LinkedIn Audience Simulator — Cowork Plugin

Know who'll engage — before you hit post.

Your LinkedIn audience is segmented by role, seniority, industry, and behavior. The Audience Simulator predicts exactly which segments will react to your content, scores your drafts 0-100, and calibrates predictions as you post. Post with confidence.

3.1

How It Works

Segments, scoring, calibration

Scores, segments, insights

3.5

Talk to Claude

Example conversations

3.6

The Ecosystem

Plugins 1, 2, and 3

3.1

How It Works

Parse your network, predict engagement, calibrate over time

The Audience Simulator depends on data from Plugin 1 (LinkedIn Feed Tracker). It reads your connections and their engagement patterns, builds audience segments, then scores any draft post you give it.

The data flow

When you run setup:

Parses 2,000+ connections — Extracts roles, seniorities, industries from headlines
Creates 4 audience segments — Executives, Technical Leaders, Growth & Sales, General Audience
Tracks engagement patterns — Measures which topics each segment cares about
Stores locally — Everything lives in your feeds.db from Plugin 1

The audience segments

Executives

VPs, C-suite, directors. Engage with leadership, strategy, business outcomes.

Typically 5-10% of network

Technical Leaders

Engineers, architects, CTOs. Engage with technical depth, infrastructure, ML/AI.

Typically 20-30% of network

Growth & Sales

Sales, marketing, ops, growth PMs. Engage with metrics, growth tactics, user insights.

Typically 15-25% of network

General Audience

Broader connections. Engage with accessible, personal, human-centered content.

Typically 40-60% of network

How scoring works

When you ask Claude to score a draft post:

Topic detection — Claude tags your draft with 1-3 topics (AI/ML, Leadership, Growth, etc.)
Hook analysis — Identifies hook type (question, story, data-driven, contrarian)
Per-segment scoring — For each of your 4 segments, calculates topic affinity + hook strength
Aggregate score — Combines per-segment predictions weighted by segment size
Calibration — After 10+ tracked posts, adjusts predictions based on actual results

Why this works

You're not guessing what your audience wants. You're predicting what your specific segments have historically engaged with, then learning from every post you publish.

3.2

Prerequisites

What you need before starting

Required: Plugin 1 (LinkedIn Feed Tracker)

The Audience Simulator is Plugin 3. It depends on Plugin 1 to provide:

Your connections list (2,000+)
Your network's posts and engagement metrics
Your own post performance data
A local SQLite database (feeds.db) to store everything

Check the dependency

If you haven't installed Plugin 1 yet, go set that up first. It takes about 10 minutes. Come back here once Plugin 1 is running.

Data readiness

Ideally, let Plugin 1 collect data for 2+ weeks before running setup for Plugin 3. Here's why:

Cold start (0-7 days of data): Predictions are less accurate (MAE ~8 reactions)
Building calibration (1-2 weeks): Accuracy improves as you post and track results
Medium confidence (2+ weeks): Predictions become reliable (MAE ~2-3 reactions)

If you're just starting Plugin 1 now, don't worry. Plugin 3 will still work immediately—it just gets better over time as you post and let it learn.

What you'll need during setup

Your Cowork workspace folder (the same one Plugin 1 uses)
Access to Claude inside Cowork
2-5 minutes to run the setup command

Pro tip

Revisit setup after 2 weeks of Plugin 1 collecting data. The initial segment calibration will be stronger with more historical posts to analyze.

3.3

Setup

One command. Everything initializes.

Installation

Install the plugin from the Cowork Plugin Directory, or clone from GitHub:

git clone https://github.com/stevegustafson32/linkedin-audience-simulator-plugin ~/cowork-plugins/

Setup in Cowork

Open Cowork and run one command:

Command

/setup-audience-simulator

Claude will:

Check that Plugin 1 is installed and has data
Parse your 2,000+ connections and extract roles/seniorities
Create and assign the 4 audience segments
Analyze your network's recent posts to build topic affinities
Initialize the database tables for predictions and calibration

The whole process takes 1-2 minutes. Once done, you're ready to score posts.

That's it

No configuration files to edit. No environment variables. Just run the command and you're live.

Available commands

/setup-audience-simulator — Initialize segments and database
/score — Score a draft post (0-100 + per-segment breakdown)
/audience — View your segment sizes and engagement patterns
/deep-dive [segment] — Explore a specific segment in detail
/accuracy — Review prediction calibration and accuracy trends

3.4

What You Get

Engagement scores, segment insights, calibrated predictions

Post Scoring

Ask Claude: "Score this post" and paste your draft. You'll get:

Overall score (0-100) — How well your draft will perform with your audience
Per-segment breakdown — Predicted reactions/comments for each of your 4 segments
Confidence level — Low/Medium/High based on calibration data available
Reasoning — Which segments will engage and why
Recommendations — How to improve the draft for better engagement

Example output

Score: 78/100 — Technical Leaders' Territory
Your post scores best with Technical Leaders (predicted 22 reactions). The ML/Infrastructure topic is highly specialized (92% affinity with TLs, 5% with Executives). The question hook is strong for driving comments.

