Performance Marketing skill 10 of 13

Ad Creative Strategy and Testing SKILL.md

The Ad Creative Strategy and Testing skill makes Claude, Codex or another AI agent build a creative system for Meta, TikTok, YouTube, Google and LinkedIn. It mines angles from reviews, calls and surveys, tests distinct concepts before small iterations, writes hooks and scripts, reads stage metrics from thumb-stop and hold rate to CTR, conversion and CAC, writes creative briefs and keeps a tagged library with substantiated, compliant claims.

ad-creative-testing/SKILL.md, version 1.0.0, 1,564 words.

As platforms automate targeting and bidding, creative has become the biggest lever advertisers still control. The ad decides who stops scrolling, who the algorithm shows it to next, and whether the click turns into a customer. Most teams test small variations of one idea and wonder why results plateau. The teams that grow test distinct concepts, learn which angles work for which audiences and build a production system that keeps fresh, on-brand creative flowing.

This skill builds a creative strategy grounded in customer insight, plans concept tests across Meta, TikTok, YouTube, Google and LinkedIn, sets metrics for each stage of an ad, and turns results into a repeatable creative system.

Operating principles

  1. Start with customers, not formats. Reviews, sales calls, support tickets and surveys hold the angles that sell.
  2. Test concepts, then iterate. Big differences first (angle, format, hook), small tweaks later.
  3. Measure each stage. Thumb-stop and hold rates for attention, click-through for interest, conversion and CAC for business value.
  4. Diversity feeds automation. Different concepts reach different people in automated delivery systems.
  5. Claims must be true. Substantiate performance claims, show real results and follow platform and advertising rules.

Where angles come from

SourceWhat to look for
Customer reviews (yours and competitors')Exact phrases about problems, outcomes and surprises
Sales and support callsObjections, questions, moments of "that's what I need"
SurveysWhy they bought, what nearly stopped them, alternatives considered
Search termsHow people describe the problem
Comments on ads and postsDoubts and enthusiasm in their words
Competitor ad librariesAngles already saturated, gaps nobody uses

When to use, and when not to

Use it when performance plateaus, when creative fatigue raises costs, when launching a product or market, when building a creative testing programme, and when briefing creators or agencies.

Don't use it for landing page optimisation (landing-page-cro) or for platform setup and bidding (the channel skills).

Inputs to ask for, in one message

  1. Product, audience segments and positioning.
  2. Customer research sources: reviews, surveys, call notes, support tickets.
  3. Current ads with performance by creative for 90 days.
  4. Brand guidelines and claims that can or can't be made.
  5. Production capacity: in-house, creators, agencies, budget.

Raw SKILL.md file

.md

This is the exact file agents read, including the credit header and licence notice. Save it as ad-creative-testing/SKILL.md.

---
name: ad-creative-testing
description: "Build an ad creative strategy and testing system for Meta, TikTok, YouTube, Google and LinkedIn: angles mined from reviews, calls and surveys, concept-level tests before iterations, hooks and scripts, stage metrics from thumb-stop and hold rate to CTR, conversion and CAC, creative briefs, a tagged library and substantiated, compliant claims. Use whenever someone asks about ad creative, creative testing, UGC ads, hooks, creative fatigue, ad concepts, creative briefs, why ads stopped working or how to scale creative production."
license: CC-BY-4.0
metadata:
  author: Afeez Adebayo
  author_url: https://afeezadebayo.com
  author_linkedin: https://www.linkedin.com/in/afeez-adebayo/
  author_x: https://x.com/za_icon
  source: https://afeezadebayo.com/skills-md/marketing/performance-marketing/ad-creative-testing/
  category: marketing
  subcategory: performance-marketing
  version: "1.0.0"
  created: "2026-10-10"
  updated: "2026-10-10"
---

# Ad Creative Strategy and Testing

As platforms automate targeting and bidding, creative has become the biggest lever advertisers still control. The ad decides who stops scrolling, who the algorithm shows it to next, and whether the click turns into a customer. Most teams test small variations of one idea and wonder why results plateau. The teams that grow test distinct concepts, learn which angles work for which audiences and build a production system that keeps fresh, on-brand creative flowing.

This skill builds a creative strategy grounded in customer insight, plans concept tests across Meta, TikTok, YouTube, Google and LinkedIn, sets metrics for each stage of an ad, and turns results into a repeatable creative system.

