What makes an outreach message work
A LinkedIn message gets a reply when three things are true: it arrived from someone vaguely familiar, it shows evidence a human looked at the recipient specifically, and it asks for nothing the reader isn't ready to give. Every technique in this guide serves one of those three. Most outreach fails by inverting them — a stranger, a template, and a calendar link.
The familiarity part is built before the message, not inside it: profile views and genuine engagement in the days before you write. That warm-up sequence is covered in the complete outreach guide; this one assumes it and focuses on the words.
The connection note: shorter than you think, often absent
The note field is 300 characters, and the strongest use of it is restraint. A good note names the real reason you are connecting and stops. No pitch, no link, no "I'd love to pick your brain."
- "Your post on churn in agency retainers matched what we're seeing — connecting so I catch the next one."
- "We're both selling into heads of RevOps; your take on comp plans was the least boring one I've read."
And frequently the right note is none at all. After real engagement — they have seen your name react to their content — a blank invite reads as natural, while a clumsy note reads as a campaign. If you cannot say something specific, say nothing.
The first message: one observation, one give, no ask
The first DM after connecting has one job: prove the connection was about them. Structure that survives contact with real prospects:
- Open with their thing, not yours: something they posted, shipped, hired for, or said.
- Add a point of view or something useful — an observation from their market, a genuinely relevant resource, a short answer to a question they raised publicly.
- End without a meeting ask. A question they can answer in one line beats a calendar link they can ignore in zero.
Worked example, generic on purpose:
- "Saw you're hiring two SDRs — usually means outbound's about to scale. Curious how you're planning to keep quality up while volume doubles? (We went through this last year and got it half right.)"
What killed it in the template era: opening with your own company, three paragraphs of anything, more than one link, and any sentence that would fit every prospect on your list equally well. Prospects pattern-match templates in under a second — the test for every line is "could this have been sent to someone else unchanged?"
Follow-ups: add, never bump
Two or three follow-ups, days apart, each one carrying something new. "Just bumping this" and "did you see my last message?" convert silence into annoyance. Instead:
- "Different angle on the same question — saw [their company] just launched in DACH; does the outbound playbook change per market for you?"
- "No response needed — this piece on SDR ramp time covered what I was fumbling toward. Worth the five minutes."
The last follow-up should close the loop politely and leave the door open. People reply to graceful exits surprisingly often.
Voice notes: the highest-signal follow-up
A short spoken message in an open thread is still rare enough to feel personal — it proves a human spent the minute, which is exactly what text templates cannot. Keep it under 30 seconds, reference the thread, and don't pitch. Voice works best as a second or third touch, once the thread exists, not as an opener from a stranger.
Three first-message shapes, torn down
Templates fail when they are used as templates — but the underlying shapes are learnable. Three that survive contact with real prospects, with the reason they work:
1. The observed-change opener
"Saw you're hiring two SDRs — usually means outbound's about to scale. Curious how you're planning to keep quality up while volume doubles?" — Works because it ties a public, verifiable change (the hire) to a problem the recipient plausibly owns this quarter. Fails when the "change" is stale or generic ("saw your company is growing!") — the specificity is the entire mechanism.
2. The point-of-view trade
"Your post on agency churn matched what we're seeing — one pattern you didn't mention: the retainers that survive are the ones with a weekly artifact. Is that true for you or are we overfitting?" — Works because it gives before it asks: a real observation the prospect can disagree with. The question invites a one-line reply, not a meeting. Fails when the point of view is flattery in disguise ("great post!") — agreement without substance is noise.
3. The honest-context ask
"We're building for exactly the problem you posted about, so full disclosure, I have a horse in this race — but the question is real: did you solve the handoff between research and first message, or just live with it?" — Works because disclosed self-interest reads as more honest than pretended neutrality, and prospects reply to candor. Fails if the follow-up pitches immediately — the disclosure buys one honest exchange, not a sales sequence.
The common thread: every shape ends somewhere the prospect can reply in one line. None of them contain a calendar link.
AI drafting, honestly
Full disclosure: we build Nova, which drafts outreach messages with AI — so here is the honest version of where AI belongs. AI is genuinely good at synthesising research (their posts, role, company signals) into a specific draft, and terrible as an unsupervised sender. The line that matters is approval: a human reads and okays every message before it goes out. That is how Nova is built — drafts land in an approval queue, nothing auto-sends — and it is the standard worth demanding from any tool (comparison here). Volume of unread AI slop is not a messaging strategy; it is a reputation liability with your exact target market.
Judge messages by the right numbers
- Reply rate on first messages — relevance check.
- Positive-reply share — a "no thanks" and a "tell me more" are different outcomes; track them separately.
- Conversations that reach a meeting — the metric the message system exists for.
If replies are low, fix the list and the research before the copy — a mediocre message to the right, warmed-up person outperforms brilliant copy to a cold list every time.
Frequently asked questions
How long should a LinkedIn outreach message be?
Short: two to four sentences for a first message, and under 300 characters for a connection note — which is also the hard limit LinkedIn imposes. The length ceiling is really an attention ceiling: a prospect decides whether to keep reading in the first line, on a phone, between meetings. A strong first message spends that budget on one specific observation about the recipient and one useful thought, then stops. If a message needs a second paragraph, it is usually trying to do the job of the second or third message — introduce the company, explain the product, ask for a call — before the reply that earns the right to any of it. Voice notes bend the rule usefully: thirty spoken seconds carry more warmth than a hundred written words, without ever reading as a wall of text.
Should I send a connection note or a blank invite?
Send a note only when you can say something specific to that person; otherwise a blank invite genuinely performs better. After real engagement — your name on a reaction or comment they have already seen — a blank invite reads as natural, while a generic note ("I came across your profile and was impressed…") pattern-matches instantly to a campaign. The test for a note is simple: could it have been sent to anyone else on your list unchanged? If yes, delete it. If you can reference their actual post, their hire, their market, a short note beats blank — no pitch, no link, no meeting ask, under 300 characters. And whatever you choose, never paste one note across a whole list: a template that gets reported does more damage to your account than a hundred blank invites ever will.
How many follow-ups should I send on LinkedIn?
Two or three, spaced days apart — and every one of them must add something new. The follow-ups that work carry a fresh angle, a genuinely relevant resource, or a reaction to something the prospect just posted. The ones that destroy threads are the bumps: "just floating this to the top" adds nothing, signals automation, and converts silence into a mute. After the final follow-up, close the loop politely and leave the door open — graceful exits get replies surprisingly often, and a prospect who ignored four messages this quarter may reply next quarter when the timing is theirs. Silence is an answer about timing, not a request for volume. Spacing matters as much as count here: same-day follow-ups read as pressure, while four to seven days apart reads as a person who remembered rather than software that scheduled.
Want researched, personalised drafts queued for your approval instead of a blank page? Start free trial — Nova is in private beta.