In short
- 01"Across 48 workspaces and 42,236 connections, the mean response rate was 15.7% and the median was 4.5%."
- 02"The bottom quartile sat at 1.6% or below; the top decile above 49.5%. That is roughly a hundredfold spread inside one platform."
- 03"68.3% of all responses were explicit disinterest, which most benchmarks count as a reply."
- 04"Every published benchmark we could check quotes a mean, and a mean is the wrong statistic for a distribution this skewed."
Every LinkedIn outreach benchmark you can find quotes an average. Belkins publishes 7.9% and 12.2%. Expandi publishes 10.3%. LeadRiver publishes 30 to 37%. All of those are real, carefully measured numbers, and we cite them across this site.
They are also, all of them, means. And a mean only describes a population well when that population is not badly skewed. So we checked ours.
What we measured, and what we did not
We looked at every workspace on our own platform that has put at least 100 contacts into outreach, computed each one's response rate separately, and then looked at the distribution rather than collapsing it into a single figure. That is 48 workspaces and 42,236 connections, from 2026-03-12 to 2026-08-03.
This is not a claim that our platform performs better than anyone else's. It is a claim about shape. The interesting result is not the level of any number here, it is how far apart the teams are, and the fact that reporting a single average hides that completely.
The distribution is the finding
The mean response rate across these workspaces is 15.7%. The median is 4.5%. When the mean is three and a half times the median, the average is being carried by a small number of outliers and describes almost nobody.
Here is the full spread, each figure the response rate of a workspace at that percentile:
The gap between the 10th and 90th percentile is more than a hundredfold, inside one product, on one platform, over the same period. A quarter of these teams are getting a response from fewer than one in sixty connections. A tenth are getting one from every other connection.
Both of those are true at once, and no single average can express it. If you have ever read a benchmark, compared it to your own numbers, and concluded something was wrong with you, this is worth sitting with: the benchmark was probably describing the top of this distribution.
Most responses are a no
The second number worth stating plainly: 68.3% of all responses in this sample were explicit disinterest.
Most benchmark methodology counts any reply as a reply. That is defensible as a measure of whether a message landed, and it is misleading as a measure of whether outreach is working. A 15% reply rate where two thirds of replies are "no thanks" is a 5% conversation rate, and the difference is the whole business case.
We do not have a comparable positive-reply figure from the published studies, because none of the ones we checked separate the two. That is not a criticism of their work; it is a gap in what the category measures.
Why we are not publishing the number you would expect
We looked at whether contacts sourced from a real engagement signal outperform the rest, because that is our own product thesis and it would have been a convenient thing to prove.
The cut came out the wrong way: 7.9% for signal-sourced against 13.6% for everyone else, and it is statistically significant at p=0.008. We are not publishing it as a finding, because it is almost certainly an artefact. Engagement signals only started being recorded on 2026-05-16, while outreach in this dataset starts on 2026-03-12, so signal-sourced contacts are structurally younger and have had less time to reply. That is a confound, not a result, and it runs in the direction that would flatter us if we reversed it.
The honest position is that we cannot yet answer that question from this data, and we would rather say so than run the comparison on a denominator we know is broken.
What this means for planning
- Plan against the median, not the average. If you are modelling a pipeline on a published benchmark, you are modelling the 75th percentile. Our ROI calculator defaults to the published cold and warm figures precisely so the output is a floor rather than a hope, and even that is generous against this distribution.
- Separate positive replies from all replies before you decide anything. Two thirds of the responses here were a no. A metric that treats those identically will tell you outreach is working right up until you look at the calendar.
- The spread is the opportunity. A hundredfold gap between teams using the same tool means the tool is not the variable. Targeting and timing are, which is the argument in signal-based selling and why warm outreach outperforms cold by a wider margin than any feature.
- Compare like with like. Before adopting any benchmark, check what it divided by. We have written up what the published acceptance-rate studies actually measure, and the denominators differ more than the headline numbers do.
Method and limits
Stated plainly, because a study that hides its method is a marketing asset rather than research.
- Sample. Workspaces on BeReach with at least 100 contacts in an outreach state, between 2026-03-12 and 2026-08-03. 48 workspaces, 42,236 connections, 5,380 responses.
- Response definition. A contact whose current state is replied, in conversation, meeting booked, converted, or not interested. Contact state is current rather than cumulative, so a contact that replied and later went quiet may be counted in a later state.
- The 100-contact floor excludes 87 smaller workspaces. Without it the distribution is noisier and the median lower still, so the floor is the conservative choice.
- Self-selection. These are teams that chose this product, so they are not a random sample of LinkedIn users. The spread within them is the finding, not the level.
- Survivorship. Workspaces that stopped entirely are included if they crossed the floor, but a team that gave up after 40 contacts is not.
- What we did not measure. Message content, industry, seniority and offer are all uncontrolled. This says nothing about which of those drives the spread.
Frequently asked questions
What is a good LinkedIn response rate in 2026?
Published benchmarks put it between 7.9% and 12.2% (Belkins, 2026, 15.1 million touchpoints) with a platform average of 10.3% (Expandi, 2026). In our own data the median team achieved 4.5%, so those published figures describe roughly the 75th percentile rather than a typical team. Plan against the lower number.
Why is the median so much lower than the average?
Because the distribution is heavily skewed. A handful of workspaces in our sample respond above 49.5% while a quarter sit at or below 1.6%. A mean is pulled toward the outliers, so it describes the top of the range rather than the middle. For skewed data the median is the honest summary.
Does counting negative replies change the numbers much?
Substantially. 68.3% of the responses in this sample were explicit disinterest. Most published benchmarks count any reply, so a headline reply rate is roughly three times the rate of replies that could become a conversation. Check what a benchmark counted before comparing yourself to it.
Does this prove signal-based outreach works better?
No, and we are not claiming it does. That cut came out against us in this data, and we believe that is a timing artefact rather than a result, because engagement signals were only recorded from 2026-05-16 while outreach starts in March. The dataset cannot answer the question cleanly yet.
Can I see the underlying data?
Not the raw records: they are customer prospect data and we do not publish or sell it. The aggregates, the sample definition, the date range and the limits are all stated above so the method can be checked and the arithmetic reproduced from the totals.
Every viral post is 100+ warm conversations waiting.
Tell your agent who you want to reach. It finds leads, qualifies them, sends personalized outreach, and follows up.
Reading this in an AI assistant? Hand it the page and let it summarise, so you can ask follow-up questions against the whole argument rather than the part you have read so far.
