// LINKEDIN OUTREACH AUTOMATION
Cold vs ICP-Scored vs Signal-Based LinkedIn Outreach
The stronger your reason to contact someone, the better every stage of LinkedIn outreach converts. We compared 10,220 connection requests across three campaigns with a cold-list benchmark.
The data
| Level | Requests | Accepted | Replied | Interested |
|---|---|---|---|---|
| 1. Unscored cold list (benchmark) | 10 to 20% | 3 to 8% of accepted | ||
| 2. AI-scored Sales Navigator list | 7,152 | 33.4% | 19.3% of accepted | 24.3% of replies |
| 3. Event attendees | 569 | 52.5% | 43.8% of accepted | 22.1% of replies |
| 4. Post engagers | 2,499 | 81.1% | 46.5% of accepted | 40.6% of replies |
Level 1: cold lists
Filters by title and company size say nothing about fit or timing. Most requests are ignored and burn your daily sending capacity.
Level 2: AI-scored lists
Same search, but every lead is scored against the ICP before anyone writes. In a 100-lead sample, 55% were filtered out before messaging. Scoring is the cheapest step in the pipeline and decides where every expensive step goes.
Level 3: event outreach
An event gives both sides a reason to talk and a natural deadline. People who post that they are attending are already thinking about who to meet.
Level 4: post engagers
They showed interest in the problem in public, days ago. Highest acceptance, highest share of interested replies.
What this means for your outbound
- The fastest win: score before you send.
- Then add signals: post engagement, events, hiring announcements.
- Higher acceptance also means fewer pending invitations, which keeps your account healthy.
On your current stack
The usual stack gets you to level 1 easily and stops there: Apollo or Sales Navigator for the list, a sequencer to send it. Level 2 means paying Clay credits to enrich and score every row before you know whether it fits. Levels 3 and 4 need event and engagement signals that neither the database nor the sequencer can find, so they mean extra scrapers, manual LinkedIn scrolling and more exports between tools.
AtomSell runs all four levels from the same building blocks and the same table, so moving your outreach up a level is a new prompt, not a new tool.
Methodology
Three separate campaigns with different offers and audiences, so compare acceptance first. Replies are measured against accepted connections, interested against replies. The event campaign ran for a client and is anonymized.
Run it yourself
Start a 3-day free trial with full access at app.atomsell.ai/register, connect AtomSell to Claude, ChatGPT or Cursor, and run the playbook on your own buyers.