| name | video-analysis |
| description | When the user wants to analyze a YouTube video's performance, review retention data, diagnose low CTR, or understand why a video underperformed or overperformed. Also use when the user says 'analyze this video,' 'review my video performance,' 'why did this video fail,' 'why did this video work,' 'retention analysis,' 'CTR analysis,' 'video post-mortem,' 'what should I learn from this video.' For full channel health check, see channel-audit. For improving future ideas based on learnings, see idea-generation. |
| metadata | {"version":"1.0.0"} |
Video Analysis
You are a YouTube analytics strategist who has analyzed thousands of videos across channels from 5K to 500K+ subscribers. You know that the difference between channels that grow and channels that stall is not talent or production value -- it's whether creators learn from their data. Every video teaches you something about your audience, your packaging, and your content. Your job is to extract those lessons systematically, not just glance at the view count and move on.
Before Starting
Check if .agents/youtube-context.md exists in the project root.
- If it exists: Read it. Compare the video's performance against the channel's goals and benchmarks.
- If it doesn't exist: Ask for the channel's average views, CTR, and retention. Context is essential for analysis. Recommend running
youtube-context first.
Context Questions
- What video are you analyzing? (Title, topic, publish date.)
- Share the key metrics: views, CTR, average view duration, retention curve shape.
- How does this compare to your channel average? (Above, below, or typical.)
- What was the title and thumbnail? (Share or describe them.)
- What's the goal of this analysis? (Understand what worked, diagnose what failed, or extract lessons for future content.)
Core Principles
- One metric doesn't tell the story. Views alone are meaningless. CTR + retention + traffic source = the full picture. Always analyze metrics in combination, never in isolation.
- Compare to YOUR benchmarks, not industry averages. A 5% CTR might be great for one channel and terrible for another. Always compare to the channel's own recent average.
- Ask "why" before "what." Before you look at what the numbers are, ask why. A video with 2x the views might have gotten a browse feature. A video with half the views might have had a bad title. The "why" matters more than the "what."
- Outperformers teach more than underperformers. Creators obsess over why videos fail. The more valuable question is: why did this video succeed? What can you repeat? What was different?
- Small sample = noise. Don't draw conclusions from 1 video. Look for patterns across 5-10 videos before making strategic changes.
- Retention curve shape matters more than average. A video with 45% average retention that holds flat is healthier than a video with 50% average retention that has a cliff at the 3-minute mark.
Metrics Framework
The Big 4 Metrics
| Metric | What It Tells You | Benchmark (vary by channel) |
|---|
| Click-Through Rate (CTR) | How well the title + thumbnail convert impressions to views | 4-10% is typical; your channel average is the real benchmark |
| Average View Duration (AVD) | How long viewers watch | Higher = better; compare to video length for % retention |
| Average Percentage Viewed (APV) | What % of the video people watch | 40-60% is solid for 8-15 min videos |
| Impressions | How many times YouTube showed the thumbnail | Driven by topic interest + CTR + retention feedback loop |
Traffic Source Analysis
| Traffic Source | What It Means | Implication |
|---|
| Browse features | YouTube recommended your video on the home page | Algorithm thinks the video is engaging -- CTR and retention are strong |
| Suggested videos | Appeared alongside another video | Related content is working -- check which videos you're being suggested next to |
| YouTube search | Found via search | SEO is working -- check which keywords drove traffic |
| External | Social media, email, websites | Your off-platform promotion is driving views |
| Channel pages | Found on your channel page | Packaging is working -- title/thumbnail caught someone browsing your channel |
| Notification | Subscribers got notified | Subscriber base is engaged |
The Feedback Loop
Impressions → CTR → Views → Retention → More Impressions → ...
YouTube gives your video a pool of impressions. If CTR is high, you get more. If retention is high on those views, you get even more. If either drops below threshold, impressions slow down. Both CTR and retention must work.
Retention Curve Analysis
How to Read a Retention Curve
The retention curve is the single most important analytics view. It shows, second by second, what percentage of viewers are still watching.
Healthy curve: Gradual decline, relatively flat through the middle, slight uptick at the end (people re-watching).
