LinkedIn outreach benchmarks, from our campaigns

How we measured

Every figure is counted from the outreach event log the client dashboards read: one row per invitation, acceptance, message, reply and booked call, each tied to a LinkedIn account and a campaign. The sample is every campaign that reached 100 invitations inside the window. That floor removes noise from two-week tests and it also removes campaigns stopped early because they were failing, so the pooled rates lean favourable by an amount we have not yet measured. The quarterly editions will state it.

Counts are of people, never of rows. One person invited twice is one invitation; one person who sends three messages back is one reply. Only events attributed to a campaign count: connections the account owner makes by hand and replies from people we never contacted are both logged and both excluded. The definitions below are transcribed from the code that produces them.

Accept rate

Distinct people with an accepted-connection event, divided by distinct people with an invitation-sent event, inside the window. People, never rows: the same person invited twice is one invitation. Capped at 100%, because invitations sent before a window can be accepted inside it. Only events attributed to a campaign. Connections the account owner makes by hand are logged by the same webhook and are excluded, because they would inflate acceptance. An account ranks only once it has 20 invitations in 30 days.

Reply rate

Distinct people who replied, divided by distinct people who were sent a message or an InMail, inside the window. The reply webhook writes one row per inbound message, about 2.3 per human and in one case 290, so counting rows would roughly double every reply figure. Only replies attributed to a campaign. A reply with no campaign is someone we never messaged, usually organic inbound or another vendor pitching the client, and folding those in is what once produced reply rates above 100%.

Positive reply rate

People whose conversation the classifier judged as interested, divided by people messaged. The verdict covers the whole thread to date and is sticky: someone neutral today becomes positive the moment they show interest and does not fall back. A bare acknowledgement ("thanks, got it") is never positive. Replies the classifier failed on are removed from both numerator and denominator. The neutral bucket exists since 22 July 2026 and campaign attribution at insert time since 5 August 2026; any series crossing those dates is marked rather than read as one trend. Before the neutral bucket, 77 to 92% of replies scored positive; after it, 27 to 59%.

Follow-up reply rate

Each send is stamped with its step number. A reply is credited to the last message sent before it, so a reply after the second follow-up counts for that follow-up and not for the opener. Rate per step is repliers divided by the distinct people who received that step. A rate is shown only once 20 people have received the step; below that the dashboard shows counts. Replies that predate step stamping are reported as unattributed rather than dropped silently.

Meeting rate

Distinct people with a booked-call event, divided by invitations, by acceptances or by positive replies, as labelled. Bookings are read from the connected calendar (Cal.com, Calendly, iClosed) and from the conversation when a prospect shares an email after a time is offered, then deduplicated per person across both. Internal and test bookings are excluded. Clicks on a tracked link are never a conversion; they are reported as clicks.

The funnel, measured

35.0% accepted, 7.9% of those replied with interest, 12.8% of those booked. Pooled totals for the 90 days to 7 August 2026; a campaign with 2,000 invitations weighs twenty times one with 100.

Read end to end: 0.35 booked calls per 100 invitations, or 1.0 per 100 acceptances. Positive replies per 100 invitations: 2.8.

What moved the numbers

Same sample, same date, split by what the opener did and who received it. Openers of 300 to 500 characters produced 42 of the 49 booked calls. The list moved positive replies by a factor of almost three with the words held constant.

Accept rate

Median and quartiles across campaigns, per 100 invitations. A campaign counts once whatever its size. These cells are filled from a full calendar quarter with the definitions above and are marked until that has been done.

Account age matters because a newly connected account sends at 25% of its budget for days 1 to 3, 50% to day 7 and 75% to day 14 before running at full rate, and because LinkedIn names many pending invitations as a restriction trigger. Pending invitations are withdrawn after seven days by default.

First-message reply rate

Median and quartiles across campaigns, per 100 people messaged. A campaign counts once whatever its size. These cells are filled from a full calendar quarter with the definitions above and are marked until that has been done.

Follow-up reply rate (steps 2 to 4)

Median and quartiles across campaigns, per 100 people who received the step. A campaign counts once whatever its size. These cells are filled from a full calendar quarter with the definitions above and are marked until that has been done.

Positive reply rate

Median and quartiles across campaigns, per 100 people messaged. A campaign counts once whatever its size. These cells are filled from a full calendar quarter with the definitions above and are marked until that has been done.

Meeting rate

Median and quartiles across campaigns, per 100 invitations. A campaign counts once whatever its size. These cells are filled from a full calendar quarter with the definitions above and are marked until that has been done.

What LinkedIn documents, what vendors publish, what we observe

LinkedIn publishes no acceptance or reply benchmark at all. The figures in circulation come from automation vendors, each measured on its own customers with a method we could not check. Ours are labelled the same way; the difference is that the period, the sample and the exclusions are on this page.

What LinkedIn does document about invitations is on our limits page, and what it names as causes of restriction is on the account safety page, each with the help page it came from.

Changes since last quarter

This is the first edition, so there is no previous quarter to compare. Each later edition replaces the figures above in place and appends the ones it replaced here, so a number quoted from an older edition can still be traced.

Cite as

Gelee, LinkedIn Outreach Benchmarks, 2026 edition 1, observed across 42 client campaigns and 13,841 invitations in the 90 days to 7 August 2026. Updated in place each quarter; previous figures are kept in the change log on the page. https://www.gelee.ai/research/linkedin-outreach-benchmarks

Keep the period and the sample with the figure. A rate quoted on its own becomes one more unsourced benchmark, which is what this page exists to replace.

Common questions

Keep reading