Performance Marketing skill 08 of 13

Meta Ads (Facebook and Instagram) SKILL.md

The Meta Ads skill makes Claude, Codex or another AI agent build, audit and scale Facebook and Instagram campaigns. It fixes signal quality with the pixel and Conversions API, consolidates structure around Advantage+ with existing-customer controls, treats creative volume and diversity as targeting, adds a testing path and scaling rules, and checks results against first-party data and lift tests rather than trusting Ads Manager alone.

meta-ads/SKILL.md, version 1.0.0, 1,559 words.

Meta advertising has moved from detailed targeting to automation. Advantage+ campaigns choose audiences, placements and budgets; the delivery system picks which ad to show which person, and creative has become the main targeting signal. Practitioners widely report that Meta's newer ad retrieval systems reward many clearly different creatives over small variations of one idea. [Consensus] Meta also introduced an incremental attribution option in 2025 that optimises for conversions it estimates would not have happened otherwise. [Reported, Meta]

This skill builds, audits and scales Meta campaigns: account structure, Advantage+ settings, conversion signals with the Conversions API, creative volume and diversity, audiences and exclusions, budgets, and measurement beyond Ads Manager.

Operating principles

  1. Signal quality first. Pixel plus Conversions API, deduplicated, with good event match quality and the right optimisation event.
  2. Fewer campaigns, more data. Consolidated structures learn faster than many small ad sets.
  3. Creative is targeting. Different angles, formats and people reach different audiences.
  4. Separate new customers from existing ones in goals and reporting where the business cares about acquisition.
  5. Measure outside Ads Manager. View-through and click windows inflate results; use lift tests and your own data.

Account structure (typical e-commerce)

CampaignPurposeNotes
Advantage+ sales (prospecting)Main acquisitionExisting customer budget cap or exclusions where available; broad delivery
Testing campaignNew creative conceptsFixed budget; graduate winners to the main campaign
Retargeting (small)Cart and product viewersOnly if lift tests show value
Catalogue (dynamic product ads)Product-level personalisationNeeds a clean catalogue
Lead generation (B2B or services)Instant forms or site leadsImport qualified outcomes via CAPI

When to use, and when not to

Use it for new Meta accounts, audits, scaling plateaus, rising CPAs, creative fatigue, Advantage+ adoption and measurement disputes.

Don't use it for TikTok, LinkedIn or other social platforms (paid-social-tiktok-linkedin) or for creative production processes in depth (ad-creative-testing).

Inputs to ask for, in one message

  1. Ads Manager exports: campaigns, ad sets, ads, results and breakdowns for 90 days.
  2. Events Manager status: pixel, CAPI, deduplication, event match quality, domain verification.
  3. Business goals: new customers, revenue, leads, payback.
  4. Creative library and production capacity.
  5. Customer lists and any lift test results.

Raw SKILL.md file

.md

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

---
name: meta-ads
description: "Build, audit and scale Meta ads on Facebook and Instagram: pixel and Conversions API with deduplication and event match quality, consolidated Advantage+ structures with existing-customer controls, creative volume and diversity as targeting, testing campaigns, budgets and scaling rules, incremental attribution and lift tests, and checks against first-party data. Use whenever someone asks about Facebook ads, Instagram ads, Meta Advantage+, rising Meta CPA, creative fatigue, Meta campaign structure, CAPI, scaling Meta spend or why Ads Manager results don't match sales."
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/meta-ads/
  category: marketing
  subcategory: performance-marketing
  version: "1.0.0"
  created: "2026-10-10"
  updated: "2026-10-10"
---

# Meta Ads (Facebook and Instagram)

Meta advertising has moved from detailed targeting to automation. Advantage+ campaigns choose audiences, placements and budgets; the delivery system picks which ad to show which person, and creative has become the main targeting signal. Practitioners widely report that Meta's newer ad retrieval systems reward many clearly different creatives over small variations of one idea. [Consensus] Meta also introduced an incremental attribution option in 2025 that optimises for conversions it estimates would not have happened otherwise. [Reported, Meta]

This skill builds, audits and scales Meta campaigns: account structure, Advantage+ settings, conversion signals with the Conversions API, creative volume and diversity, audiences and exclusions, budgets, and measurement beyond Ads Manager.

## Operating principles
1. **Signal quality first.** Pixel plus Conversions API, deduplicated, with good event match quality and the right optimisation event.
2. **Fewer campaigns, more data.** Consolidated structures learn faster than many small ad sets.
3. **Creative is targeting.** Different angles, formats and people reach different audiences.
4. **Separate new customers from existing ones** in goals and reporting where the business cares about acquisition.
5. **Measure outside Ads Manager.** View-through and click windows inflate results; use lift tests and your own data.

