# Sentiment Analysis > Analyzes customer sentiment from reviews, social media, and support tickets with trend tracking, theme categorization, and alert recommendations. Use for brand health monitoring. - Skill: `reaperinvest/sentiment-analysis` (Agent Skill) - Install (CLI): `npx skillmds@latest add reaperinvest/sentiment-analysis` - Raw SKILL.md: https://api.skillmd.com/api/skills/reaperinvest/sentiment-analysis/raw - Safety review: pending - Works with: Claude Code, Claude.ai, OpenAI Codex - Category: Data & Analytics - Author: reaperinvest (https://skillmd.com/u/reaperinvest) - Updated: 2026-09-22 - Page: https://skillmd.com/skills/reaperinvest/sentiment-analysis --- # Sentiment Analysis ## When to Use This Skill Use this skill when you need to: - Analyze customer sentiment from reviews, social posts, or support tickets - Track brand perception trends over time - Categorize feedback themes to prioritize product or service improvements - Set up a sentiment monitoring framework for ongoing use **DO NOT** use this skill for social media content creation, customer service response writing, or NLP model building. This is for interpreting and acting on customer sentiment data. --- ## Core Principle SENTIMENT IS A LEADING INDICATOR — DECLINING SENTIMENT PREDICTS CHURN, NEGATIVE REVIEWS, AND REVENUE LOSS BEFORE THEY SHOW UP IN YOUR FINANCIALS. --- ## Phase 1: Brief ### Required Inputs | Input | What to Ask | Default | |-------|------------|---------| | **Data sources** | "Where is the feedback? (Google reviews, App Store, social media, support tickets, surveys)" | Must be provided | | **Volume** | "Roughly how many pieces of feedback to analyze?" | 50-200 | | **Time period** | "What date range?" | Last 90 days | | **Focus** | "What do you want to understand? (overall sentiment, specific product, competitor comparison)" | Overall brand sentiment | | **Segments** | "Any segments to analyze separately? (product line, customer tier, channel)" | Overall first | | **Existing tracking** | "Do you have any sentiment tracking in place?" | None | **GATE: Confirm brief before proceeding.** --- ## Phase 2: Analyze ### Sentiment Scoring Framework Classify each piece of feedback: - **Positive** — praise, satisfaction, recommendation - **Neutral** — factual, no strong emotion, mixed - **Negative** — complaint, frustration, warning to others ### Theme Categorization Tag every piece of feedback with 1-2 themes: - Product quality, pricing, customer service, delivery/speed, usability, feature requests, competitor mentions, billing issues ### Analysis Dimensions 1. **Overall sentiment distribution** — % positive, neutral, negative 2. **Sentiment by theme** — which topics generate the most negativity? 3. **Sentiment trend** — is sentiment improving or declining over time? 4. **Volume trend** — are more people talking? (volume increase + negative sentiment = alarm) 5. **Competitive mentions** — how often do customers mention competitors and in what context? **GATE: Present preliminary findings and confirm focus areas for the full report.** --- ## Phase 3: Build ### Deliverables **1. Sentiment Analysis Report** - Overall sentiment score and distribution - Theme-by-theme sentiment breakdown - Trend chart over the analysis period - Top 10 representative quotes (positive and negative) - Competitive mention summary **2. Issue Priority Matrix** | Theme | Sentiment | Volume | Trend | Priority | |-------|-----------|--------|-------|----------| | Customer service | Negative | High | Worsening | Critical | | Product quality | Positive | High | Stable | Protect | | Pricing | Mixed | Medium | Stable | Monitor | **3. Alert Framework** Define triggers for ongoing monitoring: - Negative sentiment exceeds 30% in any week - New negative theme appears that was not previously tracked - Star rating drops below 4.0 on any review platform - Competitor mentioned positively more than 10% of the time **4. Response Playbook** - Template responses for common negative themes - Escalation criteria for serious complaints - Proactive outreach triggers for at-risk customers --- ## Phase 4: Polish ### Monitoring Dashboard Spec Recommend a simple tracking system: - Weekly sentiment score by source - Theme trend tracking (monthly) - Alert log for triggered notifications ### Quarterly Sentiment Review Template for a quarterly deep-dive comparing current sentiment to previous quarter with action plan updates. --- ## Example 1: App Store Reviews (150 reviews, SaaS mobile app) **Finding:** 72% positive, 18% negative, 10% neutral. Top negative theme: onboarding confusion (8 mentions). Top positive: time-saving features. **Action:** Improve onboarding flow, create tutorial videos. ## Example 2: Google Reviews (80 reviews, local service business) **Finding:** 4.2 average stars. Negative reviews cluster around wait times (6 mentions) and billing clarity (4 mentions). Positive reviews highlight staff quality. **Action:** Address wait time communication, simplify billing invoices. --- ## Anti-Patterns - **Ignoring negative feedback** — 1 negative review often represents 10 silent churners. Take complaints seriously. - **Counting stars without reading text** — a 3-star review with specific feedback is more useful than a 5-star "Great!" Read the content. - **Analyzing once and forgetting** — sentiment shifts over time. Set up ongoing monitoring, not one-time reports. - **Responding defensively** — negative sentiment analysis should drive improvement, not defensive PR responses. - **Overreacting to one bad review** — look for patterns. One complaint is an anecdote; five on the same topic is a trend. --- ## Recovery - **Very few reviews to analyze:** Supplement with support ticket data, social media comments, or direct customer interviews. - **Overwhelmingly positive (suspicious):** Dig deeper — are reviews incentivized? Check for patterns suggesting fake reviews. - **User takes negative feedback personally:** Reframe as a competitive advantage — you now know exactly what to fix. Competitors are guessing. - **No ongoing monitoring resources:** Set up Google Alerts and a monthly 30-minute review habit. Low-effort monitoring beats no monitoring.