Video & Scriptwriting / YouTube Scripts
Predict where viewers drop off based on script structure and suggest retention fixes — prevents early drop-off.
Why This Prompt Exists
Most creators look at retention analytics after publishing — when it’s too late to fix. Predictive retention analysis identifies drop-off points before you film.
You get:
- first 30-second cliff (50-80% drop) — preventable with better hook
- mid-roll drop-off (viewers bored, no retention mechanism)
- pattern of small drops (death by a thousand cuts, each fixable)
- late-video drop (CTA too early or too late)
- no understanding of why viewers leave at specific timestamps
But retention patterns are predictable:
- hook zone (0-30s): drop-off from poor hook or misleading title
- intro zone (30-90s): drop-off from slow pacing or no value promise
- value zone (90s-5min): drop-off from irrelevant tangents or poor explanations
- retention zone (5min+): drop-off from fatigue, no pattern breaks
- closing zone (last 60s): drop-off from CTA too early, content already ended
Without predictive analysis, you fix problems after the video fails.
This prompt analyzes scripts for retention drop-off points.
The Prompt
Assume the role of a YouTube retention analyst who predicts drop-off points. Your task is to analyze a script and identify where viewers will leave. Generate: 1. SCRIPT TIMESTAMP BREAKDOWN | Timestamp | Section | Content Type | Predicted Retention | Risk Level | |-----------|---------|--------------|---------------------|------------| | 0:00-0:30 | Hook | [type] | [X%] | High/Med/Low | | 0:30-1:30 | Intro | [type] | [X%] | High/Med/Low | | 1:30-3:00 | Value 1 | [type] | [X%] | High/Med/Low | | 3:00-5:00 | Value 2 | [type] | [X%] | High/Med/Low | | 5:00-7:00 | Value 3 | [type] | [X%] | High/Med/Low | | 7:00-8:00 | Closing | CTA | [X%] | High/Med/Low | 2. DROP-OFF PREDICTIONS BY ZONE | Zone | Typical Drop | Your Script Risk | Fix | |------|--------------|------------------|-----| | 0-30s (Hook) | 50-80% | High/Med/Low | [specific hook fix] | | 30-90s (Intro) | 10-20% | High/Med/Low | [pacing or promise fix] | | 90s-5min (Value) | 5-15% | High/Med/Low | [tangent or clarity fix] | | 5min+ (Retention) | 2-10% | High/Med/Low | [pattern break or payoff fix] | | Last 60s (Closing) | 5-15% | High/Med/Low | [CTA timing fix] | 3. RETENTION MECHANISMS MISSING | Mechanism | Description | Your Script Has? | Add Where | |-----------|-------------|------------------|-----------| | Preview of value | Tell them what they'll get | Yes/No | [timestamp] | | Pattern breaks | Change visuals, pacing, topic | Yes/No | [timestamp] | | Tease upcoming | "Stick around for X" | Yes/No | [timestamp] | | Micro-hooks | Small questions every 30-60s | Yes/No | [timestamp] | | Stakes reminder | Why this matters | Yes/No | [timestamp] | 4. PREDICTED RETENTION CURVE Start: 100% 0:30: [X]% 1:30: [X]% 3:00: [X]% 5:00: [X]% 7:00: [X]% 8:00: [X]% Estimated AVD (Average View Duration): [X:XX] / [total length] Estimated VVSA (Viewer Value Score): [X/10] 5. FIX PRIORITIES (highest impact first) | Priority | Timestamp | Issue | Fix | Expected Retention Gain | |----------|-----------|-------|-----|------------------------| | 1 | 0:00-0:15 | Weak hook | [specific fix] | +15-25% | | 2 | 0:30-1:00 | Slow pacing | [specific fix] | +10-15% | | 3 | 2:00-2:30 | Tangent | [specific fix] | +5-10% | 6. POST-PUBLISH VERIFICATION After publishing, check actual retention against predictions: - Actual 30s retention: [X%] (vs. predicted [X%]) - Actual 60s retention: [X%] (vs. predicted [X%]) - Highest drop-off point: [timestamp] - Compare to fix predictions to improve future scripts INPUTS: Script (full or detailed section descriptions): [PASTE SCRIPT OR DESCRIBE SECTIONS WITH TIMESTAMPS] Video length: [E.G., "8 minutes"] Content type: [TUTORIAL / COMMENTARY / STORYTELLING / REVIEW / EDUCATIONAL] Channel average retention (if known): [E.G., "45% at 30 seconds, 30% average view duration"] RULES: - First 30 seconds determine 50%+ of retention (prioritize here) - Every 30 seconds needs a micro-hook (question, transition, tease) - Tangents kill retention (if it doesn't serve the value promise, cut it) - Pattern breaks (visual changes, pacing shifts) reset attention - Previewing the value early increases retention (tell them what they'll get) - CTAs should come after the value, not before (don't ask before giving) - Test retention predictions against actual data to improve your model
How To Use It
- First 30 seconds determine 50%+ of retention — prioritize this zone above all else.
- Every 30 seconds needs a micro-hook — a question, transition, or tease to keep attention.
- Tangents kill retention — if a section doesn’t serve the value promise, cut it.
- Pattern breaks (visual changes, pacing shifts, topic transitions) reset viewer attention.
- Previewing the value early increases retention — tell them what they’ll get upfront.
- CTAs should come after the value, not before — don’t ask for engagement before delivering.
- Test retention predictions against actual data to improve your predictive model.
Example Input
Script:
“0:00-0:20: Hey guys, welcome back to the channel. Today we’re talking about camera gear. 0:20-1:30: So I’ve been testing this new lens for three weeks… [detailed specs]. 1:30-3:00: Let me show you some sample footage… 3:00-5:00: Here’s how it compares to the previous model… 5:00-6:00: Should you buy it? My verdict… 6:00-6:30: Don’t forget to like and subscribe!”
Video length:
“6.5 minutes”
Content type:
“PRODUCT REVIEW”
Channel average retention:
“55% at 30 seconds, 35% average view duration”
Why It Works
Most creators discover retention problems after publishing — when the video is already live and underperforming.
This framework improves outcomes by forcing:
- timestamp breakdown (where retention drops by zone)
- drop-off prediction (identifying risks before filming)
- retention mechanism audit (what’s missing from the script)
- predicted retention curve (estimating AVD and VVSA)
- fix prioritization (highest impact first)
Failure modes this prevents:
- First 30-second cliff that could have been fixed with a better hook
- Mid-roll boredom from no retention mechanisms
- Death by a thousand small drops (each fixable individually)
- CTA too early or too late (missed engagement opportunity)
This improves on: Post-publish retention analysis. Predictive analysis fixes problems before filming.
Related to: YT-01 (Hook) for opening retention; YT-03 (Structure) for pacing.
Build Better AI Systems
Subscribe for advanced prompt engineering, AI coding tools, debugging frameworks, and practical strategies for developers and engineers.