## Account structure (typical e-commerce)
| Campaign | Purpose | Notes |
|---|---|---|
| Advantage+ sales (prospecting) | Main acquisition | Existing customer budget cap or exclusions where available; broad delivery |
| Testing campaign | New creative concepts | Fixed budget; graduate winners to the main campaign |
| Retargeting (small) | Cart and product viewers | Only if lift tests show value |
| Catalogue (dynamic product ads) | Product-level personalisation | Needs a clean catalogue |
| Lead generation (B2B or services) | Instant forms or site leads | Import qualified outcomes via CAPI |

## When to use, and when not to
Use it for new Meta accounts, audits, scaling plateaus, rising CPAs, creative fatigue, Advantage+ adoption and measurement disputes.

Don't use it for TikTok, LinkedIn or other social platforms (`paid-social-tiktok-linkedin`) or for creative production processes in depth (`ad-creative-testing`).

## Inputs to ask for, in one message
1. Ads Manager exports: campaigns, ad sets, ads, results and breakdowns for 90 days.
2. Events Manager status: pixel, CAPI, deduplication, event match quality, domain verification.
3. Business goals: new customers, revenue, leads, payback.
4. Creative library and production capacity.
5. Customer lists and any lift test results.

## Workflow

### Step 1. Fix signals
Pixel plus CAPI with deduplication; prioritised events set; improve event match quality with hashed email and phone where consented; optimise for purchase or qualified lead, not clicks.

### Step 2. Consolidate structure
Move to fewer campaigns with broad delivery; avoid audience overlap between ad sets; keep a separate testing campaign for new concepts.

### Step 3. Configure Advantage+
Set existing customer definitions with customer lists and pixel audiences; use budget caps for existing customers when acquisition is the goal; review placement and creative enhancements for brand safety.

### Step 4. Build creative volume and diversity
Aim for many distinct concepts: different problems, benefits, formats (UGC, demo, testimonial, founder, static, carousel), hooks and people. Refresh before fatigue.

### Step 5. Set budgets and bidding
Campaign budget optimisation for consolidated campaigns; cost cap or ROAS goals only when you understand the trade-off in volume; avoid frequent large changes that reset learning.

### Step 6. Read results properly
Compare Ads Manager with GA4 and your own orders; check attribution settings (click and view windows); use breakdowns for placement, age and creative.

### Step 7. Test incrementality
Conversion lift studies or geo tests for prospecting and retargeting before scaling.

### Step 8. Scale
Increase budgets in steps; add creative as you scale; watch frequency and new-customer share.

## Creative testing routine
- Launch 3 to 5 new concepts per week or fortnight in the testing campaign, depending on budget.
- Judge on cost per acquisition or new customer from your data, plus hook rate and hold rate for video.
- Move winners into the main campaign without editing them.
- Retire ads when frequency rises and results fall for two consecutive weeks.

## Creative diversity matrix
| Dimension | Options to vary |
|---|---|
| Angle | Problem, benefit, social proof, price or value, comparison, founder story, how it works |
| Format | UGC video, demo, testimonial, static image, carousel, collection, motion graphic |
| Hook (first 3 seconds) | Question, bold claim with proof, surprising visual, result first |
| Person | Customer, creator, expert, founder, no person |
| Length | 6 to 15 seconds, 15 to 30 seconds, longer explainers |
| Offer | No offer, bundle, free shipping, limited-time (sparingly) |
A concept is a new combination of angle and format, not a new colour or caption on the same video.

## Reading Ads Manager without fooling yourself
- Check the attribution setting (for example 7-day click and 1-day view) and compare click-only results.
- Compare purchases and revenue with your commerce data for the same period.
- Separate new and returning customers in your own data.
- Treat view-through conversions as a hypothesis to test, not proof.
- Use breakdowns to find what works, then confirm with lift tests before shifting large budgets.

## Scaling rules
1. Raise budgets by about 20% at a time, a few days apart, unless a test shows strong headroom.
2. Add new creative as you scale; frequency rises fast with fixed creative.
3. Watch new-customer CPA and blended MER, not only platform ROAS.
4. If CPA rises for a week after a change, hold rather than reverse immediately; learning needs time.
5. Log every major change with the date for later analysis.

## Lead generation on Meta
- Use higher-intent instant forms (with a review step and qualifying questions) or send traffic to a fast site form.
- Send CRM outcomes back via CAPI so delivery learns which leads become customers.
- Remove fields that add friction without improving qualification.

