. ³

A bounce is an email that can’t be delivered, and too many of them hurt your sending reputation for everything after. Amazon’s email service tells its senders, “For best results, you should maintain a bounce rate below 2%”; it reviews accounts at 5% and may pause sending at 10%. ⁷ Hunter’s data from its own users puts the average bounce rate at 3.6%, mostly from unverified contact details, so the average sender in that data is already over the line. ²⁶ My rule: nothing goes out unverified.

1. 1

   Start with your database.

   Reveal the email in Apollo, or whichever database you searched. Some will come back empty or unverified.

2. 2

   Fall through to finders that charge only for results.

   Run the empty ones through Hunter, Findymail or FullEnrich, in turn. Each charges only when it finds an address, so trying a second finder doesn’t double your bill.

3. 3

   Verify every address.

   Put the whole list through a verifier such as MillionVerifier or ZeroBounce, including addresses a finder says it checked. Mark each one verified, risky or dropped.

4. 4

   Set catch-alls aside.

   A catch-all domain accepts mail for any address, so no checker can confirm the person exists. ZeroBounce says these can’t be validated “without sending a real email and waiting for a bounce,” and suggests a separate group, knowing some will bounce. ²⁷ Send them last, in a small batch, and stop if bounces climb.

5. 5

   Record where each one came from.

   Note the tool, the date you checked and the result. You’ll need the source for the EU rules below, and it shows you which tool’s data bounces.

Guessing is fine if you check. Aaron Epstein, a YC General Partner, points out that if everyone at a company is first name dot last name, you can probably guess anyone’s address there. ² Run guesses through the verifier like everything else.

A CRM (customer relationship management tool) is where you track every person, conversation and next step. You don’t need one on day one. Gustaf Alströmer, who helped start Airbnb’s growth team, says to start simple, with columns for “industry, the company, the title, the name, the email, maybe the LinkedIn,” then move to “a simple sales CRM software that tracks these conversion rates automatically.” ¹² Kazanjy says the same: “just use a Google Sheet as your initial repository of prospects.” ¹⁶

Add these columns too. Each answers a question you’ll ask later:

Give every row a stage, and every stage a next step. Kazanjy wants stages with “crisp definitions,” and “At each stage, there should be an explicit next step.” ²⁹ These are the stages I’d use:

1. To research
2. Ready: email found and verified
3. Contacted
4. Replied
5. Call booked
6. Qualified, or not a fit (write down why)
7. Trial or proposal
8. Won or lost

Kazanjy himself is “split on the necessity of using a full-blown CRM” for a sole founder. ⁹ I’d move when two of you work the same list, or when you can’t see every lead’s next step at a glance. Attio and HubSpot both have free plans (see the tools below).

Work backward from what you need. Alströmer’s rule: “you cannot close five customers from 10 leads.” His example funnel starts with 500 messages and ends with about 10 calls and 2 customers. It’s an illustration, not a benchmark, but it shows the scale. ¹²

For a first test, Kazanjy aims for “a targeted list of 50–100 prospect accounts.” ¹⁶ With one to three people at each, that’s roughly 100 to 200 people per group. Run two to four groups, each a different kind of company or a different sign, and keep them in separate lists so you can compare them. Small and specific tends to win: in Hunter’s data from its users’ 2025 emails, sequences sent to 21 to 50 people averaged 6.2% replies, against 2.4% for 500 or more. ²⁶

Be honest about what 100 people can tell you. Replies are rare, so small numbers swing a lot:

So read only big gaps: 8 replies against 1 is probably real, 4 against 3 is noise. Zero is different. If your real rate were 3%, the chance of no replies at all from 200 well-targeted, delivered emails is about 0.2%, so zero from 150 to 200 means something is wrong.

Count positive replies and calls booked, never opens: Apple Mail loads tracking pixels on its own, which is why Smartlead’s benchmark of its customers’ emails leaves open rates out. Vendors’ published reply-rate averages run from 0.74% (Smartlead’s median sender) to 4.5% (Hunter’s users), so your own groups are the only benchmark that counts. ^{30 26} Then let the replies rewrite your target:

1. Tag every reply: interested, not now, wrong person (ask who is right), no, unsubscribe, bounced.
2. Compare the groups. Then compare the interested people with each other: title, company size, sign, place.
3. Ask everyone who books a call what made them answer. Gilbert asks “what got your attention about my message” and reuses their words, because “Your customers are always going to write better copy than you do.” ³
4. Write down every objection, word for word.
5. Rewrite your one-line target, drop the weakest group, add one new group, and run the next 100.

Those rules turn into a few columns. Keep them filled in for every row:

• Country.
• Where you found the contact, and when.
• Why you can write: for example, “published work address, relevant to their role.”
• Opt-out date, plus a do-not-contact list you check before every send.

Block the time first. Gilbert’s fix for founders who never get to outbound is a plan on the calendar: how many messages a day, how long each one takes, and exactly when you’ll send them. ³

1. 1

   Day 1. Write the sentence and pick your groups.

   Write “[who] has [problem], and we help them [result].” List two to four groups that might fit and rank them by who needs it now. For each, note the title that owns the problem, the sign you’ll look for and where those people spend their day.

2. 2

   Day 2. Work your warm list.

   Go through former colleagues, friends in the industry, investors, your accelerator batch and second-degree LinkedIn connections. Put everyone who fits in the sheet with their warm path, and send intro asks with a note they can forward.

3. 3

   Day 3. Build the cold list from signs.

   Take your top two groups. Search Apollo or Sales Navigator by title, size, place and sign, check each company’s website by hand, and stop at 50 companies a group with one to three people each. Note each person’s reason to write now, with the link and the date, wherever you find one.

4. 4

   Day 4. Find and check emails.

   Run the finders in turn, verify every address, set catch-alls aside and drop the ones that fail. Fill in country, source and stage, then grade each lead: A has a clear sign and the right title, B fits but shows no sign yet (hold it until one appears), C is a maybe.

5. 5

   Day 5. Write your first 20 messages by hand.

   Start with your A leads. Keep each one under 75 words, about them, with one small ask such as a 15-minute call, and read it out loud before you send it from your own account. ¹ Log every send. The LinkedIn, email and X guides cover each channel.

6. 6

   Week 2. Follow up, read, adjust.

   Follow up three or four times over a couple of weeks. ¹ Write the next 80 messages the same way and tag every reply. When a group passes 100, compare it with the others, rewrite your sentence, drop the weakest group and add a new one.

Before anything goes out, check each row:

• There’s a reason to write now, with a link.
• Their title owns the problem.
• The email is verified, or marked risky and held back.
• Country and source are filled in.
• The message is under 75 words, with one ask.
• You’d be glad to get it. Koomen’s rule: “only send emails that you yourself would be excited to read.” ⁴