Abraham says 90% of cold email split tests are noise
Ten positions from a high-volume sender, none of them with a reply rate attached.
Nick Abraham posted a list of ten cold email beliefs he says he will fight anybody on, and the first two go straight at infrastructure orthodoxy. Branded domains, he says, matter far less than people think, and he claims his team has tested branded against non-branded infrastructure at scale with a performance gap nowhere near what is usually claimed. Sender names get the same treatment. Founder, SDR, female sender, he calls the whole debate useless in his experience.
The sharpest item is number three. Abraham argues that 90% of copy split tests are noise, and he explains the maths behind why.
"If variant A gets 2 positive replies and variant B gets 1, that is technically a 100% lift," he wrote. "It does not mean the wording caused it. Just send a better offer."
If variant A gets 2 positive replies and variant B gets 1, that is technically a 100% lift.
It does not mean the wording caused it. Just send a better offer.
That runs into item four, where he says swapping a few words or rewriting one case study sentence rarely changes results at scale, and item six, where he says a ten-step Clay workflow will not save a mediocre offer. The through line is that angle and offer move numbers and wording mostly does not, which is consistent with his earlier argument that gimmicks around the offer are a distraction.
The tactical items
Four of the ten are things you can act on this week. Abraham says recognisable client names beat bigger results in proof, because "We helped Apple with their rebrand" lands harder than helping an unknown company reach $50M. He says info@ and hello@ addresses are underrated and that for some SMB and local business campaigns they can outperform direct contacts. He says Calendly links can make booking harder, and that once somebody replies, proposing a few specific times often works better than sending them off to pick a slot.
He also says Saturdays and holidays are fine for a lot of SMB markets, with some holidays producing very strong lead volume for his team. And he argues the best campaigns often need less sophistication, since a simple relevant offer works when the prospect almost never sees that kind of pitch in their inbox.
What this is and is not
None of the ten carries a volume, a send count, a date range or a reply rate. Abraham references testing at scale on the infrastructure point but does not say how many sends, to which lists, or over what period. For people sending, this is one high-volume operator's read on what he has seen, and the split test item is the one with an argument you can check yourself. Small positive reply counts genuinely do produce enormous percentage swings, which is a statistics problem rather than an opinion.