2026-10-11

What 100 Cold Outreach Replies Actually Look Like (2026 Data)

Every reply our domain outreach received in 2026, read by hand: 100 messages scanned, 57 real inbox traffic after removing marketing mail, 23 threaded “Re:” lines, and exactly two from a human. The breakdown, the automation, and what we changed.

Two.

Out of every hundred replies our cold outreach produced this year, exactly two came from a human being who had read the message and formed an opinion about a name. The other ninety-eight were machines, broadcasts and receipts.

That single number — the human reply rate — is the one we should have been quoting all along. It is also the number almost nobody publishes, because the figure everyone else reports is the reply rate, and the reply rate is easy to inflate with noise nobody asked for.

We did not estimate this. We pulled every inbound message that landed in our outreach mailboxes in 2026 — the same threads we use to move names out of a 158-domain portfolio — and read all one hundred of them by hand. This is what a hundred replies actually look like when you stop counting the flattering way and start counting the truth.

What we counted, and how

A reply-rate claim is worthless without a definition, so here is ours.

• The unit is an inbound message, not an outbound email. Every distinct message that arrived on an outreach thread — a reply, a bounce-back, an automatic acknowledgement — counted as one candidate.
• The candidate set was 100 replies. One hundred inbound messages, each one read.
• Marketing mail came out first. Fifty-seven of those hundred were unsolicited newsletters, promotional blasts and platform digests with nothing to do with the campaign. Removing them left 57 messages.
• Then we looked at form. Of those 57, 23 arrived wearing a "Re:" prefix — the threaded form that reads, at a glance, like a conversation.
• Then we applied the only test that matters. A reply counted as human only if a person wrote it, engaged with the specific name, and was not a template, a receipt or a broadcast. On that test: two.

The strictness is the point. An out-of-office autoreply is addressed to you, lands on your thread and looks like correspondence. It is not. A helpdesk receipt carries your subject line and still is not a conversation. Count those and you can report any reply rate you like and learn nothing from it.

The hundred, in one view

Read the last row against the third. Twenty-three messages looked like a conversation. Two were.

The automated noise, in detail

The ninety-eight non-human messages were not random. They clustered into a few recognisable shapes, and each one teaches something about how to read a mailbox.

• The repeated auto-responder. The same text, word for word, arriving again and again on different threads — a holiday loop or a rigid out-of-office. It answers everything, so it tells you nothing. The giveaway is the repetition: same paragraph, same sign-off, same total absence of the name you actually pitched.
• The ticket notification with a reference. A support or procurement system confirming that a message was received and assigning it a case reference. Polite, instant, and useless. No one has read your email; a queue has.
• The inbound marketing offer. A pitch sent to us from another platform, landing in the same mailbox and looking, in a raw count, exactly like a reply. It is not a reply to anything. It is the noise floor of a busy inbox, and filtering it out is the difference between a number you can trust and a number you can brag about.

None of these messages is a failure on its own. Treating them as engagement is the failure — it hides the real signal and lets you keep sending the same way when you should be changing the approach.

The two that were human

Both real replies arrived, and both are worth more than the other ninety-eight combined — for opposite reasons.

1. "Take me off your list." A clear, unambiguous opt-out. It is not interest and it is not a lead, but it is a human who read the message, understood it, and made a decision. It is also the single most important instruction in any outreach programme: it was honoured immediately and permanently, with no follow-up question and no quiet re-add later. A sender who ignores that one sentence trades a small short-term number for a reputation that cannot be rebuilt.

2. A $50 offer against a $295 ask. The second human did what the other ninety-eight did not — they named a number. Fifty dollars against a $295 asking price is not a serious bid on its own, but it is the only message in the entire set that could ever become a transaction. We countered at $199 with a 72-hour deadline: a firm middle ground with an expiry attached, so the thread has a decision point instead of drifting into silence.

That is the whole human result — one opt-out and one negotiation, out of a hundred replies.

Three lessons we now run the programme on

1. Report the human reply rate, not the reply rate. "Replies" is a house number that can be made to say almost anything once newsletters, autoreplies and tickets are in the count. Humans who actually read the message is the figure that predicts revenue. Ours was 2%. If yours is higher, prove it the same way — read every reply and label it.

2. "Re:" means delivered, not interested. Twenty-three messages carried the threaded prefix that reads like engagement, and two were real. The prefix is an artefact of mail clients, not a signal from a buyer. When you quote a reply rate inflated by "Re:" counts, you are quoting your own plumbing.

3. The only reply with commercial value names a number. A lowball is a lead. A newsletter is a statistic. The gap between a $50 bid and a $295 ask is a negotiation problem, and negotiation problems get solved — with a counter, a floor and a deadline. Everything else in the hundred was scenery.

What we do now

Four changes came out of this exercise, and they cost nothing but discipline.

• Label before you delete. Every inbound message gets one tag — human, auto, ticket, newsletter, offer — the moment it is opened. The count becomes a fact rather than a feeling, and the human share is visible in a single line.
• Quote humans only. The weekly number we track is human replies per hundred outbound emails. Not opens, which are unreliable; not replies, which are noisy; not "conversations", which are usually receipts.
• Honour opt-outs instantly and permanently. Beyond the obvious ethics, it is self-defence: sender reputation is the asset that decides whether the next genuine reply reaches an inbox at all.
• Give every counter an expiry. The $199 counter carried a 72-hour deadline, and that deadline is now a rule rather than a one-off. A counter with no expiry is an invitation to wait, and waiting is where small offers go to die.

The pricing half of that last point runs on the same logic we use on marketplaces: an ask has to sit inside a band a buyer will argue about, or the negotiation never starts. That is the subject of the Atom Premium write-up — the commission ladder, the name-server rule and thirteen listings with real numbers, and it is why a counter of $199 is a decision rather than a guess.

What this means if you run outreach on domains

If you are a domainer or an investor running outbound — or paying someone to run it — take four things from the hundred.

1. Define the metric before the campaign, not after. "Reply rate" without a definition is marketing. "Human replies per hundred sends, counted by hand" is a number you can act on.
2. Budget for the real ratio. At two human replies per hundred, you need roughly fifty sends to reach one human who might buy. That is not a reason to stop; it is the cost of a lead, and it belongs in your planning rather than in your feelings about the campaign. It is also why a portfolio of buy-now priced names does so much of the selling for you — a listed name with a clear price lets a buyer act without waiting for your next email.
3. Treat a lowball as the beginning. A $50 offer against a $295 name is the normal opening move of someone who has not yet been told what the name is worth. Answer with a specific counter, a floor you decided in advance, and a deadline. That is exactly what we did with $199 and 72 hours.
4. A removal request is data, not defeat. It tells you the list contains people who will never buy — and every one of them you remove protects the deliverability of the messages that matter.

Two human replies out of a hundred is not a good ratio. It is, however, an honest one — and it is the number we now plan around. Everything else was noise, and we can prove it.