Cold LinkedIn benchmarks 2026

Accept rate, reply rate and volume from an outbound engine that runs every working day. Every rate on this page carries the count it came from and the days it was measured over. Last re-measured 30 August 2026.

The two rates we measure ourselves

RateOur readingCountedWindow
Invitation accept rate21.5%158 accepts on 734 invitationsInvitations sent 10 to 23 August 2026, each with at least 7 days to be answered
Reply rate after a first direct message20.3%38 replies from 187 peopleFirst message sent 30 July to 22 August 2026

Both were recomputed from the engine database on 30 August 2026 across 14 active sending profiles. The published cold-outreach reply benchmark we hold ourselves against is 14.1%.

One accept rate, one definition

Accept rate has two possible numerators, and they answer two different questions. Our stack sees both, which is how this page came to exist.

The first numerator arrives from our LinkedIn provider as a webhook: every time a sending profile gains a first-degree connection, we hear about it. Between 10 and 30 August 2026 that webhook fired 506 times. Of those 506 new connections, 254 are people we had invited. The other half arrived by other routes, most often a stranger inviting the profile after reading a post. A rate built on the raw webhook count describes how popular the profile is.

The second numerator is a poll. For each person we invited, the stack asks the provider whether the connection now exists. The poll works through the newest pending invitations first, so it reports late on older ones.

The number published above is the union of the two, restricted to people we invited ourselves: 158 accepts on 734 invitations. Every accept rate on this site now uses that definition, and every one of them carries its two counts.

A rate is a claim. A rate with its numerator, its denominator and its window is a measurement.

Volume behind the rates

  • 1,562 invitations sent between 27 July and 28 August 2026.
  • 576 direct messages, 71 InMails, 316 post likes sent between 23 July and 30 August 2026.
  • 14 active sending profiles, 6 of them premium, on 30 August 2026.
  • 3,520 people in the engine database across 7 workspaces.
  • In the week of 24 August, 14 profiles sent 458 invitations, an average of 32.7 per profile. The busiest single profile in a single calendar week sent 80.

How long an accept takes, and why it decides your window

Across 226 accepts with a recorded invitation time, the median came back in 2.4 hours, the 90th percentile in 1.9 days and the 95th in 4.9 days. A five-day window therefore catches about 95% of the accepts an invitation will ever get.

That single fact governs every comparison. On 6 August our own weekly review reported that accept rate had fallen by two thirds. On 17 August the same cohort, now matured, read normal. We had compared June invitations that had a month to be answered with August invitations that had a few hours. Compare cohorts by send period, hold every invitation younger than five days out of the verdict, and the phantom collapse goes away.

The first twenty invitations from a new profile read low

Grouping every invitation by its position in that profile's own sending history:

Position in the profile's historyInvitationsAcceptsAccept rate
First 20 from that profile2984615.4%
Invitation 21 and later1,27826821.0%

Same signals, same copy, often the same week. So a profile's first twenty touches are held out of every verdict we make about copy, about a signal, and about the profile itself. The rule earned its keep on a sender we came close to switching off in week one.

Where meetings are counted

Accept and reply are measured inside the engine, which is why they appear here with denominators. Booked calls are counted in the calendar and reported to each client in their Friday email, so a meeting rate joins this page when the calendar side of the count is as reproducible as these two.

Three rules for reading your own numbers

  1. Wait for the sample. Under 30 sends or under 3 replies in a slice, the gap between two variants is noise.
  2. One change a week. Two changes in the same week make the week unreadable.
  3. Carry the denominator everywhere. In the dashboard, in the report, and in the sentence you say out loud on a call.

Questions

What reply rate should I expect from cold LinkedIn in 2026?

The published benchmark for cold LinkedIn sits at 14.1%. Our own reading is 20.3%, from 38 replies among 187 people whose first message went out between 30 July and 22 August 2026. Ask for the denominator behind any figure you are quoted, including this one.

What is a good LinkedIn connection accept rate?

Above 20% on a cold audience means the profile is doing its job. Our reading is 21.5%, from 158 accepts on 734 invitations sent between 10 and 23 August 2026. Below 20%, rewriting the message moves very little; the profile and the audience match move everything.

How many invitations can one profile send in a week?

Our busiest profile sent 80 in a calendar week, and the fleet averaged 32.7. We hold a rolling ceiling of 100 invitations per seven days per profile, which sits under the platform allowance and well under the point where refusals start. The full measurement is on the LinkedIn limits page.

How long should I wait before judging a batch?

Five days. Across 226 accepts the median arrived in 2.4 hours and the 95th percentile in 4.9 days, so a five-day window catches roughly 95% of them. Judging a batch on day one compares a mature cohort with a fresh one.

How often is this page updated?

Every rate here is recomputed from the engine database, and the page carries the date of its last recompute at the top. The numbers move when the engine moves, and the window moves with them.

Where do these numbers come from?

From our own sending: 14 LinkedIn profiles running daily for our own pipeline and for client accounts, with every send, accept and reply written to one database. The same figures go into each client's Friday email.

Read next

LinkedIn invitation limits in 2026What the platform refuses, measured on 1,636 invitation attempts, and the pacing that took the refusal rate from 6.5% to 0.4%.Read it
What an AI SDR does, and what a week costsOne week of one sending profile, with the volume and the two rates above turned into unit economics.Read it
How operated outreach works, week by weekWeeks one and two, the first sending week, and what lands on your desk after that.Read it

Want this running on your own pipeline?

Book a 30-minute call. Tell me who you need to reach and I will say plainly whether this is a fit.