Key finding
Across 180,155 LinkedIn connection requests given at least 30 days to mature, 27.1% were accepted, 27.6% of accepted connections produced a reply, and 2.0% of accepted connections produced a booked meeting.
What we measured
Reachium is a LinkedIn outreach platform. It records every connection request its customers send, whether that request was accepted, whether the accepted connection produced a reply, and whether the conversation produced a booked meeting.
This study reports those four numbers across the full dataset. It covers 441,965 outreach sequences, of which 204,847 recorded a connection request actually sent. The window runs from 8 January 2025 to 6 August 2026.
Published benchmarks for LinkedIn outreach are unusually hard to trace. The figures that circulate tend to cite each other rather than a measurement, and they rarely state what they divide by. Everything below states its denominator.
Methodology
Dataset window. Every figure covers connection requests sent between 8 January 2025 and 6 August 2026, pulled on 6 August 2026.
Acceptance rate is accepted connection requests divided by connection requests sent.
Reply rate of accepted is connections that produced at least one inbound message divided by accepted connections. It is not a share of requests sent.
Reply rate of sent is the same numerator divided by connection requests sent. Both are reported because the two are routinely quoted interchangeably, and on this dataset they differ by a factor of 3.69 (derived from the two rates below).
Meeting rate is connections with a booked meeting divided by accepted connections.
Maturity. A connection request sent last week has had no chance to be accepted yet, and counting it as a failure understates recent performance. The headline figures therefore use only requests sent at least 30 days before the data was pulled on 6 August 2026. That leaves 180,155 of the 204,847 requests. The unrestricted figures are reported alongside so the size of the effect is visible rather than hidden.
A reply means a reply. The data records that an inbound message arrived, not whether it was positive. A reply saying "not interested" counts.
The findings
Matured cohort, 180,155 requests:
| Stage | Count | Rate |
|---|---|---|
| Sent | 180,155 | |
| Accepted | 48,836 | 27.11% of sent |
| Replied | 13,456 | 27.55% of accepted |
| Booked | 973 | 1.99% of accepted |
All requests, including those too recent to have matured:
| Stage | Count | Rate |
|---|---|---|
| Sent | 204,847 | |
| Accepted | 53,952 | 26.34% of sent |
| Replied | 14,624 | 27.11% of accepted |
| Booked | 999 | 1.85% of accepted |
The two tables both contain 27.11%, meaning different things: in the first it is acceptance as a share of requests sent, in the second it is replies as a share of accepted connections. That coincidence is a good illustration of this study's own point about denominators.
The unrestricted set runs lower at every stage. That is the maturity effect made visible rather than a different population: the extra 24,692 requests are the most recent ones (derived from the two totals above), and a share of them will still be accepted after this was measured.
Two derived figures, computed from the matured counts above: replies were 7.47% of all matured requests sent, and booked meetings were 0.54%.
What this means
The headline number most people quote is acceptance, and it is the least interesting of the four. Acceptance says someone clicked a button. The distance between a click and a conversation is where the funnel actually lives.
Read down the matured column as a hundred requests. Roughly 27 of them connect. Roughly 7 of the original hundred produce any reply at all, including the rejections. Under one produces a booked meeting.
That compounding is why "reply rate" is such a slippery benchmark. A vendor reporting 27.6% and a vendor reporting 7.5% can both be honest and both be describing this exact dataset. One is dividing by accepted connections, the other by requests sent. Any benchmark quoted without its denominator is not a benchmark.
In practice
A sales leader sizing a quarter off a target meeting count is working with the last row of that table. At the rates measured here, booked meetings were 1.99% of accepted connections and 0.54% of matured requests sent, so the request volume sitting behind a meeting target is governed by the 0.54% figure rather than by the 27.11% one that usually gets quoted. That is the arithmetic the data supports. It does not say the same ratio will hold for a different offer or a different audience, and the data cannot tell anyone what their own ratio is.
An SDR being measured on acceptance rate is being measured on the stage with the least signal in it. Acceptance was 27.11% across this dataset while replies were 27.55% of those acceptances, so two reps with identical acceptance rates can be running very different conversation volumes and the acceptance number will not show it.
A RevOps lead comparing a vendor's quoted reply rate against their own dashboard is the most common version of this problem. The same 13,456 replies in this dataset are 27.55% or 7.47% depending only on what sits under the line. Before two reply rates can be compared, both denominators have to be known, and one of them usually is not stated.
Limitations
This is one platform's customer base, not LinkedIn. Everyone in this dataset is running deliberate, sequenced outreach. It says nothing about connection requests in general.
The dataset is concentrated. The largest single customer workspace accounts for 21.97% of all requests sent, the largest five for 57.66%, and the largest ten for 76.19%. These figures are therefore weighted toward heavy senders, and a differently distributed dataset could produce different rates.
The meeting rate is a floor. A meeting is counted when it is booked through the platform's own booking step. Conversations that moved to email, to a calendar link pasted by hand, or to a call arranged in the LinkedIn thread are not counted. The true meeting rate is higher than 1.99% by an amount this data cannot measure.
The all-sent totals are a moving snapshot. 204,847 sent, 53,952 accepted and 14,624 replied describe the dataset as it stood on 6 August 2026. That cohort grows: new requests are sent daily, and acceptances keep landing on requests already sent. The matured cohort of 180,155 does not grow, because the 30 day cutoff closes it, which is why the headline figures use it. Anyone re-running the unrestricted numbers later will get larger totals.
Replies are counted, not classified. No sentiment is recorded, so the reply rate includes declines.
Targeting is not controlled for. Acceptance depends heavily on who is being contacted, and this study reports a single blended rate across every audience in the dataset.
Time is not controlled for. These are pooled figures across a 19 month window during which the underlying rates moved. The year over year movement is reported separately.