Cold Email That Gets a Reply: What 1,153 of Ours Showed

A cold email is a commercial message sent to someone you have had no prior relationship with. The one that gets a reply does not differ from the rest through writing technique. It differs because it says something true and specific about the recipient’s company, something that person recognises immediately as being about them.

TL;DR. Between 1 and 29 August 2026 we sent 1,153 cold emails to Romanian industrial companies and received 23 replies, most of them positive. The reply rate on the large campaign was 5.51%, at a bounce rate of 1.09%. What produced the replies was not surface personalisation, it was a real figure from the company’s filed accounts, placed in the first line. What almost broke our deliverability were guessed addresses that a verification service had reported as acceptable.

What does a cold email that gets a reply actually mean?

It means a message in which the person recognises a problem they have right now and that nobody outside their company has named for them before. Every other element, from length to sending time, matters far less than people believe, and those elements only become important once the first thing is solved.

The test we run before sending any message is the same test we apply to brand messaging. If the sentence in the first line could be sent identically to five hundred other companies, then it says nothing about the recipient, however many variables it contains. A variable auto-filled with the company name is not personalisation, it is a populated field.

Will Allred, co-founder of the Lavender platform, describes exactly this failure when he talks about surface personalisation. “I see you’re an Atlanta Braves fan, or you’re in Europe so Manchester United. And it’s like, yeah, yeah, yeah. So what?” His point is that the information you find about a person has to serve a purpose, meaning it has to connect to a real problem, or it stays an ornament.

How good are our numbers against the market?

That depends entirely on what gets divided by what, and this is where the largest confusion in the whole field hides. Belkins reports an average reply rate of 0.45%, measured across 7.5 million emails sent in 2025, while Instantly reports 3.43% on their own 2025 data. The near eightfold gap between the two does not come from the market, it comes from the denominator, because Belkins divides replies by total messages sent, follow-ups included, while other providers report differently.

The practical conclusion for a founder receiving a report from someone is to ask first what was divided by what. A reply percentage without a declared denominator is not a figure, it is an impression.

Here are our numbers, read straight from the platform on 29 August 2026, with the denominator stated openly.

Our reply rate is calculated the same way Instantly calculates it, meaning replies over total messages sent. The number of people reached is lower than the number of emails because each person receives, on average, between two and three messages from the sequence.

What these figures do not say, and it deserves saying, is that a reply is not a contract. The 23 replies produced scheduled conversations, requested analyses and a few proposals, and the road from “agreed” to a signature stays long. Anyone selling you cold email as a contract machine is selling you something other than what you receive.

Why does deliverability fail before the reply rate does?

Because a wrong address does not produce a negative answer, it produces a technical rejection, and rejections accumulate in your domain reputation until the good messages stop arriving. The difference between the two campaigns in the table above, 1.09% against 5.36%, does not come from the quality of the text. It comes from how the addresses were obtained.

On the clinic campaign, some addresses had been constructed by permuting names against the company domain, meaning by guessing, and the verification service had marked them acceptable because those domains were configured to accept any mailbox. A server set up that way says yes to any address you ask about, so the verifier had nothing to confirm. The lesson we wrote into our procedure after that episode is that such a response means “cannot be known” rather than “verified address”, and since then first contact goes only to individually confirmed addresses.

On top of this sit the rules the large platforms have imposed, and those are not negotiable. Google has required, since 1 February 2024, that all senders to Gmail accounts have SPF or DKIM, valid reverse DNS and a spam complaint rate below 0.3%, and above 5,000 messages a day it also requires DMARC and one-click unsubscribe. Yahoo introduced identical thresholds the same month, plus an obligation to honour unsubscribes within two days. Microsoft went further on 5 May 2025 and now rejects outright the messages of high-volume senders that fail authentication.

There is a nuance most translated articles miss. These thresholds are written for consumer mailboxes, meaning personal Gmail and Outlook.com, whereas an industrial prospect usually sits on the business infrastructure of the same providers, where filters behave differently. Your domain reputation stays the same everywhere, though, so there is no reason to lean on that distinction.

What goes into the first email to get a reply?

One verifiable observation about the person’s company, followed by one clear proposal. That is all.

We build the first sentence from the company’s filed accounts, meaning from public data, and the observation usually sounds like a change the person knows about and that nobody outside has mentioned. A net margin that dropped from above eleven percent to just over four, inside a single year, is a fact rather than a compliment, and the founder who lived through that year recognises immediately what it refers to. One recipient disputed our figure, which was also a reply, and checking against the full filed series showed the figure was correct.