Per-segment breakdown:
• Technical Leaders: 22 reactions (85/100 hook strength)
• Growth & Sales: 6 reactions (35/100 hook strength)
• Executives: 3 reactions (28/100 hook strength)
• General: 1 reaction (15/100 hook strength)

Recommendations:
✓ Add a business outcome sentence for broader appeal
✓ Consider leading with results before technical details

Audience Segments Overview

Ask Claude: "Show my audience"

Get a high-level overview of your segments: size, engagement rates, top topics, and activity patterns.

Deep-Dive Analysis

Ask Claude: "Deep dive into [segment]" (e.g., "Deep dive into Executives")

Explore a specific segment in detail:

Role and seniority breakdown
Topics this segment engages with most
Best-performing content types for them
Notable patterns and behaviors

Variation Comparison

Ask Claude: "Compare these two post variations" and paste multiple drafts. The plugin scores each and shows you which will perform better with which segments.

Prediction Accuracy Tracking

Ask Claude: "How accurate are my predictions?" or "/accuracy"

Review:

Overall prediction accuracy (Mean Absolute Error)
Confidence levels by segment (how many predictions tracked)
Calibration coefficients and trends
Build plan (what to post more of to improve accuracy)

Learning over time

After 10-15 posts tracked, calibration kicks in and accuracy jumps significantly. After 50+ posts, predictions become highly reliable.

Integration with Plugin 2 (Content Engine)

If you've also installed Plugin 2, you can score its generated variations directly:

/score variation-1 — Score a specific Content Engine draft
/compare-variations — Rank multiple variations by predicted performance

3.5

Talk to Claude

Ask questions like you're texting a friend

Once setup is done, open Cowork and ask Claude anything about your audience and post performance. You don't need special syntax—just talk naturally.

Example conversations

Score this post draft: "Just shipped a new feature that cuts deployment time in half. How are you thinking about CI/CD?"

Score: 72/100 — Technical Leaders dominate. Technical Leaders will get 18 reactions, but Executives might only engage at 4. The technical hook is strong, but consider adding a business impact line for broader appeal.

Who in my network would engage with this?

Primarily Technical Leaders (92% affinity with Infrastructure topics). Secondary engagement from Growth & Sales (they care about delivery speed). Executives might engage if you lead with ROI impact. General audience will see it but probably won't comment.

Compare these two variations for me.

[Scores both, shows per-segment breakdown, recommends which will perform better overall and which targets specific segments best]

What topics should I post more about?

Based on your segments' engagement patterns, you have high affinity for AI/ML (especially with Technical Leaders), but low coverage in Leadership & Strategy (high affinity with Executives). Try 2-3 posts on leadership in the next week.

How accurate are my predictions getting?

You've scored 18 posts and tracked results on 4. Current MAE is 4.2 reactions. After 10-15 tracked posts, accuracy will jump to ~2-3 reactions. Focus on tracking results on every post for the next 2 weeks.

Deep dive into Technical Leaders for me.

[Detailed breakdown: 420 people, 92% engage with Infrastructure/ML, 45% engage with Leadership, top roles are Senior Engineer/Architect, best content types are technical deep-dives and case studies, etc.]

Other things to ask

"What's my audience breakdown?"
"Which of my last 3 posts did Executives engage with most?"
"Show me the prediction accuracy trend."
"What hook types work best for Growth & Sales?"
"Am I covering enough of the topics my Executives care about?"

3.6

The Ecosystem

Three plugins that work together

The Audience Simulator is Plugin 3 of a 3-plugin LinkedIn ecosystem built for Cowork. Each plugin builds on the others.

The three plugins

Plugin 1: LinkedIn Feed Tracker

Collects your network's posts, your own post performance, and your connections list. Runs nightly on your Mac. This plugin is the foundation—all others depend on it.

→

Plugin 2: Content Engine

Generates post variations and content ideas based on trends from Plugin 1. Uses your network's data to suggest what will resonate.

→

Plugin 3: Audience Simulator (you are here)

Predicts which segments will engage with your content before you post. Scores drafts from Plugin 2, calibrates against real results from Plugin 1.

Currently installed

How they work together

Typical workflow:

Plugin 1 collects data — Your network, posts, engagement patterns
Plugin 2 generates ideas — Suggests variations based on trends
Plugin 3 scores them — Predicts which variation will work best with your audience
You post the winner — On LinkedIn
Plugin 1 tracks results — Captures actual engagement
Plugin 3 learns — Calibrates predictions based on actual vs predicted

The power

You're not guessing what to post. You're generating ideas from real trends, predicting audience response, learning from every post, and getting better over time. All inside Cowork with Claude.

Ready to predict your audience?

Setup takes 2-5 minutes. Start scoring drafts immediately.