## Edge cases and judgement calls
| Situation | What to do |
|---|---|
| Low budget | One campaign, few ads, broad delivery; give learning time |
| Lead quality poor | Higher-intent forms, qualifying questions, CAPI with CRM outcomes |
| Special ad categories (housing, employment, credit, politics) | Follow restricted targeting rules and authorisations |
| Brand safety concerns | Review placements and inventory filters; exclude where necessary |
| Creative production bottleneck | Use customer UGC, creators, simple statics and iterations of winning angles |

## Handoffs to related skills
- Tracking and CAPI: `conversion-tracking-capi`
- Creative concepts and testing: `ad-creative-testing`
- Audiences and customer lists: `audience-first-party-data`
- Lift and geo tests: `incrementality-testing`
- Landing pages: `landing-page-cro`

## Worked example
A skincare brand ran 22 Meta campaigns with interest targeting and five ads each. CPA had risen for months. The rebuild created one Advantage+ sales campaign with an existing-customer budget cap, a testing campaign launching four new concepts a week (creator demos, dermatologist explainer, before-and-after with disclaimers, founder story, static benefit cards), CAPI with deduplication raising event match quality, and a geo test on retargeting. CPA for new customers fell, retargeting proved weakly incremental and was cut back, and budget scaled in 20% steps while new creative kept frequency in check.

## Output format
```
# Meta Ads plan: <account>, <date>
## Signal health (pixel, CAPI, dedup, EMQ, events)
## Structure and Advantage+ settings
## Creative plan (concepts, formats, cadence)
## Budgets, bidding and scaling rules
## Measurement (Ads Manager vs own data, lift tests)
## Monitoring routine
```

## Common mistakes
- Many small ad sets fighting each other.
- Too few, too similar creatives.
- Optimising for clicks or landing page views.
- Counting existing customers as acquisition.
- Trusting view-through results without tests.

## Questions to ask before you change the account
- Is the goal new customers, revenue or leads, and how is each measured?
- How much creative can the team produce each week?
- What does your own data say compared with Ads Manager?

## Self-check before you answer
- Are pixel and CAPI deduplicated with good match quality?
- Is the structure consolidated with a testing path?
- Is creative diverse and refreshed?
- Are results checked against first-party data and tests?

## Official documentation
- Meta Conversions API: https://developers.facebook.com/docs/marketing-api/conversions-api
- Meta Business Help Centre: https://www.facebook.com/business/help
- Meta Advantage+ overview: https://www.facebook.com/business/ads/meta-advantage-plus

---
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/meta-ads/

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. Fix signals

Pixel plus CAPI with deduplication; prioritised events set; improve event match quality with hashed email and phone where consented; optimise for purchase or qualified lead, not clicks.

Step 2. Consolidate structure

Move to fewer campaigns with broad delivery; avoid audience overlap between ad sets; keep a separate testing campaign for new concepts.

Step 3. Configure Advantage+

Set existing customer definitions with customer lists and pixel audiences; use budget caps for existing customers when acquisition is the goal; review placement and creative enhancements for brand safety.

Step 4. Build creative volume and diversity

Aim for many distinct concepts: different problems, benefits, formats (UGC, demo, testimonial, founder, static, carousel), hooks and people. Refresh before fatigue.

Step 5. Set budgets and bidding

Campaign budget optimisation for consolidated campaigns; cost cap or ROAS goals only when you understand the trade-off in volume; avoid frequent large changes that reset learning.

Step 6. Read results properly

Compare Ads Manager with GA4 and your own orders; check attribution settings (click and view windows); use breakdowns for placement, age and creative.

Step 7. Test incrementality

Conversion lift studies or geo tests for prospecting and retargeting before scaling.

Step 8. Scale

Increase budgets in steps; add creative as you scale; watch frequency and new-customer share.

Creative testing routine

  • Launch 3 to 5 new concepts per week or fortnight in the testing campaign, depending on budget.
  • Judge on cost per acquisition or new customer from your data, plus hook rate and hold rate for video.
  • Move winners into the main campaign without editing them.
  • Retire ads when frequency rises and results fall for two consecutive weeks.

Creative diversity matrix

DimensionOptions to vary
AngleProblem, benefit, social proof, price or value, comparison, founder story, how it works
FormatUGC video, demo, testimonial, static image, carousel, collection, motion graphic
Hook (first 3 seconds)Question, bold claim with proof, surprising visual, result first
PersonCustomer, creator, expert, founder, no person
Length6 to 15 seconds, 15 to 30 seconds, longer explainers
OfferNo offer, bundle, free shipping, limited-time (sparingly)

A concept is a new combination of angle and format, not a new colour or caption on the same video.