The data on form confirms the intuition. Gong analysed more than a million sales cycles and found that reply rates drop sharply above 100 words, with the best emails landing between 50 and 100 words, and subject lines of one to four words performing best. The same analysis shows that executive-level directors reply 30.2% less often than others, which does not mean you should not write to them. It means the message to a founder has to be better, not merely more persistent.

Two things we got wrong and corrected deserve a mention here. The first is that value gets delivered in the first message rather than promised for after a call, because someone who receives something useful replies, while someone asked for an hour of their time weighs it up. The second is that you never announce what you are not doing, because a sentence like “I am not asking you for anything right now” plants in the reader’s head exactly the idea you were trying to avoid.

What happens after someone replies “agreed”?

This is where most campaigns lose, and the loss hurts more because the hard part was already done. Someone who replied “agreed” said yes to the concrete thing you promised in the email, not to a generic conversation, and if you send them an automated scheduling message, you have cancelled the very reason they replied.

The rule we apply now is that a promise made about a specific company gets honoured on that company, with the work done. If the email said we would show them why their margin fell, then the first thing going back is that analysis, not a calendar link. It costs more and it cannot be automated, which is precisely why it works.

Why is a reply not a decision?

Because the person who replied does not decide alone, and that changes how a campaign should be read completely. Forrester measured in 2024 that a typical B2B buying process involves thirteen people, that 89% of purchases cross at least two departments, and that 86% of processes stall somewhere along the way. Amy Hayes, research director at Forrester, sums the situation up in a single sentence, “a broken B2B buying process creates chaos for both buyers and sellers”.

An even more uncomfortable figure sits on top of that. 6sense found in 2025 that first contact with a salesperson happens only after 61% of the buying journey, and Kerry Cunningham, who leads research there, says that “in 84% of cases, buyers reported that their first conversation was with the vendor that eventually won the contract”.

Those two data points, put together, explain why cold email works differently than it appears to. It convinces nobody to buy, because people convince themselves earlier, from what they find about you. The email does one single thing, and that thing matters enormously. It puts you in the conversation at the moment somebody has already started looking, and the rest depends on what they find after they search your name.

This is why we do not treat the campaign as a separate channel. A cold email that earns a reply, followed by a website where the person finds nothing that confirms the promise in the message, produces a conversation that dies on its own within two weeks.

Is it legal in Europe?

Here a point deserves stating plainly, because most articles get it wrong by assuming Europe is one jurisdiction on this. The European regulation permits processing personal data on the basis of legitimate interest, under Article 6(1)(f), and the ePrivacy Directive leaves member states free to decide whether the protection extends to legal persons as well, through Article 13(5).

That single sentence is why the answer changes at the border. Some member states apply prior consent only to natural persons, which leaves room for business-to-business outreach on a legitimate interest basis. Others extended the protection to companies. Romania, where our own campaign ran, chose the strict route: Law 506/2004, at Article 12(4), states that the provisions on prior consent apply “correspondingly to subscribers that are legal persons”, meaning the requirement for express prior consent covers companies too. The practical consequence is that you cannot lift a playbook from one European market and assume it transfers to another.

What we do in practice, and what we recommend to anyone who asks, is to treat every message as though it had to be defensible in front of the recipient. The sender identity is visible, the reply address is real and monitored by a person, unsubscribe works from the first message and is honoured immediately, and the data used is public, taken from filed accounts and the company’s own website. For a legal assessment of your specific situation, ask a lawyer, because this article describes what we read in the legislation rather than standing in for advice.

Frequently asked questions

How many emails do I need to send to know whether it works? Enough for one variant to accumulate at least a hundred sends, otherwise the difference between two variants is statistical noise. Below that threshold there is nothing to interpret, however tempting it looks.

Does open rate still mean anything? Almost nothing, since Apple introduced Mail Privacy Protection, which hides the IP address and prevents the sender from seeing whether a message was opened. We do not track opens at all, because the tracking pixel damages deliverability without producing a number you can trust.

What bounce rate is acceptable? Industry convention settles around 2%, although no major email provider publishes an official threshold. Below 1% means the list is clean, and above 5% means you stop the campaign and fix the address source rather than the text.

Is it worth outsourcing? That depends on what you outsource. The technical part, meaning sending infrastructure and address verification, outsources well. The observation about the recipient’s company does not, because it requires understanding their business, and that is where it gets decided whether you receive a reply.

Cold email or paid ads? The two answer different questions. An ad buys the attention of someone already searching, while a cold email reaches someone who is not searching but has the problem. When you lack clarity on who the ideal client is, both cost the same and produce equally little, which is why the correct order starts with positioning and with the architecture of the growth system, not with the channel.

If you want to see what verifiable observation could be written about your company, meaning exactly the kind of sentence that produces a reply, you can request a positioning analysis. We run it on your public data and show you how you look from the outside, before any campaign.