Cold Email Templates That Got Me Referrals at FAANG (Without Being Cringe)

Why most cold messages get ignored

The default cold message you see on LinkedIn looks like this:

"Hi Alex, I'm a Data Scientist with 3 years of experience looking for opportunities at Google. I'd love to connect and learn more about your team. Please let me know if there are any openings."

This gets 1-2% reply rates. The reason isn't that the recipient is mean — it's that the message could have been sent to literally any person at any company. There's zero signal that you specifically are reaching out to them specifically.

The fix is one sentence: a real, specific reference to something the recipient has done. Everything else flows from there.

The 4-line cold email structure that actually works

Every high-reply-rate cold email I've ever sent or seen follows this structure:

1. The hook — one specific thing they shipped, wrote, or talked about. Not "I follow your work." A real artifact.
2. Your credibility line — your role plus one quantified achievement that maps to what their team cares about.
3. The ask — one small, specific, easy-to-say-yes-to ask. Not "can we hop on a call."
4. The graceful exit — "No worries if not a fit" or similar. Removes pressure, ironically increases replies.

That's it. Four lines, often under 100 words. Long cold emails do worse than short ones.

Template 1: Cold email for a referral

Subject: Loved your post on RAG eval at Acme — quick question

Hi Alex,

Your post on swapping faithfulness scoring into Acme's RAG eval pipeline was the clearest explanation I've read this year — I rebuilt our internal eval the same week using your hybrid retrieval approach.

I'm a Data Scientist at Beta Corp; I shipped an LLM-based intent classifier that cut support ticket misroutes by 41%, and I'm looking at the open ML Engineer role on your team.

Would you be open to forwarding my resume internally, or pointing me at the right hiring manager? Either is hugely appreciated, no pressure if it's not a fit.

Thanks,
Kuldeep

Why each line is there:

• Subject — specific to their work, not yours. Opens 70-80% of the time.
• Hook — names a concrete artifact ("post on RAG eval") plus a verifiable claim ("I rebuilt our internal eval the same week"). Burns 30 seconds of their attention on something that sounds like a real person.
• Credibility — role + one quantified result that maps directly to the role you're targeting. Quantified is the keyword.
• Ask — forward resume OR point to hiring manager. Two micro-asks, both 30-second favors.
• Exit — releases pressure. Counterintuitive but consistently lifts reply rate.

Template 2: Recruiter inbound reply (you got the InMail, this is your follow-up)

Hi Sarah, thanks for reaching out about the ML Engineer role at Acme.

Short intro: I'm a Senior Data Scientist at Beta Corp; recently shipped a churn model that reduced false positives by 32% and a real-time recommendation pipeline serving 8M users at p99 < 80ms. Stack overlaps closely with what's in the JD — Python, PyTorch, Ray, AWS, dbt.

I'm available for a chat Tuesday or Thursday afternoon next week. What's the best email or calendar link?

Thanks,
Kuldeep

Recruiter replies have a different bar: they already opened the conversation. Your job is to match yourself to the JD in 60 seconds. Lead with two quantified results that map to JD bullet points. Name the stack overlap. Propose a time. Done.

Template 3: LinkedIn DM (use the 300-character limit on your side)

Hi Alex, your RAG eval write-up was the clearest I've read this year — rebuilt ours after reading it. I'm a DS at Beta Corp (shipped LLM intent classifier, -41% misroutes). Eyeing the ML Engineer role on your team. Open to a quick referral conversation? No worries if not a fit.

Same structure compressed. Under 300 characters fits the InMail preview window and shows up fully in the recipient's notification panel — they see the whole message before they decide to open.

The 5 things that kill cold-message reply rate

1. Generic compliment — "I love your work / your posts" with no specific reference. Reads as bot output.
2. Asking for time without giving them anything — "Can we hop on a 30-minute call?" is a huge ask for a stranger. Smaller asks (forward, point me, one-line question) convert 5x better.
3. Long preamble — "Hope you're doing well, I know you're busy, I'll keep this brief" is the verbal equivalent of asking for permission to ask. Skip it.
4. Pasting your resume into the message body — clutters the message and dilutes the ask. Attach it OR drop a link. Better: get the reply first, send resume in reply.
5. Sending the same message to 50 people same day — LinkedIn flags it. So do humans when a friend forwards three identical messages.

Follow-up cadence

Send one follow-up 5-7 days after the original message. Keep it short:

Hi Alex, bumping this in case it got buried — happy to leave you alone if not a fit, just didn't want to assume.

That's the entire follow-up. Two lines. No new ask, no new context dump. If still no reply after the follow-up, move on. Two unanswered messages is plenty; three feels like nagging.

Reply rates you should expect

• Highly personalized, specific hook, small ask: 15-25% reply rate.
• Personalized but generic ask ("can we chat"): 5-10%.
• Slot-filled but generic hook ("I follow your work"): 1-3%.
• Pure spray-and-pray: <1%, and usually LinkedIn-flagged.

If your reply rate is below 5%, the hook is the problem 9 times out of 10. The fix is always more specificity — what exact post / project / talk / repo are you referencing?

Generate your message in 30 seconds

If you'd rather skip the blank-page problem, the free cold email + LinkedIn DM generator on getjob4u uses this exact structure — referral asks, recruiter replies, alumni outreach, follow-ups. Fill in the slots and copy a working draft to personalize.

The single field that matters most is "Their work" — be specific (mention a paper, repo, talk, or shipped feature). That field is the difference between 2% and 20%.