TL;DR: After a LinkedIn connection is accepted, the first message got the most replies at 25 to 49 words (20.06%). Replies fell as messages got longer, to 11.78% at 75 to 99 words. This is the first direct message, not the connection note.
Our analysis of 355 distinct first messages sent to 58,207 accepted LinkedIn connections found that first messages of 25 to 49 words drew a 20.06% reply rate, against 14.47% at 50 to 74 words and 11.78% at 75 to 99 words, with a campaign-weighted measure showing the same shape. Our analysis
Key Takeaways
- First messages of 25-49 words drew a 20.06% reply rate, the highest of any band with enough messages to read.
- Replies fell with length: 14.47% at 50-74 words and 11.78% at 75-99 words.
- Very short messages (under 25 words) did not win: 16.90%, below the 25-49 band.
- The median first message was 42 words, inside the best-performing band.
- This measures the first direct message after acceptance. It says nothing about connection-note length.
What did we measure?
"How long should a LinkedIn message be?" gets answered with confident rules that rarely come with a reply rate attached. This study measures one specific message: the first direct message a person receives after accepting a connection request. It is not the note attached to the request, which is a different message with its own study (Do Connection Notes Work?).
Reachium's outreach sequences store the text of each step. That makes it possible to count the words in the first message of each sequence and compare reply rates across length bands, across 355 distinct first-message texts.
How was it measured?
Dataset window. Connection requests sent between 8 January 2025 and 16 October 2026, pulled 16 November 2026, matured at least 30 days. InMail sends are excluded.
The first DM. The first message node inside the "connection accepted" branch of an outreach sequence, read as stored at the pull.
Denominator. Matured, accepted requests that were sent the first DM. Outcome: reply to the first DM, meaning the person replied while the first DM was still the last message sent. Sequences stop on a reply, so this attributes the reply to the first message rather than a follow-up. "Replied at any point" is shown alongside. Our analysis
Exclusions, in order. Accepted requests whose first DM was never sent (10,402); requests whose sent step is not in the campaign's current sequence because the copy was edited after sending (55); campaigns whose first DM carried more than one variant text; and 63 requests whose first DM would have rendered a blank company or title token. That leaves a clean sample of 58,207 accepted people and 355 distinct first-message texts, with an overall reply to the first DM of 17.05% (26.26% replied at any point). Our analysis
Length. Word count of the first DM, with a counting as one word.
The effective sample is messages, not people. A band with thousands of people but only a few distinct messages tells you about those messages, not about length. Every row reports distinct messages and a campaign-weighted reply rate (campaigns with 50 or more recipients, each weighted equally). A band with fewer than 10 distinct messages is labelled too few messages to read. The largest single workspace's share of each band is shown as a concentration check.
What did we find?
The curve rises from very short messages to a peak at 25-49 words, then falls steadily.
| First DM length (words) | People who got the first DM | Reply to first DM | Replied at any point | Distinct messages | Campaign-weighted reply | Largest workspace share |
|---|---|---|---|---|---|---|
| Under 25 | 3,095 | 16.90% | 26.72% | 33 | 15.44% | 25.75% |
| 25-49 | 28,219 | 20.06% | 29.77% | 181 | 19.87% | 15.20% |
| 50-74 | 19,947 | 14.47% | 22.93% | 102 | 14.87% | 24.78% |
| 75-99 | 6,172 | 11.78% | 20.54% | 32 | 11.06% | 41.12% |
| 100+ (too few messages to read) | 774 | 16.28% (derived) | 7 |
The two measures agree. Pooled across people, 25-49 words drew 20.06%; weighting every campaign equally, 19.87%. At 75-99 words the figures are 11.78% and 11.06%. The pooled gap is 8.28 points and the weighted gap 8.81 points (both derived). A shape that survives equal weighting is not one large campaign pulling the average. Equal weighting does not remove workspace concentration, though: the 75-99 band is 41.12% one workspace. Our analysis
Most copy already sits in the best band. The median first message was 42 words. The 25-49 band holds 181 of the 355 distinct messages and 48.48% of recipients (derived). Our analysis
The 100+ row is not a finding. The 100-149 and 150+ bands have 6 and 1 distinct messages. Merged, that is 7 messages and 774 people, too few messages to read. Its 16.28% (derived) reflects a handful of specific texts, not long messages in general.
Concentration. The 75-99 band is the least spread: the largest workspace sent 41.12% of it. The 25-49 band is the most spread, at 15.20%.
What does this mean for cold LinkedIn messages?
On this data, the first message that works best is short but not minimal: a short paragraph rather than a single line or a full pitch. Each step past 49 words came with a lower reply rate, and the 25-49 band replied 1.70 times as often as the 75-99 band (derived from 20.06% and 11.78%).
Why very short messages trail the peak is not measured. One plausible reading is that a one-line message gives the recipient too little to answer. Another is that teams who write one-liners differ in other ways. Length travels with other choices: what the message asks, whether it carries tokens, and who it goes to. The companion studies on personalisation tokens and how the message ends test two of those choices on the same 58,207 people, and the ending study shows the length effect is not the whole story: a closing question did worse than a statement at 75-99 words.
The first message is also not the last. Follow-ups add a modest share of replies, and reply rates among accepted connections have drifted down across the platform over time. Teams rewriting sequence copy can compare setups on Reachium's pricing page.
How do you use this in practice?
An SDR whose first message after acceptance runs 80 words is in the 75-99 band, which replied at 11.78% here. The same SDR's peers writing 25-49 words saw 20.06%. The data does not say that cutting the message will move that SDR's rate, because length is chosen alongside everything else in the message. Our analysis
A founder who writes one-line openers to seem casual is in the under-25 band: 16.90%, below the 25-49 band but above the 50-74 and 75-99 bands. Our analysis
A team lead auditing templates can count words and place each template in the table. Templates in the 50-74 band, where 102 of the distinct messages sit, replied at 14.47%, 5.59 points below the 25-49 band (derived). Our analysis
The Ancaster Capital case study shows one team's outreach numbers on the same platform.
What are the limits of this study?
Observational, not an experiment. Workspaces chose their own copy. Nobody was assigned a message length, so other copy choices, targeting and offer ride along with length. A/B variants are excluded, not analysed.
Messages, not people, are the sample. 355 distinct texts. A band's reply rate is driven by the specific messages in it, which is why distinct-message counts and campaign-weighted rates are shown.
Not the connection note. This is the first DM after acceptance. Nothing here applies to the note on a connection request.
Concentration. The 75-99 band is 41.12% one workspace. The 100+ bands are too few messages to read.
Copy as stored. Sequences are read as stored at the pull. 55 requests whose copy had been edited since sending are excluded, and edits that kept the same step are not detectable.
Reply is not interest. A reply can be a no. This study counts replies, not positive ones. Teams that want help writing and testing first messages can talk to the agency team.
FAQ
What is the best LinkedIn message length?
For the first message after a connection is accepted, 25 to 49 words drew the highest reply rate in our analysis: 20.06%, against 11.78% at 75 to 99 words.
How long should a LinkedIn message be?
On this data, a short paragraph. The median first message was 42 words, and the 25-49 band beat both shorter and longer messages.
Are short or long cold messages better on LinkedIn?
Shorter, up to a point. Replies fell from 20.06% at 25-49 words to 11.78% at 75-99, but under 25 words drew 16.90%, below the peak.
Does this apply to LinkedIn connection request notes?
No. This study measures the first direct message after acceptance only. Connection notes are a different message, covered in Do Connection Notes Work?