## Operating principles
1. **Start with customers, not formats.** Reviews, sales calls, support tickets and surveys hold the angles that sell.
2. **Test concepts, then iterate.** Big differences first (angle, format, hook), small tweaks later.
3. **Measure each stage.** Thumb-stop and hold rates for attention, click-through for interest, conversion and CAC for business value.
4. **Diversity feeds automation.** Different concepts reach different people in automated delivery systems.
5. **Claims must be true.** Substantiate performance claims, show real results and follow platform and advertising rules.

## Where angles come from
| Source | What to look for |
|---|---|
| Customer reviews (yours and competitors') | Exact phrases about problems, outcomes and surprises |
| Sales and support calls | Objections, questions, moments of "that's what I need" |
| Surveys | Why they bought, what nearly stopped them, alternatives considered |
| Search terms | How people describe the problem |
| Comments on ads and posts | Doubts and enthusiasm in their words |
| Competitor ad libraries | Angles already saturated, gaps nobody uses |

## When to use, and when not to
Use it when performance plateaus, when creative fatigue raises costs, when launching a product or market, when building a creative testing programme, and when briefing creators or agencies.

Don't use it for landing page optimisation (`landing-page-cro`) or for platform setup and bidding (the channel skills).

## Inputs to ask for, in one message
1. Product, audience segments and positioning.
2. Customer research sources: reviews, surveys, call notes, support tickets.
3. Current ads with performance by creative for 90 days.
4. Brand guidelines and claims that can or can't be made.
5. Production capacity: in-house, creators, agencies, budget.

## Workflow

### Step 1. Mine insights
Pull 50 to 100 customer phrases; group them into problems, desired outcomes, objections and proof points.

### Step 2. Write angles
Each angle is one reason to buy for one type of customer: "for busy parents, dinner in 15 minutes", "for sceptics, a dermatologist explains the ingredient".

### Step 3. Pick formats per platform
UGC and creator video, demos, testimonials, founder stories, statics, carousels, motion graphics; vertical for TikTok, Reels and Shorts.

### Step 4. Write hooks and scripts
Hook in the first 2 to 3 seconds (problem, result first, bold claim with proof, question). Body shows proof. End with a clear call to action.

### Step 5. Plan the test
A testing campaign or budget with several new concepts at once, judged on the stage metrics and CAC from your own data.

### Step 6. Read results by stage
| Stage | Metric | What it tells you |
|---|---|---|
| Attention | Thumb-stop rate (3-second views ÷ impressions), hook rate | Does the opening stop people? |
| Retention | Hold rate (views to 50% or 15 seconds ÷ 3-second views) | Is the story engaging? |
| Interest | Click-through rate | Does it make people want more? |
| Conversion | Conversion rate, CAC, new-customer share | Does it bring the right buyers? |

### Step 7. Iterate winners
New hooks, lengths, people and formats on winning angles; retire losers.

### Step 8. Build the library
Tag every ad by angle, format, hook, person and result, so the team learns which combinations work.

## Creative brief template
```
Product and audience segment:
Angle (one reason to buy):
Insight behind it (customer quote):
Format and platform:
Hook options (3):
Proof to show:
Call to action:
Mandatory claims and disclaimers; claims to avoid:
Length and specs:
Success metrics:
```

## Hook formulas
| Hook type | Example |
|---|---|
| Problem call-out | "If you dread 6pm because nobody knows what's for dinner..." |
| Result first | "This is what my skin looked like after 30 days" (with real, consented footage) |
| Myth-buster | "You don't need eight products for clear skin" |
| Question | "Still paying for three project tools?" |
| Demonstration | Product doing the job in the first second |
| Social proof | "Why 10,000 teachers switched this term" (only if true and sourced) |

## Concept test design
1. Pick 3 to 6 concepts that differ in angle or format.
2. Give each similar budget and time in a testing campaign.
3. Wait for enough conversions to compare (or use stage metrics early for clear losers).
4. Promote winners unedited to scaling campaigns.
5. Iterate winners on hook, length, person and format.
6. Tag everything in the library.

## Creative library tags
| Tag | Values |
|---|---|
| Angle | decision fatigue, food waste, family, price, quality |
| Format | UGC, demo, testimonial, founder, static, carousel |
| Hook type | problem, result, myth, question, demo, proof |
| Person | customer, creator, expert, founder, none |
| Length | under 15s, 15-30s, over 30s |
| Result | CAC, new-customer share, hold rate, CTR |
Review the library monthly to find patterns: which angles win for which segments, and how long winners last before fatigue.

## Fatigue signals
- Frequency rising while CTR and conversion rate fall for two weeks.
- Hook rate dropping on a previously strong video.
- CPA rising in the campaign without changes elsewhere.
Refresh with new hooks or formats on the same angle before retiring the angle entirely.

