Articles·Go-to-market

“I noticed you noticed me.” Delete. What replaced personalized cold email.

Chris Cordon · 4 min read
A smug man in aviator sunglasses leans back with his boots on the desk, grinning, after sending a cold email.

Personalized cold email stopped working the day everyone started doing it. The fix is to keep the research and add an opinion about what you found. I learned this the expensive way, across three rounds of campaigns this year that added up to a very complete list of what not to do.

Why do personalized cold emails get deleted?

“I saw your LinkedIn post last week” is now the fastest spam signal in the inbox. Two years ago, opening with a scraped detail proved you were paying attention. Today it reads as a template with a merge field, and it’s deleted before the second sentence.

AI made it worse. Left alone, an AI copywriter congratulates people on whatever it finds, every time. “I saw you held an all-hands in April. That’s not easy for someone in your seat.” I sent a version of that line in March and spent 9 campaigns and 371 leads to learn that it wasn’t working, and I got really tired of seeing the word “seat.”

What do most agencies try first?

Most agencies try a merge field and a bigger list. “Hi {First Name}, I see you’re the owner of {Company}.” That’s the 2023 version.

Here’s what I do know: you can’t send cold email from your business domain. I once worked with a very “seasoned” producer who insisted on sending volume from the company account. It would have been hilarious if I hadn’t been sharing that domain. When a domain gets flagged, every email on it goes to spam, renewals included. Some owners hand it to a lead-gen vendor with a shared template. The list is fine. The email is the same one the recipient got from four other people that week.

Wanna fix it?

Personalized cold email gets fixed with separate sending plumbing, fresh signal data, an opinion on every signal, and a person who approves every line.

  1. Separate plumbing.
    • Spend $10 on a domain that kind of sounds like yours.
    • Get no more than 5 mailboxes: some version of your name@the domain.
    • Then you have to warm that domain up. Either find a platform that’ll do the warming for you, or call all of your cousins so you can email back and forth for a month.
    • Don’t send any more than 30 per day, or you run the risk of landing in spam.
    • If you start landing in the spam folder, you’ll keep landing in the spam folder.
  2. Signal data in three tiers. A research agent looks for something the person did in the last 30 days. Nothing there? Something the organization did: a contract won, a location added, an acquisition. Still nothing? The state or the industry: a rate filing, a law that changed. Every contact gets the freshest tier that exists.
  3. Judgment on the signal, never applause. The email says what the signal means for them. From a construction segment: “Your union crews’ plan just renewed for five years, so they’re set. Your non-union crews are getting the same increases the rest of us pedestrians are.” Nobody got congratulated.
  4. Personalization moves to the second email. By August, my third wave led with the plain problem and put the personal detail in the follow-up. It read less like surveillance.
  5. A human approves the templates and the replies. My copywriter agent drafted, and I rejected its templates for months before a batch went out clean. Replies landed in my CRM with a suggested response already written, and I approved, edited, or killed each one.

What does it look like once it’s running?

Once it’s running, you’re the judgment layer and nothing else. The agent finds the signal, writes the line, and drafts the reply. You decide: approve, edit, or kill. Campaigns run about 70 contacts each, 40 to 45 of them sendable after enrichment. Small, focused segments are the best practice as of this writing, and best practice in cold email changes fast.

Volume takes the emotion out of sales. More at-bats, fewer feelings about any one of them. You can’t even smell your KPIs until you have the at-bats.

Can you do this yourself?

Yes, for a short list. ChatGPT can find a company’s recent news and draft an opinionated line if you hand it the facts. A sending tool like Instantly handles the mailboxes, the warmup, and the daily caps.

Where DIY stops is scale. At a few hundred contacts, the research has to run without you, the AI has to hold your voice on every draft, and replies need to land somewhere with a next step attached. That’s three systems talking to each other. If you’ve got 20 prospects, do it by hand. If you’ve got 2,000, build the machine.

Quick answers

Does personalized cold email still work?
Yes, with an opinion attached, and usually in the second email instead of the first.
Can I send cold email from my agency’s domain?
No. If it gets flagged, your client emails go to spam with it.
How many mailboxes per sending domain?
No more than five.
How many emails per mailbox per day?
About 30. More than that and you risk landing in spam.
What if there’s no recent news on the person?
Drop to the organization, then the state or industry. There’s always a third tier.
How long before the first campaign is ready?
Longer than you think. I spent months rejecting templates before the AI sounded like me.