Last month we sent 13,898 emails across 11 campaigns. We got 164 replies, 80 of them positive. We used zero personalisation and zero AI.
I am putting that at the top because almost every article on this subject tells you to personalise harder, and almost all of them are published by companies that sell personalisation. Enrichment credits, AI first line generators, research tools. The advice is not dishonest. It is just worth noticing who benefits from you believing it.
Here is what I actually think: the list is the personalisation. Not the copy.
What that claim is not
It is not "personalisation does not work." It is a claim about order.
If the person genuinely has the problem you are writing about, a plain opener and a clear sentence about that problem gets a reply. If they do not have the problem, no amount of research into their recent posts creates one. You end up with a beautifully personalised email about something they do not care about, which is a more expensive way to be ignored.
The pattern I see over and over: heavy personalisation is what a weak offer leans on to compensate. When the offer is relevant and the data list is tight, you do not need the theatre. When it is not, the theatre does not save you.
Vague copy is a symptom, not a disease
The most common copy problem is writing for everyone. "Hey, I help B2B companies book more meetings" could be sent to anyone, which is exactly why it lands with no one.
But that is not a writing failure. The copy is vague because the list is vague. You cannot write a specific email to a list you have defined loosely, because there is no specific thing that is true of everyone on it.
Nail the definition and the copy almost writes itself.
The specificity upgrade
Here is the move, and it is the most useful single thing in this post.
Take your target description and make it narrower until it names a group that would recognise itself.
"B2B cold emailers" becomes "founders at Clay certified agencies."
The first is a category. The second is a room full of people with the same tooling, the same problems and the same week. You can write one email to the second group that is true for all of them, which is what relevance actually means. You cannot do that for the first group at any length.
Do that on the list and you get the effect people are chasing with first line generators, for free, at the top of the funnel instead of the bottom. The free play walks through the targeting approach I use to get there.
Build the personas from closed won, not from assumptions
The way to find those groups is not a brainstorm.
Take every closed won customer and map them properly: industry, role, growth stage, tech stack, buying signals. Look for the patterns that are actually there rather than the ones you expected. Build the campaign personas off that.
Then run several at once. Each campaign of ours has four distinct personas with A, B and C copy variants going simultaneously, which means more experiments running at any moment than we have ever had, and the list tells us which segment is real rather than us guessing in advance.
The same problem lands differently depending on who is holding it. A senior leader experiences it as revenue risk. A manager experiences it as whether their team hits the number. The person doing the work experiences it as the part of their week it ruins. One email written for all three speaks to none of them, and that is a targeting decision, not a writing one.
The same problem, three different people
There is a second targeting decision inside every list, and almost nobody makes it deliberately.
You can have the right company and still have the wrong person, because the same problem is experienced three different ways depending on who is holding it.
The Observer sits above the problem. A founder, a C-level leader, a VP. They meet it as risk and as numbers: revenue that did not arrive, a plan drifting off track, a cost that keeps climbing. They do not feel it daily and they will not be the one implementing whatever you sell. What they want to know is whether this is a threat and whether it is already handled.
The Manager owns the outcome. They are measured on the thing you affect. They meet the problem as pressure, because their team is not hitting the number and they are the one explaining that upwards. They are usually the most motivated of the three, since they carry the result without the authority to ignore it.
The Doer lives in it. They do the work your product changes, and they meet the problem as friction: the hours every Thursday, the spreadsheet that breaks, the tool that will not talk to the other tool. They often cannot buy. They can absolutely kill a deal, and they will tell you in one reply whether your understanding of the problem is real.
Write one email for all three and it reaches none of them. The Observer finds it lost in the weeds. The Doer reads it as vague marketing. The Manager, who was your best shot, gets an email that never names the pressure they are actually under.
Deciding which of the three you are writing to is a list decision, not a writing one. It costs nothing, it happens before you open a document, and it does more than any opening line.
What this buys you in numbers
My best week was 738 sends per positive reply. My worst was 6,355.
Same product. Same person writing. The only variable was how tight the targeting was.
That is roughly an eight times swing, and it came entirely from the list. I have never got a swing like that from rewriting an email.
The exception, which matters more than the rule
Everything above describes tier two and tier three accounts, where automation is correct and personalisation is mostly decoration.
There is a tier where the opposite is true. Your dream accounts. The named list. Anything where the contract value is in six figures.
Do not automate those. Do not blast them. Go one to one, with proper research, every time.
The logic is just arithmetic. On a list of thousands, an hour of research per prospect is an absurd trade. On an account worth six figures, an hour is nothing and the email being obviously hand written is part of the signal. Automating your tier one list to save time is saving the wrong resource.
Most of the argument about personalisation is people applying tier one instincts to tier three lists, or tier three economics to tier one accounts. Sort the tiers first and the argument mostly dissolves.
Where I draw the line on AI
I use AI heavily. I don't use it for fake familiarity.
No hallucinated personalisation lines. No "I noticed you posted about X" when the post has nothing to do with why I am writing. A merge field is not knowledge and a scraped detail is not a relationship, and the reader can tell.
If you want the longer version of that argument, it is in AI cold email: you do not need agents to book meetings.
What to do with the hours you get back
Personalisation at scale is expensive in the one resource you cannot buy more of, which is your attention. If you stop spending it on first lines, spend it on the list instead.
That means tightening the definition until it names a room rather than a category, verifying the data before it goes out, and mapping your closed won properly rather than trusting the profile you wrote a year ago.
Less satisfying than rewriting an email. Nobody gets a dopamine hit from narrowing an ideal customer profile. It is still where the result lives.
If your campaign is currently flat, the order to work in is in cold email not getting replies.