// BENCHMARKS AND DATA

Cold vs ICP-Scored vs Signal-Based LinkedIn Outreach: Real Data

4 min readBenchmarks and data

`The stronger the reason you have to contact someone, the better every stage of LinkedIn outreach converts. Across 10,220 connection requests in three campaigns, acceptance rose from 33% for AI-scored Sales Navigator leads to 53% for event outreach and 81% for people who engaged with posts in our niche. Unscored cold lists typically get 10 to 20%.

The data

Lead sourceRequestsAcceptance (of requests)Replies (of accepted)Interested (of replies)
Unscored cold list (benchmark)10 to 20%3 to 8%
AI-scored Sales Navigator list7,15233.4%19.3%24.3%
Event outreach56952.5%43.8%22.1%
LinkedIn post engagers2,49981.1%46.5%40.6%

Level 1: cold lists

A cold list is a Sales Navigator or LinkedIn search sent to everyone who matches the filters. Filters describe a job title and a company size, but they say nothing about whether the person is a real fit or has any reason to talk to you now. Most requests go to people who ignore them, and every ignored request is sending capacity you did not use on someone better.

Level 2: AI-scored lists (33% acceptance)

Scoring changes who you contact, not how many searches you run. Before sending anything, each lead from a Sales Navigator search gets an ICP score from 1 to 10 with a reason, based on headline, role, company, location, industry, and summary.

In our ICP-scored flow, from June 1 to September 25, 2026, we sent 7,152 requests. 2,387 were accepted (33.4% of requests), 460 replied (19.3% of accepted connections), and 112 were interested (24.3% of replies).

In a 100-lead sample from the same setup, 55% of leads were filtered out before any message was written. Scoring one lead with Claude Haiku cost about $0.002 and took about a second. That is the cheapest step in the pipeline, and it decides where all the expensive steps go.

Level 3: event outreach (53% acceptance)

An event gives you a shared context and a deadline. People who post that they are going to a conference, or who engage with speakers' posts before it, are already thinking about the topic and about meeting people.

In a client campaign before a European tech conference, from August 12 to 31, 2026, two senders sent 569 requests. 299 were accepted (52.5% of requests), 131 replied (43.8% of accepted connections), and 29 were interested (22.1% of replies). Leads came from posts about the event, engagers on speakers' and organizers' posts, and the speakers themselves, scored by AI for real attendance intent and ICP fit.

Level 4: post engagers (81% acceptance)

People who react to or comment on posts in your niche have shown interest in the exact problem you solve, publicly and recently. From May 29 to September 25, 2026, we sent 2,499 requests to them. 2,026 accepted (81.1% of requests), 943 replied (46.5% of accepted connections), and 383 were interested (40.6% of replies). Full breakdown here.

What this means for your outbound

  1. Score before you send. Moving from a cold list to an AI-scored list is the fastest win, because it only changes one step in your pipeline.
  2. Add signals as you grow. Post engagement, events, and hiring posts are all signals you can pull automatically with the same tools.
  3. Higher acceptance protects your accounts. Fewer ignored requests means fewer pending invitations per new connection. The infrastructure behind Periodix has lost 0 LinkedIn accounts across 25,000 connected profiles in the last 24 months.

Methodology and limits

These are three separate campaigns, with different offers, audiences, and dates, not a controlled experiment. Acceptance is the most comparable metric across them. Reply rates are shown as a share of accepted connections to match how the benchmarks count them, and interested rates as a share of replies. "Interested" counts replies that showed real interest, excluding neutral replies, objections, and requests for information. The post engager and ICP-scored campaigns were our own; the event campaign was run for a client and is shown anonymized. In every campaign, leads were sourced with Periodix Actions and conversations were run by the Periodix AI agent.

Run each level yourself

Periodix Actions is GTM infrastructure for AI agents and automation builders: the same pipelines run as n8n workflows, through an MCP-connected agent like Claude, or over the REST API.

Each playbook has a ready n8n template and MCP prompts for Claude, Cursor, or any MCP client. Start a free 3-day trial.`

Frequently asked questions

›What is a good LinkedIn connection acceptance rate?

Unscored cold lists typically see 10 to 20%. In our campaigns, AI-scored Sales Navigator leads reached 33%, event outreach 53%, and LinkedIn post engagers 81% of requests.

›What is signal-based prospecting on LinkedIn?

Reaching out to people because of a recent action that shows interest, such as engaging with a post in your niche, announcing they will attend an event, or their company hiring for a relevant role, instead of contacting everyone who matches search filters.

›Does AI lead scoring improve LinkedIn outreach?

Yes. Scoring each Sales Navigator lead against your ICP before sending raised acceptance to 33% of requests and replies to 19.3% of accepted connections, compared with 10 to 20% acceptance and 3 to 8% replies for unscored cold lists.

›How much does AI lead scoring cost per lead?

About $0.002 per lead with Claude Haiku, in about a second. In a 100-lead sample, 55% of leads were filtered out before any message was written.

›Which LinkedIn signal converts best?

In our data, post engagers: 81% acceptance, 46.5% of accepted connections replied, and 40.6% of replies were interested.

›Are these results from a controlled experiment?

No. They come from three separate campaigns with different offers, audiences, and dates. Acceptance is the most comparable metric across them.

Related playbooks

Periodix Actions used

Keep reading

Olga Yasynska avatar

// AUTHOR

Olga Yasynska

Olga Yasynska is GTM and Chief Revenue Officer at Periodix. She has spent 20+ years in B2B sales and marketing and six years building Periodix, where she runs GTM and designs the playbooks behind Periodix Actions. She holds master's degrees in mathematics and psychology.

LinkedIn

// GET STARTED

Ship your first Periodix playbook

Get an API key, connect a profile, and run a production-ready playbook end-to-end in your stack.

Create your account