# How to Write a Cold Message That Gets Replies (3 Rules, 2-Min Video)

Canonical URL: https://pumpgtm.com/blog/cold-message-that-gets-replies-three-rules

[← All guides](/blog) Outbound Cold messaging

# How to write a cold message that gets replies: 3 rules from 10+ YC startups

Start with the problem, show why you're relevant, and never ask for a meeting in the first message. The two-minute version of the cold message rules we run for 10+ YC startups, with examples and the one mistake AI-written outreach keeps making.

Namanyay Goel · September 22, 2026 · 7 min read On this page
- [Rule 1: start with the problem](#rule-1-start-with-the-problem)
- [Rule 2: include evidence that you are relevant](#rule-2-include-evidence-that-you-are-relevant)
- [Rule 3: never ask for a 15-minute meeting in the first message](#rule-3-never-ask-for-a-15-minute-meeting-in-the-first-message)
- [Bonus: stop over-personalizing](#bonus-stop-over-personalizing)
- [Put the three together](#put-the-three-together)
- [Questions founders ask](#questions-founders-ask)

Most cold messages fail before the second sentence. Not because the product is bad, but because the message opens with the sender, quotes something the prospect posted last week, and closes with "do you have 15 minutes?" I read hundreds of these a week while running outbound for 10+ Y Combinator startups, and the ones that get replies all share the same three moves. This is the two-minute version.

## Rule 1: start with the problem

One line, about them, about a problem they are probably living with right now. Not your company, not your title, not a compliment.

If you sell to early-stage founders: "Are you trying to build your next MVP but find all the tools confusing?" If you sell to engineering teams: "Are PR reviews getting stuck between GitHub and Slack?" The test is whether the reader can answer yes or no in their head before they finish reading.

## Rule 2: include evidence that you are relevant

After the problem, one piece of proof that you have the right to talk about it. It does not have to be big. "We have 2,000 stars on GitHub." Your last job. Your college. A customer they would recognize. Anything concrete that answers the silent question "why are you the person telling me this?"

Skip this and the message reads as a pitch from a stranger. Include it and the same words read as a peer who has seen the problem up close.

## Rule 3: never ask for a 15-minute meeting in the first message

This is the biggest mistake I see. If a stranger walked up to you on the street and asked for 15 minutes of your time, you would feel a little creeped out. A cold message is the same thing.

Ask a question about the problem instead. If your startup helps companies cut AI inference spend: "Is AI spend something you're thinking about this quarter?" A yes starts a conversation. A no starts a conversation. A meeting request gets ignored.

This is also why PumpGTM's sequences end on a question rather than a calendar link, and why the reply drafts it writes pick up from whatever the prospect said. More on that in [how PumpGTM works](/blog/what-is-pumpgtm-how-it-works).

## Bonus: stop over-personalizing

The number one mistake in AI-generated cold messages is spending the opening on the prospect's job, career, or last LinkedIn post. Everyone can tell it was pasted in from ChatGPT, and everyone knows you did not actually read their profile. It wastes the few words you have before they stop reading.

Spend those words on the problem and your authority instead. If you want a checklist for catching this in messages an AI wrote for you, use the [anti-slop review list](/blog/review-ai-written-cold-messages).

## Put the three together

Problem, evidence, question. Three short lines. For an infra startup selling to CTOs:

"Are you trying to cut AI inference spend this quarter? We run inference for two YC companies at about a third of the cost they paid before. Is that on your list right now, or handled already?"

No introduction, no compliment, no meeting request. If they answer, you have a conversation, and the meeting comes out of that.

This is the message structure PumpGTM writes and sends for you, from your own LinkedIn, email and X accounts, then learns from which lines get replies. The rest of the system is in [using AI to run founder-led outbound for 10 YC companies](/blog/ai-founder-led-outbound-system) and the [4-step LinkedIn outbound playbook](/blog/linkedin-outbound-playbook-enterprise-meetings). Pricing is on the [pricing page](/pricing).

## Transcript

There's three simple components of a great cold message that I'm helping 10 plus Y Combinator startups do.

Number one, start with the problem. In one simple line, talk about the problem that the other person might be encountering. So if you've got a Y Combinator startup, something you can ask is, hey, are you trying to build your next MVP but find all the tools confusing?

Number two, always include some evidence, authority or proof of why you're relevant about this problem. Something simple, you can't say that, hey, I've got 2,000 stars on GitHub. Or you can talk about your college, your past job, anything to give some evidence why you're the right person to think and talk about this.

