By name: retention-surgeon description: "Diagnose retention problems using benchmarks, identify which of the 4 curve shapes you have, recommend cutting patterns, and fix the first 30 seconds. Use when a video underperforms or when editing." argument-hint: "[paste retention data, or describe the video and its performance]"
Retention Surgeon
Diagnose and fix retention for: $ARGUMENTS
Step 1: Identify the Retention Curve Shape
The 4 Shapes
The Cliff — Steep early drop (>30% lost in first 30 seconds)
- Diagnosis: Failed hook. Title/thumbnail promise doesn't match opening.
- Fix: Rewrite first 30 seconds. Deliver on the promise IMMEDIATELY.
Gradual Decline — Steady drop throughout
- Diagnosis: Natural attrition. Healthy if above 40% at midpoint.
- Fix: Add pattern interrupts every 20-30 seconds. More Setup-Tension-Payoff cycles.
The Bump — Spike at a specific point (people rewatching a segment)
- Diagnosis: You have a rewatch-worthy moment. Extract it as a Short/clip.
- Fix: Study what made that moment work. Replicate the structure.
Flat Line — Consistent engagement throughout
- Diagnosis: Ideal. Rare. Your content is holding.
- Fix: Nothing. Study this video and replicate.
Step 2: Benchmark Against Standards
By Video Length
| Length | Good | Great | Problem |
|---|---|---|---|
| Under 5 min | 50% | 60%+ | Below 40% |
| 5-15 min | 40% | 50%+ | Below 30% |
| 15-30 min | 30% | 40%+ | Below 25% |
| Over 30 min | 25% | 35%+ | Below 20% |
| Shorts | 70% | 90%+ | Below 50% |
By Content Type
- Tutorials: 45-55% (viewers skip to relevant parts — expect dips)
- Thought leadership: 35-50%
- Product demos: Bimodal curves typical
- Interviews/Podcasts: 25-35%
Beast-Style Segment Targets
Use these as aggressive targets for highly packaged browse content:
| Time | Target | Strategy |
|---|---|---|
| 0-1 min | 90%+ retention | Perfect hook, no wasted setup |
| 1-3 min | 80%+ retention | First escalation and clear reason to stay |
| 3-6 min | 65%+ retention | Twist, reveal, or upgrade every 3 minutes |
| 6 min+ | 50%+ retention | Keep stair-stepping toward the peak moment |
| Final 30 sec | Hold or redirect | CTA, payoff, or next-video bridge |
Engagement Ratios
- Like-to-view: 4-8% (long-form), 3-6% (Shorts)
- Return viewer rate: Above 10% = building an audience
Step 3: First 30 Seconds Audit
The first 30 seconds are make-or-break. Check:
- Does it deliver on the title's promise within 5 seconds?
- Is there ZERO preamble before the value starts?
- Does it introduce specific value the viewer will get?
- Is there a pattern interrupt at 25-35 seconds?
- Does it establish purpose ("By the end, you'll know...")?
- MrBeast test: Does the viewer see what they gain or lose in first 5 seconds?
- Does it show the best visual proof, final result, or biggest problem before explaining background?
- Is action underway by 15-30 seconds?
If any box is unchecked, rewrite the opening.
Step 4: Recommend a Cutting Pattern
5 Proven Patterns (AIR Media-Tech)
1. Progressive Rhythm — Best for: educational, explainer
- Min 0-3: Tight pacing, frequent visual changes
- Min 3-7: Stabilize, fewer cuts, more B-roll
- After min 8: Mix calm with energy bursts
- Used by: Veritasium, Ali Abdaal
2. Contrast Pattern — Best for: commentary, essay
- Simple pacing (15-25 sec/cut) most of the time
- Every 2-3 min: burst of 5-10 quick cuts
- Return to calm. Mimics conversation rhythm.
- Used by: Ryan Trahan, Drew Gooden
3. Narrative Loop — Best for: challenge, story, MrBeast-style
- Open with hook/question
- Every 2-3 min, remind viewers of core premise
- Smoother transitions, ambient bridges
- Used by: MrBeast, YesTheory
4. Hybrid Tempo — Best for: tutorial with energy
- Fast mode: micro-cuts every 10-15 sec with zooms/graphics
- Slow mode: holds up to 40 sec during examples
- Used by: Better Ideas, Think Media
5. Anchor Pattern — Best for: personal, documentary
- Edit at emotional beats, not time intervals
- Hold shots through full statements before cutting
- Ambient sound guides pacing
- Used by: Nathaniel Drew, Johnny Harris
Step 5: The Minute 8 Threshold
YouTube starts rewarding with higher suggested placement after minute 8. If your video is 8+ minutes:
- Maintain high energy through minute 8
- Place a strong payoff or reveal just past the 8-minute mark
- This triggers algorithmic momentum for suggested placement
Step 5.5: Re-Engagement Moments
If the curve drops every few minutes, add a re-engagement moment every 3-5 minutes:
- New rule or constraint
- New visual environment
- Unexpected failure
- Bigger prize, cost, timer, or consequence
- Open loop payoff followed by a new open loop
Step 6: Sound Design Check
- Voice dominates, music 5-25dB quieter
- Calm segments: -20 to -25dB music
- Energetic sequences: -8 to -12dB music
- Music track changes when story shifts
- Brief silence before major reveals
- Heavy transitions reserved for BIG moments only
Output Format
## Retention Diagnosis: "[Video Title]"
### Curve Shape: [Cliff / Gradual / Bump / Flat]
**Current retention**: [X]% average
**Benchmark**: [Good/Below/Great] for [length] [content type]
### Problem Areas
1. [Timestamp] — [what's happening] — [why viewers leave here]
2. [Timestamp] — [what's happening] — [why viewers leave here]
### Recommended Cutting Pattern: [Pattern Name]
**Why**: [reason this pattern fits this video]
### Fix List (priority order)
1. **First 30 seconds**: [specific rewrite]
2. **[Timestamp]**: [specific fix]
3. **[Timestamp]**: [specific fix]
### Opportunities
- **Rewatch spike at [timestamp]**: Extract as Short
- **Minute 8 checkpoint**: [is energy maintained? what to add?]
### Sound Design Notes
- [any sound issues identified]