Pourquoi itérer sur vos "Top créas" ne sert à rien (et comment ça vous coûte cher)

Core business message

• “Top créas” (the creatives with the highest spend/conversions) create a biased, incomplete view of what truly drives performance.
• Iterating by repeating/deriving from top-performing creatives often fails because performance depends on contextual and structural factors, such as:
  • timing/seasonality
  • the testing environment
  • competing creatives
• KODAC proposes a data-driven creative iteration playbook that turns each test into structured learning, enabling continuous improvement rather than blind copying.

---

Problems with the “Top créas” approach (what goes wrong)

Brands often optimize toward the most widely distributed creatives, but those are frequently:

• Not reproducible
• Not the best for long-term effectiveness

As a result:

• Teams can “relaunch” essentially the same creative yet see no improvement.
• You miss the mechanisms that actually sell, because top results don’t reveal the true “why.”
• Performance may also be caused by invisible external factors (e.g., timing, seasonality, other concurrent creatives), so “what worked today” may not work tomorrow.

---

KODAC “lab method” / iteration framework (process & scoring)

Shift in principle

• From: “A/B test variations of the top creatives based on intuition”
• To: “Iterate with data to understand underlying mechanisms across the whole account.”

Structured analysis & tool

• Uses a tool called Motion to analyze creative performance beyond top performers.
• Builds a multidimensional scoring and diagnostic model.

---

Dimensions / what they measure

• Hooks: audience-triggering lines/angles
• Marketing angles (examples):
  • gift vs price
  • product framing differences
• Product framing (examples):
  • “mistress” vs “family outings” as different product narratives
• Engagement signals (examples cited):
  • CPC
  • Thumbstop (video attention)
  • Old rate on videos (likely watch/retention-related metric)
• Business impact / conversion efficiency (examples):
  • ROS
  • CPA efficiency

---

Scoring system (to reduce bias from spend)

Creative scoring weights include:

• Distribution across all campaigns
• Engagement
• Business impact

This produces:

• common patterns among winners
• identification of “hidden gems” that may not have the highest spend

---

Operational taxonomy

A clear naming/classification system categorizes tests by:

• Category
• Products concerned
• Campaign

Then analysis breaks down creative elements such as:

• Message
• Hook
• Format length

Goal: convert each test into action-oriented learning so patterns become explicit and comparable.

---

Concrete example: “Top Creative” iteration fails to reproduce results

Setup

• Original top creative (left) vs multiple new versions (right)
• New tests varied:
  • hooks
  • lengths
  • other parameters

Observed performance changes (key metrics)

• Spend: from €2,490 down to much lower amounts
• ROS: from 1.61 down to 0.93
• CPC: from €0.50 up to > €1.54

Engagement improved, but business efficiency collapsed

• Hook score: 74 vs 59 (original)
• Watch score: 82 (higher engagement signals)
• But conversion/business efficiency fell sharply → engagement ≠ conversion

Core lesson

Optimizing only for top creative performance without understanding the mechanics can create paradoxical outcomes:

• better attention metrics
• worse revenue efficiency

---

Concrete example: two strategies that split engagement vs conversion

Client example: “pairs party”

Strategy A (organic POV style)

• strong engagement signals
• good spend
• decent CPC
• excellent hook score: 77
• but CPA very bad (weak conversion efficiency)

Strategy B (more branded/classic)

• lower engagement (lower “spin,” average hook score 53)
• but converts twice as well (better conversion efficiency)

Lesson

Creative performance requires balance:

• One approach can attract & spread
• Another can convert

So iteration should aim for a smart mix, not a one-size-fits-all creative style.

---

Actionable recommendations (implied playbook)

• Treat every campaign/test as an opportunity for structured learning, not replication.
• Build learning around creative mechanisms:
  • identify what drives attention (hooks)
  • identify what drives spread (engagement/distribution behavior)
  • identify what drives conversion (business impact / CPA/ROS efficiency)
• Stop using top creatives as a proxy for causality; instead:
  • compare performance across angles, hooks, product framing, formats, and lengths
  • evaluate with a scoring method that accounts for distribution bias
• Anticipate rather than react to performance fluctuations:
  • “What will work” becomes more predictable when you understand the why.

---

Metrics / KPIs explicitly mentioned

• Spend (e.g., €2,490 in the example)
• ROS:
  • 1.61 → 0.93
• CPC:
  • €0.50 → > €1.54
• Hook score (examples):
  • 74 vs 59
  • 77 vs 53
• Watch score:
  • 82
• CPA
  • described as very bad for the engagement-strong strategy
  • about 2x better conversion for the branded strategy
• Video engagement/attention indicators:
  • Thumbstop
  • Old rate on videos (watch/retention-related)

---

Presenters / sources

• Presenter: “I’m going to try to show you…” (no specific name provided in subtitles)
• Company/source: KODAC
• Tool/software: Motion (KODAC’s creative analysis system)
• Team referenced: “our lab team” / “our team of creative strategists” (no individual names provided)