Reading Ads Manager without fooling yourself

  • Check the attribution setting (for example 7-day click and 1-day view) and compare click-only results.
  • Compare purchases and revenue with your commerce data for the same period.
  • Separate new and returning customers in your own data.
  • Treat view-through conversions as a hypothesis to test, not proof.
  • Use breakdowns to find what works, then confirm with lift tests before shifting large budgets.

Scaling rules

  1. Raise budgets by about 20% at a time, a few days apart, unless a test shows strong headroom.
  2. Add new creative as you scale; frequency rises fast with fixed creative.
  3. Watch new-customer CPA and blended MER, not only platform ROAS.
  4. If CPA rises for a week after a change, hold rather than reverse immediately; learning needs time.
  5. Log every major change with the date for later analysis.

Lead generation on Meta

  • Use higher-intent instant forms (with a review step and qualifying questions) or send traffic to a fast site form.
  • Send CRM outcomes back via CAPI so delivery learns which leads become customers.
  • Remove fields that add friction without improving qualification.

Edge cases and judgement calls

SituationWhat to do
Low budgetOne campaign, few ads, broad delivery; give learning time
Lead quality poorHigher-intent forms, qualifying questions, CAPI with CRM outcomes
Special ad categories (housing, employment, credit, politics)Follow restricted targeting rules and authorisations
Brand safety concernsReview placements and inventory filters; exclude where necessary
Creative production bottleneckUse customer UGC, creators, simple statics and iterations of winning angles
  • Tracking and CAPI: conversion-tracking-capi
  • Creative concepts and testing: ad-creative-testing
  • Audiences and customer lists: audience-first-party-data
  • Lift and geo tests: incrementality-testing
  • Landing pages: landing-page-cro

Worked example

A skincare brand ran 22 Meta campaigns with interest targeting and five ads each. CPA had risen for months. The rebuild created one Advantage+ sales campaign with an existing-customer budget cap, a testing campaign launching four new concepts a week (creator demos, dermatologist explainer, before-and-after with disclaimers, founder story, static benefit cards), CAPI with deduplication raising event match quality, and a geo test on retargeting. CPA for new customers fell, retargeting proved weakly incremental and was cut back, and budget scaled in 20% steps while new creative kept frequency in check.

Output format

# Meta Ads plan: <account>, <date>
## Signal health (pixel, CAPI, dedup, EMQ, events)
## Structure and Advantage+ settings
## Creative plan (concepts, formats, cadence)
## Budgets, bidding and scaling rules
## Measurement (Ads Manager vs own data, lift tests)
## Monitoring routine

Common mistakes

  • Many small ad sets fighting each other.
  • Too few, too similar creatives.
  • Optimising for clicks or landing page views.
  • Counting existing customers as acquisition.
  • Trusting view-through results without tests.

Questions to ask before you change the account

  • Is the goal new customers, revenue or leads, and how is each measured?
  • How much creative can the team produce each week?
  • What does your own data say compared with Ads Manager?

Self-check before you answer

  • Are pixel and CAPI deduplicated with good match quality?
  • Is the structure consolidated with a testing path?
  • Is creative diverse and refreshed?
  • Are results checked against first-party data and tests?

Official documentation

When to use this skill, and when not to

Use it for new Meta accounts, audits, scaling plateaus, rising CPAs, creative fatigue, Advantage+ adoption and measurement disputes.

Meta has moved from detailed targeting to automation. Advantage+ chooses audiences, placements and budgets, and creative has become the main targeting signal; practitioners widely report that newer delivery systems reward many clearly different creatives over small variations. Meta also introduced an incremental attribution option in 2025 that optimises for conversions it estimates wouldn’t otherwise happen.

Don’t use it for TikTok or LinkedIn (use TikTok, LinkedIn and Other Paid Social) or creative production systems in depth (use Ad Creative Testing).

Example prompts

  • “Our Meta CPA rose for six months. Rebuild the account.”
  • “Set up the Conversions API with deduplication.”
  • “How should we structure Advantage+ for new-customer growth?”
  • “Plan a weekly creative testing routine.”
  • “Ads Manager shows far more purchases than Shopify. What’s real?”

Example output

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

# Meta Ads plan: example skincare brand, 10 October 2026

## Signals
Pixel + CAPI, event_id dedup, event match quality improved with hashed email/phone

## Structure
| Campaign | Purpose |
| Advantage+ sales | Acquisition, existing-customer budget cap |
| Testing | 4 new concepts per week |
| Retargeting (small) | Pending geo test |

## Creative mix
Creator demos, dermatologist explainer, founder story, static benefit cards

## Scaling: +20% steps, new creative each week

Ads Manager results are always compared with your own orders and new customers, and big shifts wait for lift or geo tests.

