Why are my cold emails getting no replies? Debug them in this order

I sent 100+ cold messages to scraped lead lists for my products and got 0 replies. Then I sent 42 DMs to people who had posted about the exact problem I was solving, and 12 of them answered.

It was me sending a similar message both times. What changed was who received it.

That's the idea behind the debug order below. It comes from Christina Gilbert, a YC Visiting Partner and the CEO of OneSchema, in her talk on why you're getting zero replies to your cold emails. When nobody answers, the instinct is to rewrite the copy. She checks who got the message first, and only looks at the copy after the first three checks pass.

What reply rate should you expect from cold email?

Gilbert's bar is 2 to 3 replies per 100 cold emails. At that rate a campaign is worth iterating on. (I found that number through an infographic summary of her talk, which is also how I found the talk.)

Warmer outreach should do better. The target I use for DMs to people who already talk about the problem is 5% or more. So if a cold campaign sits at 0 or 1 reply per 100, something earlier in the process is broken, and it's probably not the wording.

One more thing before the list: a reply is only the first step. I had a week with 63 cold messages, 4 replies and zero customers. Track how many people actually start using the product from the first week, otherwise a decent reply rate can hide that nobody signs up.

The debug order

Fix the first broken step before you touch the next one. It works the same for email and for LinkedIn or X DMs.

1. Is it the right person?

Is this the job title that actually buys? Gilbert tells of a founder selling an infrastructure product that senior engineering leaders buy, who kept emailing DevRel and GTM managers. He got zero replies, and no email would have fixed that.

If you have customers, look at who pushed the last deal through and who paid. That's the title. With no customers yet, pick the role with the most to gain if the problem goes away.

2. Is it the right company, right now?

Someone can match your ideal customer profile perfectly and still not have the problem yet. Founders still building their first product will tell you a distribution tool sounds useful, and they won't buy it, because that pain starts after launch.

Check that each lead is at the stage where the problem happens now. If you're not sure who your ideal customer is in the first place, score your segments first.

3. Would you open it?

The subject line decides if the email gets read at all. Keep it to a few words, with no category buzzwords: "Quick question" still beats "AI-powered platform for agencies".

On LinkedIn your profile is the subject line, since it's all they see before accepting the request. A real photo and a plain line on what you do and for whom.

4. Is the message about them?

Only now look at the copy. It should be specific to that person, use the words customers use for the problem, and explain the product in one sentence in terms of what they need to hit this quarter. Gilbert's example is that executives don't care that your database is built in Rust (as someone who likes Rust, I have to agree).

Keep it under 80 words for the US and UK, up to 100 for Europe. Open with a question about their problem and don't ask for a call in the first message. "Want me to send a 2-minute demo?" is a yes or no. "Do you have 30 minutes this week?" is a lot to ask from a stranger.

When someone does reply, ask what got their attention in your message, and reuse the answer in the next ones. The way they describe it usually works better than what you'd write yourself.

5. Does what you send hold up?

If the email points to a demo, a sample or your site, does it hold up when someone skeptical clicks it? A landing page with no proof on it can ruin a campaign where everything else was fine.

6. Is it arriving?

For email, that means the inbox and not spam. Gmail and Yahoo have required SPF, DKIM and DMARC on sending domains since February 2024, Microsoft since May 2025, and Gmail started rejecting non-compliant mail outright in November 2025 (Google's sender guidelines). Keep the spam complaint rate under 0.1% in Google Postmaster Tools, send 30 to 50 a day per inbox, and leave links out of the first email.

For LinkedIn, check whether the connection request was even accepted. A pending request is a message nobody saw.

Delivery is 6th on purpose. It's the easiest thing to blame, but a perfectly delivered email to the wrong person still gets 0 replies.

7. Is it product-market fit?

If all six check out and replies are still near zero, more copy tweaks won't fix it. At that point the problem is probably the offer itself, and that's a validation question more than an outreach one.

Send the first 100 by hand

Gilbert's other rule, and the one I'd underline: send at least your first 100 by hand before automating anything. A thousand automated emails with zero replies can't tell you which step broke. A hundred by hand show you every bounce, reply and silence, and you can change one thing at a time.

A sending tool is useful once you have a message that gets replies, so you can send more of it.

What I'd do with a dead campaign today

1. Stop sending.
2. Pull 20 leads and check steps 1 and 2 by hand. Right title? Problem happening now?
3. Look for people who posted about the problem in public, on Reddit, Indie Hackers or in LinkedIn comments. For me that was the best list, and it's where my 12 replies out of 42 came from.
4. Send those 20 by hand, and compare the replies before you change anything else.

This debug order is one page of the Distribution Framework, the playbook I use to find customers for my own products.