## Edge cases and judgement calls
| Situation | What to do |
|---|---|
| Tiny budget | Test fewer concepts at a time; use statics and simple UGC |
| Regulated claims (health, finance) | Substantiate claims, include required disclaimers, get approvals |
| Brand team resists UGC style | Test it alongside brand creative and show the data |
| Creative fatigue in days (TikTok) | Plan production for weekly refreshes |
| AI-generated images or video | Follow platform disclosure rules; don't use them to fake results or testimonials |

## Handoffs to related skills
- Platform execution: `meta-ads`, `paid-social-tiktok-linkedin`, `google-search-ads`, `performance-max-shopping`
- Landing page match: `landing-page-cro`
- Audience insight: `audience-first-party-data`
- Reporting creative results: `paid-media-reporting`

## Worked example
A meal-kit brand's Meta ads had used the same polished video for months. Mining 400 reviews surfaced three strong angles: no more "what's for dinner" decisions, less food waste and cooking with kids. The team produced 12 concepts across UGC, recipe demos and static benefit cards. Two angles (decision fatigue and cooking with kids) beat the old ad on CAC; UGC from parents performed best for the kids angle. Iterations with new hooks extended their life, and the library tags showed problem-first hooks consistently outperformed product-first hooks for this audience.

## Output format
```
# Creative strategy: <brand>, <date>
## Customer insights (grouped quotes)
## Angles by segment
## Concept plan (angle x format x platform)
## Hooks and scripts for priority concepts
## Test design and stage metrics
## Library tagging and iteration rules
```

## Common mistakes
- Testing button colours while concepts never change.
- Judging creative only on CTR.
- No link between ad promise and landing page.
- Making claims you can't substantiate.
- No tagging, so lessons are lost.

## Questions to ask before you plan
- What do customers say in their own words about why they bought?
- Which claims can you prove?
- How many new concepts can you produce each week?

## Self-check before you answer
- Are angles grounded in customer language?
- Is the test plan concept-level first, with stage metrics?
- Are results judged on CAC from first-party data?
- Are claims substantiated and compliant?

## Official documentation and tools
- Meta Ad Library: https://www.facebook.com/ads/library/
- TikTok Creative Center: https://ads.tiktok.com/business/creativecenter/
- Google Ads Transparency Center: https://adstransparency.google.com/
- UK ASA advertising codes: https://www.asa.org.uk/codes-and-rulings/advertising-codes.html

---
Created by Afeez Adebayo, Marketing, SEO & Growth Expert
Website: https://afeezadebayo.com
LinkedIn: https://www.linkedin.com/in/afeez-adebayo/
X: https://x.com/za_icon
Source: https://afeezadebayo.com/skills-md/marketing/performance-marketing/ad-creative-testing/

Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). You may copy, share and adapt this skill, including for commercial use, as long as you credit Afeez Adebayo and link to the source URL above. Keep this notice in any copy or adapted version.

Workflow

Step 1. Mine insights

Pull 50 to 100 customer phrases; group them into problems, desired outcomes, objections and proof points.

Step 2. Write angles

Each angle is one reason to buy for one type of customer: "for busy parents, dinner in 15 minutes", "for sceptics, a dermatologist explains the ingredient".

Step 3. Pick formats per platform

UGC and creator video, demos, testimonials, founder stories, statics, carousels, motion graphics; vertical for TikTok, Reels and Shorts.

Step 4. Write hooks and scripts

Hook in the first 2 to 3 seconds (problem, result first, bold claim with proof, question). Body shows proof. End with a clear call to action.

Step 5. Plan the test

A testing campaign or budget with several new concepts at once, judged on the stage metrics and CAC from your own data.

Step 6. Read results by stage

StageMetricWhat it tells you
AttentionThumb-stop rate (3-second views ÷ impressions), hook rateDoes the opening stop people?
RetentionHold rate (views to 50% or 15 seconds ÷ 3-second views)Is the story engaging?
InterestClick-through rateDoes it make people want more?
ConversionConversion rate, CAC, new-customer shareDoes it bring the right buyers?

Step 7. Iterate winners

New hooks, lengths, people and formats on winning angles; retire losers.

Step 8. Build the library

Tag every ad by angle, format, hook, person and result, so the team learns which combinations work.