What the agent needs from you

The skill asks for 90 days of Ads Manager exports with breakdowns, Events Manager status (pixel, CAPI, deduplication, event match quality, domain verification), business goals, your creative library and production capacity, customer lists and any lift test results.

How this skill hands off to related skills

Meta work connects to the set. Tracking comes from Conversion Tracking, creative from Ad Creative Testing, customer lists from First-Party Audiences, lift tests from Incrementality Testing and landing pages from Landing Page CRO.

How to run Meta Ads step by step

  1. Install the skill as meta-ads/SKILL.md.
  2. Check Events Manager: pixel, CAPI, deduplication and event match quality.
  3. Export 90 days of campaigns, ad sets, ads and breakdowns.
  4. Share goals, customer lists and creative capacity.
  5. Review structure, Advantage+ settings, the creative plan and scaling rules.
  6. Compare results with your own data and test incrementality before big shifts.

Situations this skill handles

Meta accounts stall for familiar reasons. The skill handles these situations:

SituationWhat the skill does
Many small ad setsConsolidates into fewer campaigns with broad delivery
Few, similar creativesBuilds a diversity matrix of angles and formats
Optimising for clicksSwitches to purchase or qualified lead events
Existing customers counted as newSets customer definitions and budget caps
Poor lead qualityHigher-intent forms and CRM outcomes via CAPI
Special ad categoriesFollows restricted targeting rules

What makes this skill different

Most Meta advice still talks about interest targeting and audience stacking. This skill works with how delivery now operates: clean signals, consolidated structure, creative diversity as the main lever and existing-customer controls so acquisition budgets acquire.

It also refuses to grade Meta on Meta’s numbers alone, checking results against your own sales and using lift tests before scaling.

Practical tips from real projects

  • Signals first: pixel plus CAPI, deduplicated.
  • A concept is a new angle and format, not a new caption.
  • Graduate winners unedited from testing to scaling.
  • Scale in steps and log every change.

A worked example: a skincare brand ran 22 Meta campaigns with interest targeting and five ads each, and CPA had risen for months. One Advantage+ sales campaign with an existing-customer cap, a testing campaign launching four new concepts a week, CAPI with deduplication and a geo test on retargeting cut new-customer CPA, showed retargeting was weakly incremental and let budget scale in 20% steps while new creative kept frequency in check.

Creative diversity is built from a matrix: angle (problem, benefit, proof, value, comparison, founder story, how it works), format (UGC, demo, testimonial, static, carousel, motion), hook, person, length and offer. Each new concept changes at least the angle or format, which gives delivery clearly different signals to work with.

Reading Ads Manager safely means checking the attribution setting, comparing click-only results, matching purchases with commerce data, separating new from returning customers and treating view-through conversions as hypotheses to test.

For lead generation, higher-intent instant forms with a review step and qualifying questions, or fast site forms, plus CRM outcomes sent back through CAPI, teach delivery which leads become customers.

Scaling rules keep growth stable: raise budgets about 20% at a time, add creative as you scale, watch new-customer CPA and blended MER rather than platform ROAS alone, hold steady for a week after changes before reversing, and log every major change.

A typical e-commerce structure has four parts. An Advantage+ sales campaign handles most acquisition, with existing customers capped or excluded where the goal is growth. A separate testing campaign launches new concepts on a fixed budget and graduates winners. A small retargeting campaign for cart and product viewers runs only if lift tests show it adds sales. A catalogue campaign with dynamic product ads personalises by product once the catalogue is clean. Lead businesses add lead campaigns that import qualified outcomes through CAPI.

Creative testing follows a routine: three to five new concepts every week or fortnight depending on budget, judged on cost per acquisition or new customer from your own data plus hook and hold rates for video; winners move into the main campaign unedited; ads retire when frequency rises and results fall for two weeks in a row. Production bottlenecks are eased with customer UGC, creators, simple statics and iterations on winning angles.

Frequently asked questions

Fewer campaigns with broad delivery, usually an Advantage+ sales campaign for acquisition, a testing campaign for new concepts and small, tested retargeting where it proves incremental.

Automated delivery uses the ad itself to find who responds. Different angles, formats and people reach different audiences, so creative diversity works like targeting.

Yes for most advertisers. It sends events from your server alongside the pixel, recovering signal lost to browsers and ad blockers, with deduplication and consent.

View-through credit, overlapping attribution with other platforms and modelled conversions. Compare with your own orders and test incrementality.

In steps of about 20% a few days apart, adding new creative as you go, unless a test shows strong headroom.

Test it. It often shows high ROAS but low incrementality in geo tests.

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.