Key finding
Across 43,084 accepted LinkedIn connections with a recorded message step, 14.7% of all replies arrived after a follow-up message was sent, and among accepted prospects who stayed silent long enough to receive a follow-up, 13.9% then replied.
What we measured
Follow-up advice is a genre. Five-touch sequences, seven-touch sequences, "the fortune is in the follow-up". None of it comes with a denominator.
Reachium records which message step in a sequence was the last one sent to each accepted connection, and whether that person ever replied. Follow-up messages carry an explicit marker in the step record. That makes two honest questions answerable at scale: how much of the total reply volume arrived after a follow-up went out, and how often a follow-up turned a silent prospect into a replier.
Methodology
Dataset window. Connection requests sent between 8 January 2025 and 27 July 2026, pulled 27 August 2026, matured at least 30 days: 196,696 requests, 53,050 accepted.
Coverage, measured. 43,084 of the accepted connections (81.21%) carry a record of the last message step sent; the remaining 9,966 carry no step record, and 14.78% of them replied. All step-level figures use the 43,084.
How the split works. A reply ends the sequence, so a prospect whose last recorded step is the first message either replied to it or never progressed further, and a prospect whose last recorded step is a follow-up is someone who did not reply to the first message and was sent at least one more. That second group is the clean one: 13.9% of them replying is a true follow-up conversion among the silent.
The trap this design avoids. The first-message group replied at 32.64%, and putting that next to 13.9% as a message-quality comparison would be wrong: everyone who was ever going to reply quickly is in the first group by construction. The rates answer different questions, and this study only compares like with like.
The findings
| Last message step sent | Accepted prospects | Replied | Reply rate |
|---|---|---|---|
| First message only | 30,647 | 10,004 | 32.64% (see methodology) |
| A follow-up | 12,437 | 1,723 | 13.85% |
Of the 11,727 replies with a recorded step, 1,723 (14.69%) arrived after a follow-up was sent. Put the other way: a sequence that stopped after one message would have kept 85.3% of these replies and forfeited one in seven.
One structural fact from the same dataset reframes the follow-up debate entirely: every campaign in this cohort that ever sent a follow-up message after acceptance sent exactly one. The 5-touch and 7-touch sequences the genre recommends are essentially absent from 196,696 real requests. Whatever the right number of follow-ups is, the operating reality here is binary: one follow-up, or none.
What this means
Follow-ups earn their keep, modestly and measurably. A 13.85% conversion of silent prospects is not the "most deals close on touch five" story; it is roughly one extra reply for every seven silent prospects nudged. On the volumes in this dataset that is 1,723 conversations that would not otherwise exist.
The absent deep sequences cut both ways. This data cannot say whether a third or fifth touch works, because almost nobody sent one; anyone claiming those numbers from their own data should be asked for a table like the one above. What the data does say is that teams sending at real volume converged on a single follow-up, and that single follow-up produced 14.69% of their reply volume.
Timing context from the same platform: replies cluster fast after acceptance, with a median of 31 hours, and two thirds of post-acceptance bookings land within a week. The silent-prospect pool the follow-up addresses forms quickly.
In practice
An SDR whose accepted connection has not replied is looking at the 12,437-person group in this table. Sending one follow-up moved 13.85% of that group to a reply. The data measures one follow-up because that is what the dataset contains; it is silent on the second and beyond.
A team lead auditing sequence templates can put a number on the stakes: in this data, dropping the follow-up step entirely would have cost 14.69% of step-recorded replies. A team getting 100 replies a month from outreach is deciding whether to hand back roughly 15 of them.
Limitations
Observational, not randomized. No prospect was randomly assigned to receive or not receive a follow-up. The follow-up group is defined by their own silence, and campaigns that send follow-ups may differ from campaigns that do not in copy, targeting, and everything else.
"After" is not "because". A reply that arrived after the follow-up is attributed to the follow-up window, but some of those prospects might have replied anyway. 13.85% is a ceiling on the follow-up's causal effect in this data, not a point estimate of it.
Coverage is 81.21%. The 9,966 accepted connections without a step record replied at 14.78% overall, well below the covered group's 27.22% (11,727 of 43,084, derived), so the uncovered rows are not missing at random; the cause of the missing records is not itself recorded.
Reconciliation. The matured cohort has 14,421 replies in total; 11,727 carry an acceptance-linked step record, 1,473 sit in the no-step-record group, and 1,221 replies have no recorded acceptance at all, a class the timing study also measured and attributed to pre-existing conversations.
One platform's sequencing. All sequences ran through one platform whose users converged on one follow-up. Concentration as elsewhere: the largest workspace holds 22.88% of matured requests.