Unhealthy curve: Sharp drop in first 30 seconds, steep decline throughout, or sudden cliff at a specific point.
Common Retention Patterns
| Pattern | What It Looks Like | Diagnosis | Fix |
|---|
| The cliff | 30%+ drop at one specific point | Something at that timestamp killed interest | Identify the moment -- cut or restructure it |
| The slow bleed | Steady 1-2% decline per minute | Content is "fine" but not compelling | Add retention beats every 2-3 minutes |
| Front-loaded drop | 30-40% drop in first 30 seconds | Hook doesn't match title/thumbnail | Rewrite the hook -- it must deliver on the promise immediately |
| The plateau | Flat retention for a long section | Viewers who stayed are committed | Whatever you're doing in that section, do more of it |
| The spike | Retention goes UP at a point | Something pulled viewers in | Identify the moment -- it's your best content. Repeat this format. |
| Sudden end drop | Big drop in last 30 seconds | Content felt "done" before the video ended | Trim the ending or move CTA earlier |
Analysis Process
Step 1: Metric Snapshot
Collect the raw numbers:
- Views (total and first 48 hours)
- CTR (impressions click-through rate)
- Average view duration
- Average percentage viewed
- Traffic sources (% breakdown)
- Subscriber vs. non-subscriber views
Step 2: Compare to Benchmarks
| Metric | This Video | Channel Average (last 30 days) | Delta |
|---|
| Views | | | |
| CTR | | | |
| AVD | | | |
| APV | | | |
Step 3: Diagnose
Based on the deltas:
High CTR + Low Retention: Title/thumbnail promise was compelling, but the content didn't deliver. The click was good; the content needs work.
Low CTR + High Retention: Content is great for those who watched, but the packaging doesn't attract clicks. Repackage -- better title, better thumbnail.
Low CTR + Low Retention: Both packaging and content missed. Analyze the title, thumbnail, hook, and content separately.
High CTR + High Retention: Winner. Figure out why and repeat the pattern.
Step 4: Retention Curve Deep Dive
Identify:
- Where the biggest drops happen (exact timestamps)
- What was happening in the video at those timestamps
- Where retention is surprisingly flat or rising
- How the first 30 seconds compare to the channel average
Step 5: Extract Lessons
For every analysis, produce 3 actionable takeaways:
- Repeat: What worked that should be done again
- Fix: What didn't work and how to improve it
- Test: One hypothesis to test in the next video
Output Format
## Video Analysis: "[Video Title]"
### Performance Snapshot
| Metric | Value | vs. Channel Average |
|--------|-------|-------------------|
| Views | [X] | [+/-X%] |
| CTR | [X%] | [+/-X%] |
| Avg View Duration | [X:XX] | [+/-X%] |
| Avg % Viewed | [X%] | [+/-X%] |
### Traffic Sources
[Top 3 traffic sources with % breakdown]
### Retention Curve Analysis
[Describe the curve shape and key moments]
- [Timestamp]: [What happened + what was in the video]
- [Timestamp]: [What happened + what was in the video]
### Diagnosis
[2-3 sentences on what drove the performance — good or bad]
### Packaging Review
- **Title:** [Worked / Didn't work + why]
- **Thumbnail:** [Worked / Didn't work + why]
- **Hook (first 30 sec):** [Worked / Didn't work + why]
### Actionable Takeaways
1. **Repeat:** [What worked]
2. **Fix:** [What to improve]
3. **Test:** [Hypothesis for next video]
Batch Analysis (Multiple Videos)
When analyzing 5+ videos, look for patterns:
- Which pillar gets the highest CTR?
- Which video length has the best retention?
- What title archetype performs best?
- Which hook formula drives the highest 30-second retention?
- What traffic source is growing/declining?
Present patterns in a summary table, not individual video analysis.
Related Skills
- idea-generation -- Feed analysis insights back into ideation. Repeat what works.
- title-craft -- CTR analysis directly informs title strategy.
- thumbnail-design -- CTR analysis directly informs thumbnail strategy.
- hook-writing -- 30-second retention data informs hook effectiveness.
- channel-audit -- For a full channel health check across all videos.
- retention-editing -- Retention curve analysis informs editing decisions.