Creative brief template

Product and audience segment:
Angle (one reason to buy):
Insight behind it (customer quote):
Format and platform:
Hook options (3):
Proof to show:
Call to action:
Mandatory claims and disclaimers; claims to avoid:
Length and specs:
Success metrics:

Hook formulas

Hook typeExample
Problem call-out"If you dread 6pm because nobody knows what's for dinner..."
Result first"This is what my skin looked like after 30 days" (with real, consented footage)
Myth-buster"You don't need eight products for clear skin"
Question"Still paying for three project tools?"
DemonstrationProduct doing the job in the first second
Social proof"Why 10,000 teachers switched this term" (only if true and sourced)

Concept test design

  1. Pick 3 to 6 concepts that differ in angle or format.
  2. Give each similar budget and time in a testing campaign.
  3. Wait for enough conversions to compare (or use stage metrics early for clear losers).
  4. Promote winners unedited to scaling campaigns.
  5. Iterate winners on hook, length, person and format.
  6. Tag everything in the library.

Creative library tags

TagValues
Angledecision fatigue, food waste, family, price, quality
FormatUGC, demo, testimonial, founder, static, carousel
Hook typeproblem, result, myth, question, demo, proof
Personcustomer, creator, expert, founder, none
Lengthunder 15s, 15-30s, over 30s
ResultCAC, new-customer share, hold rate, CTR

Review the library monthly to find patterns: which angles win for which segments, and how long winners last before fatigue.

Fatigue signals

  • Frequency rising while CTR and conversion rate fall for two weeks.
  • Hook rate dropping on a previously strong video.
  • CPA rising in the campaign without changes elsewhere.

Refresh with new hooks or formats on the same angle before retiring the angle entirely.

Edge cases and judgement calls

SituationWhat to do
Tiny budgetTest fewer concepts at a time; use statics and simple UGC
Regulated claims (health, finance)Substantiate claims, include required disclaimers, get approvals
Brand team resists UGC styleTest it alongside brand creative and show the data
Creative fatigue in days (TikTok)Plan production for weekly refreshes
AI-generated images or videoFollow platform disclosure rules; don't use them to fake results or testimonials
  • Platform execution: meta-ads, paid-social-tiktok-linkedin, google-search-ads, performance-max-shopping
  • Landing page match: landing-page-cro
  • Audience insight: audience-first-party-data
  • Reporting creative results: paid-media-reporting

Worked example

A meal-kit brand's Meta ads had used the same polished video for months. Mining 400 reviews surfaced three strong angles: no more "what's for dinner" decisions, less food waste and cooking with kids. The team produced 12 concepts across UGC, recipe demos and static benefit cards. Two angles (decision fatigue and cooking with kids) beat the old ad on CAC; UGC from parents performed best for the kids angle. Iterations with new hooks extended their life, and the library tags showed problem-first hooks consistently outperformed product-first hooks for this audience.

Output format

# Creative strategy: <brand>, <date>
## Customer insights (grouped quotes)
## Angles by segment
## Concept plan (angle x format x platform)
## Hooks and scripts for priority concepts
## Test design and stage metrics
## Library tagging and iteration rules

Common mistakes

  • Testing button colours while concepts never change.
  • Judging creative only on CTR.
  • No link between ad promise and landing page.
  • Making claims you can't substantiate.
  • No tagging, so lessons are lost.

Questions to ask before you plan

  • What do customers say in their own words about why they bought?
  • Which claims can you prove?
  • How many new concepts can you produce each week?

Self-check before you answer

  • Are angles grounded in customer language?
  • Is the test plan concept-level first, with stage metrics?
  • Are results judged on CAC from first-party data?
  • Are claims substantiated and compliant?

Official documentation and tools

When to use this skill, and when not to

Use it when performance plateaus, when creative fatigue raises costs, when launching a product or market, when building a creative testing programme and when briefing creators or agencies.

As platforms automate targeting and bidding, creative is the biggest lever advertisers still control. It decides who stops scrolling, who the algorithm shows the ad to next and whether the click becomes a customer. Teams that test distinct concepts and learn which angles work for which audiences keep growing; teams that test button colours plateau.

Don’t use it for landing pages (use Landing Page CRO) or platform setup (use Meta Ads or Google Search Ads).

Example prompts

  • “Mine our reviews for ad angles and plan a concept test.”
  • “Our best ad has run for 18 months. What do we test next?”
  • “Write hooks and scripts for three UGC concepts.”
  • “Create a creative brief template for our creators.”
  • “Set up a creative library with tags so we learn from tests.”

Example output

The skill returns a fixed structure. A shortened example for an imaginary site:

# Creative strategy: example meal kit, 10 October 2026

## Angles from 400 reviews
1. No more "what's for dinner" decisions (busy parents)
2. Less food waste (eco-minded)
3. Cooking with kids (families)

## Concept plan (12 concepts)
| Angle | Formats | Hooks |
| Decision fatigue | UGC, static | Problem call-out |
| Cooking with kids | Parent UGC, demo | Result first |

## Stage metrics
Thumb-stop, hold rate, CTR, CAC from Shopify

Every angle starts from a customer quote, and every claim must be one the business can prove.

