Your outreach report is measuring the wrong number.
Nearly 3x the reply rate. The same meetings.
Here is the study
It is also on its way to your inbox, so you can come back to it later.
Open the studyOpening it for you now.
Key findings
What 2,279 LinkedIn conversations showed
Seven findings from the study, each with its sample size. Rates come from the 1,082 conversations that had been live for at least seven days.
-
The benchmark: 13.4% cold, 57.4% warm
Of 942 mature cold conversations, 13.4% got a reply. One company announcement sent to 527 people replied at 4.6% and pulled that down; the other 415 cold messages replied at about 25%. Conversations that began warm replied at 57.4% (n=141). Reply rate is only the start, as the next finding shows.
-
Reply rate did not predict meetings
A fiduciary services campaign into a new European market replied at 13.3%, and 8 of its 11 replies (72.7%) reached meeting intent. A trade credit insurance campaign replied at 35.9%, and 26% of its replies did. Per 100 messages sent, both reached about 9 meeting conversations. (n=83 and n=66)
-
Be specific, not personal
Messages personalised with the name only replied at 10.7% (n=460). Adding company or role references moved it to 15.5% (n=458). One specific, verifiable observation about the recipient took it to 25% (n=24, directional).
-
Ask exactly one question
Messages with no question replied at 9.5% (n=693). Messages with exactly one question replied at 23.8% (n=239). Two or more questions did no better than one (n=10, directional).
-
A soft close wastes the replies it gets
When the close was a soft question such as "happy to share if useful", 18% of replies reached meeting intent (n=143). Stating the ask plainly took it to 40% (n=69), and holding the ask for the second message took it to 56% (n=123).
-
Follow up after 3 to 5 days
One follow-up sent 3 to 5 days after the first message revived 14.6% of silent threads (n=355). Waiting 6 to 9 days nearly halved that, to 8.2% (n=73). The follow-ups that worked were short, under 200 characters.
-
Cold messages should not be too short
Cold messages under 350 characters engaged 7.1% of recipients (n=28). Messages of 650 to 800 characters engaged 15.6% (n=45), with no falloff below 800. For warm conversations it flips: shorter wins.
To cite: AscendAI, The 2,279 LinkedIn DM Study, 2026, ascendai.tech/dm-study.html. Observational data from AscendAI's own client programmes, not a controlled experiment.
The asset
What's in the study
Around twelve pages. Five pieces of standard outreach folklore put against the outcome that pays, plus the anatomy of the messages that produced meetings and a scorecard for your own.
- Specificity against personalisationWhat one verifiable observation about the recipient is worth, measured against plain name and company merge fields.
- How many questions a message should carryWhether none, one or several changes who answers, and whether it matters where in the message the question sits.
- Length, by temperatureWhy the right length for a cold message and a warm one are not the same number, and where each one stops working.
- The closing askHow the shape of your close changes the quality of the replies it earns, not only how many of them arrive.
- Where replies and meetings part companyThe campaign metric that tracked revenue in this data, and the louder one that did not.
- The follow-up windowWhat a single follow-up is worth, and the point at which sending it stops helping.
- Warm against cold, judged properlyHow conversations that began warm compare per message and in total, and what each lane is actually for.
- Anatomy of a winning messageThe six-part skeleton shared by the 85 conversations that reached meeting intent or better, in order.
- Score your last tenA seven-line scorecard for grading your team's recent outreach against the data, with banded results.
The dataset
2,279 cold LinkedIn conversations, started by senior people at professional and financial services firms across Africa, Europe and the Middle East. 1,082 of them had been live for at least seven days when the numbers were run, and every rate in the study comes from that mature set.
Every reply was read and classified by a human across five tiers, from no reply through to a meeting held, rather than scored by a sentiment model. Each conversation is joined to what happened afterwards, which is why the study can report meetings instead of stopping at replies.
It is observational data, not a controlled experiment, so correlation is not causation. Every figure quotes its sample size, anything under roughly n=20 is flagged as directional, and every campaign appears as a sector descriptor so no client, sender or recipient is identifiable.
Questions
Common questions
Does LinkedIn reply rate predict meetings?
Not on its own. In the study, a campaign that replied at 13.3% and one that replied at 35.9% both reached about 9 meeting conversations per 100 messages sent, because far more of the lower campaign's replies reached meeting intent (72.7% against 26%). Judge outreach on the replies that reach meeting intent, not on reply rate.
How long should a cold LinkedIn message be?
Longer than most advice suggests. In the study, cold messages under 350 characters engaged 7.1% of recipients (n=28), while messages of 650 to 800 characters engaged 15.6% (n=45), with no falloff below 800. A cold reader needs to know who you are, why them and why now. Warm messages are the opposite: shorter wins.
When should you follow up on a LinkedIn message that gets no reply?
After 3 to 5 days. In the study, one follow-up sent 3 to 5 days after the first message revived 14.6% of silent threads (n=355), while waiting 6 to 9 days nearly halved that, to 8.2% (n=73). Keep the follow-up short, under 200 characters.
What do I get, and what does it cost?
The full study, free. Five pieces of standard outreach folklore tested against booked meetings, the warm against cold comparison, the anatomy of the messages that produced meetings, and a scorecard for grading your team's last ten. The link opens the moment you submit the form, and a copy goes to your inbox.
How reliable are these numbers?
2,279 cold LinkedIn conversations started by senior people, of which 1,082 had been live at least seven days when the numbers were run. Every rate comes from that mature set and every reply was classified by a human, not a sentiment model. It is our own client programmes, anonymised to sector descriptors so no client is identifiable. Observational data, not a controlled experiment, and every figure in the study quotes its sample size.
What happens to my email address?
It is added to the AscendAI list so we can send you the study and, from time to time, the newsletter this research came out of. You can unsubscribe from any email in one click. We do not sell or share your details. See our privacy policy for the full position.