Personalization that isn't
“I saw you were recently active on LinkedIn.” That sentence is sitting in thousands of inboxes right now, and every single recipient knows in the same second what it means: a tool found them, not a human. The email is dead before the second sentence gets read.
The absurd part: the sender thinks this is personalization. First name, company, industry, maybe the latest funding round or the job title as a merge field, five variables neatly populated. Technically, that is personalization. In terms of effect, it’s the opposite, because variables only prove that a database works. They don’t prove that someone engaged with the recipient, and that proof is the only currency a cold email has.
The recipient checks this unconsciously with one simple question: could this sentence have gone to 500 other people? “Congrats on the Series A” fails the test, that’s in every Crunchbase feed. “Your job posting for the second SDR mentions HubSpot, but your careers form visibly sends its data nowhere” passes it. The difference is not a better variable. The difference is an observation that cost a real minute and touches a real problem.
Of course this scales worse. That’s exactly why it works. The value of a researched sentence lies precisely in the fact that your competition can’t automate it, at least not credibly. AI research tools are currently promising to close this gap, and mostly produce just the next generation of recognizable patterns: “I read your post about leadership, powerful thoughts!” Anyone who gets a lot of cold email has long since built an inner spam filter for this sort, and it gets trained sharper every day.
The consequence for the process is uncomfortable but clear. Better 30 emails per week with one real observation each than 300 with merge fields. The reply rates don’t differ by percentage points, they differ by a factor of ten, and the 30 emails don’t burn the market on the side. Research belongs in the playbook like a required field: what exactly was looked at about the prospect, what is the one observation the email hangs on? If that question has no answer, the email isn’t done.
The simplest quality test remains the self-test. Send yourself your own sequence. Would you reply, or would you know the score within a second?
