Creative Copilot is Nanga's AI module dedicated to optimizing your advertising creatives. It automatically analyzes your ads and classifies them into three categories to help you make the right decisions quickly:
- Top Performers — Creatives delivering exceptional results
- Bad Performers — Underperforming creatives that should be paused
- Ad Fatigue — Previously strong creatives showing signs of exhaustion
This classification relies on a weighted scoring system, comparing each creative against dynamic benchmarks from your own account.
Prerequisites: Creative Filtering
Before any analysis, creatives must meet these minimum thresholds:
| Criterion | Value | Reason |
|---|---|---|
| Status | Active | Only analyze running ads |
| Minimum Spend | ≥ €50 | Ensure statistical significance |
Creatives below €50 spend are excluded from both analysis and benchmark calculations to avoid statistical noise.
Dynamic Benchmarks
Why Personalized Benchmarks?
Rather than using arbitrary thresholds, we compare each creative to your account's performance history. A creative is evaluated against what's actually achievable for your audience, product, and market.
- Window: Rolling 12 months
- Geographic scope: Benchmarks calculated only on creatives from the same country
Segmentation by Type and Age
Benchmarks are calculated separately for each type/age combination to ensure fair comparisons.
Creative Types:
| Category | Formats |
|---|---|
| VIDEO | video, reel, dynamic |
| IMAGE | static, carousel, feed, collection, gif |
Age Brackets:
| Bracket | Active Days | Description |
|---|---|---|
| EARLY | 1-7 days | Learning phase, volatile metrics |
| ESTABLISHED | 8-30 days | Stable phase, reliable performance |
| MATURE | 31+ days | Long-running, saturation risk |
A 5-day-old video is compared to other early-stage videos, not to 60-day-old images.
Percentiles Used
For each segment, we calculate percentiles P10, P25, P50, and P75:
- P75 = top 25% of creatives (Top Performer threshold)
- P25 = bottom 25% (Bad Performer threshold)
- P50 = median
Key Metrics
Performance Metrics
| Metric | Formula | Interpretation |
|---|---|---|
| ROAS | Revenue / Spend | Return on ad spend. ROAS 2.5 = €2.50 generated per euro spent |
| CPC | Spend / Clicks | Cost per click. Lower = better |
| CTR | (Clicks / Impressions) × 100 | Click-through rate (%) |
Video Metrics
| Metric | Formula | Interpretation |
|---|---|---|
| Hook Rate | (3s Views / Impressions) × 100 | % of viewers captured in the first seconds |
| Hold Rate | (ThruPlays / 3s Views) × 100 | % of viewers who watched until the end |
Creative Signal
The "creative signal" is the most predictive engagement metric for each format:
| Format | Signal | Reason |
|---|---|---|
| Images | CTR | A static image must generate clicks |
| Videos | Average (Hook + Hold) | A video must capture AND retain attention |
Scoring System
Each creative receives three scores (0-100%), one per category. It's assigned to the category with the highest score, if that score reaches at least 70%.
Top Performer Score
| Criterion | Weight | Condition |
|---|---|---|
| Objective metric > P75 | 40% | In the account's top 25% |
| Spend > P50 | 20% | Significant budget allocated |
| Conversions > P50 | 5% | Conversion volume |
| Creative signal > P50 | 15% | Good engagement |
| Stable/growing performance | 10% | Trend ≥ -5% |
| CPC < P50 | 10% | Efficient clicks |
Bad Performer Score
| Criterion | Weight | Condition |
|---|---|---|
| Objective metric < P25 | 40% | In the bottom 25% |
| Spend > P25 | 20% | Wasted budget |
| Conversions < P10 | 5% | Almost no conversions |
| Creative signal < P25 | 15% | Low engagement |
| Declining performance | 10% | Trend < -10% |
| CPC > P75 | 10% | Expensive clicks |
Ad Fatigue Score
| Criterion | Weight | Condition |
|---|---|---|
| Declining objective metric | 40% | Trend < -10% |
| Running > 7 days | 25% | Saturated audience |
| Rising CPM | 20% | CPM +15% vs last 7 days |
| Declining creative signal | 15% | -10% vs last 7 days |
Final Classification
Decision Logic
IF score_top ≥ 70% AND trend > -50%:
→ TOP_PERFORMER
ELSE IF score_bad ≥ 70%:
→ BAD_PERFORMER
ELSE:
→ Unclassified
Fatigue as an Overlay
Fatigue is an additional flag, not an exclusive category:
-
A Top Performer can be flagged as fatigued if fatigue_score ≥ 85% → "Still performing but prepare a replacement"
-
An unclassified creative is flagged as fatigued if fatigue_score ≥ 70% → "Declining due to audience saturation"
-
Bad Performers are never flagged as fatigued (should just be paused)
Image vs Video: Key Differences
| Aspect | Images | Videos |
|---|---|---|
| Creative signal | CTR only | Hook Rate + Hold Rate |
| "Bad signal" condition | CTR < P25 | Hook OR Hold < P25 |
| Main optimization lever | Visual, copy, CTA | First 3 seconds (hook) |
Why Hook + Hold for Videos?
A video can fail in different ways:
- Good hook, bad hold → "Clickbait" effect, viewers leave after a few seconds
- Bad hook, good hold → Slow start, viewers lost before the value
Both metrics matter, and changing the hook is often the fastest lever to optimize a video.
Summary
Creative Copilot lets you instantly identify your best creatives, those to cut, and those showing signs of fatigue. All based on your own historical data, not generic benchmarks.
Coming soon: choice of objective metric (ROAS, CAC, CPA, CPL...) based on your campaign goals.