What the agent needs from you

The skill asks for your product, audience segments and positioning, customer research sources such as reviews, surveys, call notes and support tickets, current ads with performance by creative for 90 days, brand guidelines and claims that can or can’t be made, and your production capacity.

How this skill hands off to related skills

Creative feeds every channel. Execution happens in Meta Ads, TikTok and LinkedIn, Google Search and Performance Max; the promise continues on landing pages; audience insight comes from First-Party Audiences; and results go into Paid Media Reporting.

How to run Ad Creative Strategy and Testing step by step

  1. Install the skill as ad-creative-testing/SKILL.md.
  2. Gather customer research: reviews, surveys, call notes and support tickets.
  3. Export creative performance for 90 days.
  4. Share brand rules and claims you can prove.
  5. Review angles, concepts, hooks and the test plan.
  6. Launch concept tests, iterate winners and tag everything in the library.

Situations this skill handles

Creative problems have different causes. The skill handles these situations:

SituationWhat the skill does
One ad running for monthsBuilds new angles from customer research
Small budgetFewer concepts at a time, statics and simple UGC
Regulated claimsSubstantiation, disclaimers and approvals
Brand team resists UGCTests UGC beside brand creative and shows the data
TikTok fatigue in daysPlans weekly production
Lessons keep getting lostTagged library reviewed monthly

What makes this skill different

Most creative testing compares small variations of one idea. This skill tests distinct concepts first, built from customer language, reads results at every stage from attention to CAC and only then iterates on winners.

Its tagged library turns tests into knowledge: which angles win for which segments, which hooks work for which formats and how long winners last before fatigue.

Practical tips from real projects

  • Concepts first, iterations second.
  • Judge on CAC from your own data, not CTR alone.
  • Match the ad promise on the landing page.
  • Only make claims you can prove.

A worked example: a meal-kit brand had used the same polished video for months. Mining 400 reviews surfaced three angles: no more dinner decisions, less food waste and cooking with kids. Twelve concepts across UGC, recipe demos and static benefit cards followed. Two angles beat the old ad on CAC, parent UGC performed best for the kids angle, and library tags showed problem-first hooks consistently beat product-first hooks for this audience.

Hook formulas give writers a starting point: problem call-outs, result first with real and consented footage, myth-busters, questions, demonstrations in the first second and social proof that is true and sourced. Each hook is tested against the same angle so the team learns what stops this audience.

Stage metrics explain where an ad wins or loses. Thumb-stop rate (three-second views divided by impressions) shows whether the opening works. Hold rate shows whether the story keeps attention. Click-through rate shows interest. Conversion rate, CAC and new-customer share show business value. An ad with a great hook and weak hold needs a better middle, not a new opening.

The brief template covers product and segment, angle, the customer insight behind it, format and platform, three hook options, proof to show, call to action, mandatory claims and disclaimers, length and specifications and success metrics, so creators and agencies start from the same page.

Fatigue is managed rather than suffered: refresh winners with new hooks, lengths, people or formats on the same angle before retiring the angle altogether.

Insight sources each add something different. Customer reviews, yours and competitors’, hold the exact phrases people use about problems, outcomes and surprises. Sales and support calls reveal objections and the moment someone says “that’s what I need”. Surveys explain why people bought and what nearly stopped them. Search terms show how people describe the problem. Comments on ads show doubts and enthusiasm in their own words. Competitor ad libraries show which angles are already saturated and where nobody is speaking. Grouping 50 to 100 phrases into problems, outcomes, objections and proof points gives a bank of angles that lasts months.

Frequently asked questions

A distinct combination of angle (one reason to buy) and format (UGC, demo, static and so on). A new caption on the same video is an iteration, not a new concept.

Three to six distinct concepts at a time is common, depending on budget, with winners iterated afterwards.

Thumb-stop and hook rate for attention, hold rate for retention, CTR for interest and conversion rate and CAC from your own data for business value.

Customers: reviews, sales calls, support tickets, surveys and comments, in their own words.

Frequency rising while CTR and conversion rate fall for two weeks, hook rate dropping and CPA rising without other changes.

Yes within platform disclosure rules, but never to fake results, testimonials or evidence.

Changelog

  • 1.0.0, 10 October 2026: first release.

Use, adapt and share this skill under CC BY 4.0: credit Afeez Adebayo and link to this page.