linkedin-audience-simulator

LinkedIn Audience Simulator Plugin (Plugin 3)

Predict which LinkedIn content will perform best with your audience using AI-powered segment analysis and calibrated scoring.

Part of the 3-plugin LinkedIn ecosystem:

1. Plugin 1 — LinkedIn Feed Tracker (Data collection)
2. Plugin 2 — Content Engine (Idea generation)
3. Plugin 3 — Audience Simulator (This plugin — Audience prediction)

What It Does

Plugin 3 analyzes your LinkedIn connections to predict audience segments and forecast post performance:

• Parses your network — Extracts roles, seniorities, industries from 2,000+ connection headlines
• Creates audience segments — Groups connections into Executives, Technical Leaders, Growth & Sales, General Audience
• Tracks engagement patterns — Maps which topics each segment engages with (AI/ML, Leadership, Growth, etc.)
• Scores draft posts — Analyzes a post and predicts reactions/comments per segment
• Predicts overall performance — Combines per-segment predictions into a 0-100 score
• Calibrates predictions — After 10+ posts, adjusts predictions based on actual performance
• Explains reasoning — Shows which segments will love your post and why

Installation

Prerequisites

• Plugin 1 (LinkedIn Feed Tracker) must be installed, configured, and running
• Cowork (the AI automation platform) — All skills run in Cowork
• Python 3.8+ with sqlite3

Quick Start

1. Install Plugin 3 from the Cowork Plugin Directory (or clone from GitHub)
2. Run /setup-audience-simulator in Cowork
3. Follow the guided setup (takes 1-2 minutes)
4. Run /score to score your first draft

Commands

Command	What it does
/setup-audience-simulator	First-run setup: parse connections, identify segments, initialize database
/score	Score a draft post and get per-segment predictions
/audience	View high-level overview of your audience segments, sizes, and engagement patterns
/deep-dive [segment]	Explore a specific segment in detail (roles, topics, past performance)
/accuracy	Review prediction calibration metrics and accuracy trends

How It Works

The Scoring Process

When you run /score with a draft post:

1. Topic Assignment → Regex + Claude analyze the draft and tag it with 1-3 topics (AI/ML, Leadership, Growth Hacking, etc.)
2. Hook Detection → Identify hook type (story, question, data-driven, contrarian, etc.)
3. Segment Analysis → For each of your 4 audience segments:
   • Measure topic affinity (does this segment care about this topic?)
   • Evaluate hook strength (does this hook appeal to this segment?)
   • Predict reactions, comments, reach
4. Aggregate Score → Combine per-segment predictions weighted by segment size and activity
5. Reasoning → Explain which segments will engage and why

The Data Flow

Plugin 1 (Feed Tracker)
├─ Collects network posts → feeds.db
├─ Collects your own posts → feeds.db
└─ Tracks connections → feeds.db

Plugin 3 (Audience Simulator)
├─ Parses connection headlines
├─ Creates audience segments (Executives, Technical Leaders, Growth & Sales, General)
├─ Tags posts with topics (AI/ML, Leadership, etc.)
├─ Calculates segment × topic affinities
└─ Stores in connection_profiles, audience_segments, segment_topic_affinity

User scores a draft → /score command
├─ Analyzes draft topics and hook
├─ Queries segment affinities
├─ Applies calibration (if 10+ predictions tracked)
└─ Returns 0-100 score + per-segment breakdown

User posts on LinkedIn → Plugin 1 tracks actual engagement

User reviews performance → /accuracy command
├─ Compares predicted vs actual
├─ Updates calibration coefficients
└─ Improves future predictions

The Scoring Model

Score formula:

overall_score = weighted_average(segment_scores)

where:
  segment_score = (hook_strength + topic_affinity) / 2
  weight = segment_size × activity_rate

Confidence levels:

• Low: < 10 predictions per segment/topic (uses heuristics)
• Medium: 10-50 predictions (uses calibration)
• High: 50+ predictions (reliable predictions)

Calibration:

• After 10 predictions per segment × topic, we fit a linear regression
• Formula: calibrated_score = (predicted_score × slope) + intercept
• Updates after each new post that you track results for

Example Workflow

1. Setup (First Time)

> /setup-audience-simulator

✓ Checking Plugin 1... Found 1,966 connections and 8,234 posts.
✓ Parsed 1,966 connection headlines.
✓ Identified 4 core audience segments:
  - Executives (150 people)
  - Technical Leaders (420 people)
  - Growth & Sales (380 people)
  - General Audience (1,016 people)
✓ Audience Simulator initialized!

2. Score a Draft

> /score
[Paste your draft...]