Growth-Today/claude-skills / gt-meta-ads
gt-meta-ads
VerifiedGT Meta Ads — expert Meta (Facebook and Instagram) Ads strategist for B2B companies, built by Growth Today (growthtoday.co). Use when asking about Meta ads, Facebook ads, Instagram ads, Advantage+, custom audiences, lookalike audiences, creative fatigue, Pixel and Conversions API (CAPI), learning phase, Meta funnel structure, retargeting, Instant Forms, Meta lead gen, ad creative strategy, frequency capping, or B2B paid social on Meta. Triggers on: Meta ads, Facebook ads, Instagram ads, Advantage+, custom audience, lookalike, creative fatigue, CAPI, Conversions API, Pixel, learning phase, Instant Forms, Opportunity Score, offline conversions, Meta retargeting, frequency cap, Reels ads. Do NOT use for LinkedIn Ads (use gt-linkedin-ads), Google Ads (use gt-google-ads), or organic social content. More GT skills: growthtoday.co.
| File | Kind | Size |
|---|---|---|
| .claude/skills/audiences/gt-SKILL.md | markdown | 4.2 KB |
| .claude/skills/audit/gt-SKILL.md | markdown | 5.0 KB |
| .claude/skills/campaign-setup/gt-SKILL.md | markdown | 4.8 KB |
| .claude/skills/creative-fatigue/gt-SKILL.md | markdown | 3.9 KB |
| .claude/skills/creative/gt-SKILL.md | markdown | 5.1 KB |
| .claude/skills/lead-forms/gt-SKILL.md | markdown | 4.5 KB |
| .claude/skills/learning-phase/gt-SKILL.md | markdown | 3.7 KB |
| .claude/skills/measurement/gt-SKILL.md | markdown | 5.6 KB |
| .claude/skills/tracking/gt-SKILL.md | markdown | 3.9 KB |
| CHANGELOG.md | markdown | 8.1 KB |
| README.md | markdown | 1.5 KB |
| SKILL.md | skill md | 8.4 KB |
| resources/references/2026-ai-updates.md | markdown | 2.7 KB |
| resources/references/audit-checklist.md | markdown | 5.0 KB |
| resources/references/b2b-vs-b2c-guardrail.md | markdown | 3.1 KB |
| resources/references/benchmarks.md | markdown | 3.3 KB |
| resources/references/creative-formats.md | markdown | 2.7 KB |
| resources/references/crm-attribution.md | markdown | 4.8 KB |
| resources/references/tracking-hygiene.md | markdown | 2.7 KB |
.claude/skills/learning-phase/gt-SKILL.md
View on GitHub---
name: meta-ads-learning-phase
description: Meta Ads learning phase — the 50-events rule, budget floors to exit learning, what resets it, and learning-limited diagnosis. Use for learning phase, 50 events, exiting learning, learning limited, budget to exit learning, what resets learning, edits resetting learning. Triggers on "learning phase", "50 events", "exit learning", "learning limited", "budget floor", "resets learning", "learning limited status". Do NOT use for bid strategy detail (covered in campaign-setup) or measurement (use measurement).
---
# Meta Ads Learning Phase
Every ad set goes through a learning phase while Meta figures out who to show the ad to. Until it exits, performance is unstable and CPA is unreliable. Underfunded or constantly-edited ad sets get stuck "Learning Limited" and never stabilise — a quiet, common cause of weak B2B accounts.
## Instructions
1. Fund each ad set to clear ~50 optimisation events per week
2. Pick a conversion event frequent enough to hit that threshold
3. Avoid edits that reset learning mid-phase
4. Consolidate ad sets so budget is not spread too thin
5. Judge performance only after the ad set exits learning
## The 50-Events Rule
- An ad set needs roughly **50 optimisation events per week** to exit the learning phase. Below that it shows "Learning Limited" and delivery stays inefficient.
- Budget floor follows from your CPA: if a conversion costs ~$20 and you need 50/week, that is roughly **$140+/day** for that ad set. Higher CPA → higher floor.
- Rough B2B floors: ~$50/day minimum for retargeting ad sets; enough on prospecting to generate ~7+ conversions/day toward the 50/week target.
## Choose an Event You Can Actually Hit
- If your true conversion (qualified lead / demo) is too rare to hit 50/week, optimise to a higher-funnel event (lead, landing-page view of the right depth, or Instant Form open) to give the algorithm volume, then use offline data to keep quality honest (see measurement).
- This is the central B2B tension: optimise to an event frequent enough to exit learning, but feed back quality so you do not just buy cheap junk.
## What Resets Learning (avoid mid-test)
- Editing budget significantly, changing the audience, swapping the creative set, or changing the optimisation event resets learning.
- Do not make these edits mid-test. Batch changes, then let the ad set re-stabilise. Constant tinkering keeps an ad set permanently in learning.
## Consolidate to Escape "Learning Limited"
- Too many ad sets split the budget so thin no single one clears 50 events. Consolidate audiences (broad + Advantage+) and ad sets so each gets enough volume (see campaign-setup, audiences).
- Fewer, better-funded ad sets exit learning and stabilise; many tiny ones all sit Learning Limited.
## Examples
Example 1: "My ad sets say 'Learning Limited' — what do I do?"
→ They are not getting 50 events/week. Consolidate ad sets, raise budget toward the floor (CPA x 50 / 7 per day), or optimise to a more frequent event. Stop mid-test edits that reset learning.
Example 2: "How much budget do I need on Meta?"
→ Enough per ad set to clear ~50 optimisation events/week. At a $20 CPA that is ~$140/day per ad set; scale by your real CPA and consolidate so budget is not spread thin.
## 2026 updates
For Opportunity Score, Advantage+ 2026 changes (25/wk threshold, Predictive Budget Allocation), gen-AI Creative Enhancements, Advantage+ Leads, and Threads placements → Read `{SKILL_BASE}/resources/references/2026-ai-updates.md`.
---
*Created by [Growth Today](https://www.growthtoday.co) — AI-native GTM engineering firm. Maintained and updated by [Brigitta Ruha](https://www.linkedin.com/in/brigittaruha/). More open Claude Skills for go-to-market teams: https://www.growthtoday.co/claude-skills*
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 323k | Updated | Claude CodeCodex | Verified | Install |