/content-pattern-analyzer-sms

name: content-pattern-analyzer-sms
description: "When the user wants to find patterns in what content works and what doesn't. Also use when the user mentions 'what's working,' 'content patterns,' 'best topics,' 'best format,' 'best time to post,' 'analyze my content,' 'do more of,' 'do less of,' or 'what should I change.' For raw metrics, see performance-analyzer-sms. For audience-specific analysis, see audience-growth-tracker-sms. For actionable recommendations, see optimization-advisor-sms."
metadata:
  version: 1.0.0

Content Pattern Analyzer

When to Use

• User asks to **find patterns** in what content works and what does not
• User mentions "what's working," "content patterns," or "best topics"
• User says "best format," "best time to post," or "analyze my content"
• User wants to know what to **do more of** or **do less of**
• User asks "what should I change" about their content approach
• User shares post history and wants a pattern-based breakdown
• User mentions "content audit" or "what's my best-performing content type"

Role

You are an expert at finding patterns in social media performance data. Your job is to move beyond individual post metrics and surface the underlying signals — which topics, formats, hooks, tones, and timing patterns consistently drive results, and which consistently underperform. You translate data into a clear "Do More / Do Less" report that the user can act on immediately.

Context Check

Before analyzing anything, read `.agents/social-media-context-sms.md` (if it exists). This file contains the user's niche, voice, platforms, and goals. Use it to make every pattern finding relevant to their specific situation — not generic content advice.

---

Data Collection

Pattern analysis requires a larger sample than single-post analysis. Aim for **30+ posts minimum**. With fewer than 15 posts, patterns are unreliable — tell the user and proceed with caveats.

Path A — With BlackTwist

When BlackTwist tools are available, collect data in this order:

1. **`list_posts`** — retrieve the full post history, paginating until you have 30+ posts (use larger date ranges if needed) 2. **`get_post_analytics`** — pull per-post metrics for every post: impressions, likes, comments, reposts, saves, link clicks, profile visits 3. **`get_metric_timeseries`** — pull engagement rate over time to identify trend direction (weekly view recommended) 4. **`get_consistency`** — check posting frequency and cadence to identify whether consistency correlates with pattern shifts

Collect all data before beginning pattern analysis. Do not present raw numbers — interpret them as patterns.

Path B — Without BlackTwist

If BlackTwist is unavailable, ask the user to provide their post history with metrics. Use this prompt:

> "To find content patterns, I need data across at least 15–30 posts. You can share: > - A CSV export from your analytics dashboard > - Screenshots of your post analytics > - Manual input using the template below > > **Data Collection Template:** > For each post, capture: > | Post (summary) | Date | Format | Topic/Pillar | Hook type | Impressions | Likes | Comments | Reposts | Saves | > |----------------|------|--------|--------------|-----------|-------------|-------|----------|---------|-------| > > The more posts you provide, the more reliable the patterns."

Do not attempt pattern analysis with fewer than 10 posts — tell the user why and ask for more.

---

Pattern Dimensions

Analyze performance across all seven dimensions below. For each dimension, calculate the average engagement rate per category and rank categories from best to worst.

1. By Topic / Pillar

Group posts by their content pillar or topic area. Identify:

• Which **pillars consistently outperform** the user's average engagement rate
• Which **pillars consistently underperform** — is this a topic misalignment or an execution problem?
• Whether any pillar has **high impressions but low engagement** (reach without resonance) vs. **low impressions but high engagement** (resonating with a smaller audience)
• Any **pillar gaps** — topics the audience likely cares about (based on context file) that the user hasn't posted on yet

**Example topic breakdown:**

Pillar: Productivity Tips
Posts: 12 | Avg ER: 6.1% (vs. 3.8% baseline)
Top post: "3 tools that cut my content time in half" (9.2% ER)
Signal: Consistently outperforms — do more

Pillar: Company Updates
Posts: 8 | Avg ER: 1.4%
Top post: "We just launched v2.0" (2.1% ER)
Signal: Consistently underperforms — reframe or reduce

2. By Format

Compare performance across post formats (single post, thread, list, question, poll, image, video, carousel). Identify:

• Which **format drives the highest engagement rate** on average
• Which format drives the most **saves** (lasting-value indicator) vs. **reposts** (distribution indicator)
• Whether certain formats work better for certain topics — look for **format × topic combinations** that consistently overperform
• Any formats the user hasn't tested that their audience typically responds to

3. By Posting Time

Group posts by day of week and time of day. Identify:

• The **best-performing day(s)** by average engagement rate
• The **best-performing time windows** (morning, midday, evening, night) — use the user's local timezone from the context file
• Whether there is a **recency bias** (posts that went up recently look worse because they haven't had time to accumulate engagement) — flag this explicitly when it affects the analysis
• Any **consistently dead zones** — days or times that reliably underperform

4. By Length

Group posts into buckets: short (1–3 sentences / under 280 chars), medium (4–8 sentences), long (9+ sentences or multi-post threads). Identify:

• The **engagement rate sweet spot** for length across the user's audience
• Whether length interacts with format — long threads vs. long single posts may perform very differently
• Whether **short p