Number three, never ask for a 15-minute meeting at the end of that. That's one of the biggest mistakes I'm seeing everyone make. But if you ask someone for a meeting even in your first cold message, honestly, that's a little weird and creepy. Think about it this way. If a stranger came up to you and asked for a 15-minute meeting, wouldn't you feel a bit creeped out? So the better thing to do is to ask them a question about the problem they're having. So for example, if your startup helps companies save on AI inference spend, you can simply ask them, hey, is AI spend something you're thinking about this quarter? This is because if they say yes or no, you can at least start a conversation with them, versus asking for a 15-minute meeting and they just completely ignore you.

And this is a bonus tip. The number one mistake I'm seeing these AI generated cold messages have is over-focusing on personalization. Listen, man, talking about someone's job or their career or something they posted on LinkedIn or X. Everyone knows that you just copied and pasted it from ChatGPT and you didn't actually bother and care and check their profile. You stop lying and stop wasting crucial text. You have very little time to get their attention. So do it by talking about the problem and your authority in presenting the solution.

The reason I'm so confident with this advice is because I'm here in San Francisco and I'm helping 10 plus of the latest Y Combinator startups automate all of their GTM and actually get meetings and sales booked on the calendar. I'm Namanyay, the CEO of PumpGTM, and if you want to get the same playbook top startups in San Francisco are using, just check out the links in the description. Thank you.

## Questions founders ask

How long should a cold message be? Three short lines: the problem in one sentence, one piece of evidence that you are relevant, and one question about the problem. If it needs scrolling on a phone, it is too long.

Should I ask for a meeting in a cold message? No. Asking a stranger for 15 minutes in the first message is the single most common reason cold outreach gets ignored. Ask a yes-or-no question about the problem, and let the meeting come out of the conversation.

Is personalization bad in cold outreach? Personalization about the person (their post, their job change) reads as pasted from ChatGPT and wastes your opening. Personalization about their problem is what gets replies.

What counts as evidence or authority? Anything concrete that earns you the right to talk about the problem: GitHub stars, a recognizable customer, your last job, your college, a number from your own product.

Does this work for cold email as well as LinkedIn? Yes. The three rules are the same on LinkedIn, email and X. PumpGTM runs the same problem-first structure across all three channels from your own accounts.

How does PumpGTM use these rules? Every sequence it writes opens with the prospect's problem, includes your proof, and ends on a question instead of a calendar link. When someone replies, it drafts the next message for you to send in one click.

[Namanyay Goel](https://nmn.gl/) · Founder, PumpGTM

Founder of PumpGTM (YC). Writes about founder-led GTM, LinkedIn outbound, and the systems behind it.

[X](https://x.com/NamanyayG)[LinkedIn](https://www.linkedin.com/in/namanyayg/)

On this page

- [Rule 1: start with the problem](#rule-1-start-with-the-problem)
- [Rule 2: include evidence that you are relevant](#rule-2-include-evidence-that-you-are-relevant)
- [Rule 3: never ask for a 15-minute meeting in the first message](#rule-3-never-ask-for-a-15-minute-meeting-in-the-first-message)
- [Bonus: stop over-personalizing](#bonus-stop-over-personalizing)
- [Put the three together](#put-the-three-together)
- [Questions founders ask](#questions-founders-ask)

Share

Founders doing this with PumpGTM

[Workers.io Workers.io tests its message with real engineering leaders before the whole team joins](/customers/workers-io)[Supermemory Supermemory put founder-led LinkedIn outreach on autopilot](/customers/supermemory)[Docmost Docmost follows up with trial signups and abandoned checkouts on LinkedIn](/customers/docmost)

## Your first batch of buyers goes out this week.

Give us your website. PumpGTM finds the people who have this problem right now, you approve the list, and connection requests start from your own LinkedIn.

## More guides

[### GTM engineering: why a lead is fit, timing, intent and confidence Finding leads is easy now. Knowing which lead matters is the hard part. The GTM engineering approach we run for 10+ YC companies: score every account on fit, timing, intent and confidence, then close the loop from signal to revenue. 8 min read](/blog/gtm-engineering-lead-scoring)[### Is LinkedIn automation safe? How PumpGTM protects your account The honest answer: LinkedIn automation can absolutely get you restricted, and I got my own account restricted twice before I figured out why. Here is what actually keeps an account safe, and how PumpGTM does it by default. 4 min read](/blog/is-linkedin-automation-safe-pumpgtm-account-protection)