Core Philosophy
Before you touch a campaign — understand the system you're operating in.
What we test
Creatives.
Not Audiences.
Not Audiences.
We never test audiences. We always run broad and let Meta's algorithm do the audience work. What we test are creatives — the visual, the hook, the angle, the offer. This is the only variable we control.
Targeting rule
Broad.
Always broad.
Always broad.
No interests. No lookalikes. No detailed targeting. You may only narrow by age or gender when it genuinely fits the product. Everything else gets left to Meta's algorithm — it's better at this than you are.
What you're buying
Data.
Not sales.
Not sales.
Most creatives fail. Most test campaigns lose money. That's not failure — that's the process. Every dollar spent teaches the algorithm what your buyer looks like. You are buying data, not results.
Optimization event
Purchase.
Always purchase.
Always purchase.
Never run engagement campaigns. Never optimize for ATC. Always set your optimization event to Purchase (Conversions). If your pixel isn't warmed up, that's a separate fix — not a reason to optimize for the wrong event.
Strong Creative
Hook · Visual · Offer
→
Algorithm Signal
Meta learns your buyer
→
Right Audience
Automatically found
The Diagnostic Funnel — Why Ads Fail
🚫 No clicks on ad
→
Creative problem — hook isn't working
👆 Clicks but no ATC
→
Offer or product problem — page doesn't convert
🛒 ATC but no purchase
→
Website problem — checkout, trust, friction
Campaign Structure
Two structures, two purposes. Know when to use each.
Ad Set Budget Optimization
ABO
Budget controlled at ad set level
Campaign1 campaign
Ad Set1 ad set
Ad1 ad
Daily budget (ad set level)
$50
Fully isolated test — one creative, clean data
No internal competition between ads
Precise spend tracking per creative
Easier to kill individual underperformers
Campaign Budget Optimization
CBO
Budget controlled at campaign level
Campaign1 campaign
Ad Set1 ad set
Ad 1
Ad 2
Ad 3
Ad 4
Ad 5
+5 more
5–10 creatives per ad set
Daily budget (campaign level)
$50
Meta auto-allocates spend to winning creatives
Faster to find a winner across many creatives
Efficient use of budget during testing
Best for scaling discovery at moderate budgets
⚡ Structure Insight
There is no universal winner between ABO and CBO. Use both and let real performance data tell you which structure works better for your specific product and audience. Start with both, compare results after 3–5 days, then double down on the structure that delivers.
Day 0–7 Execution Plan
A strict, step-by-step sequence. Follow it exactly — no improvising.
0
Day 0 — Launch
Set Up & Launch. Then Stop.
Launch ABO campaign (1 campaign → 1 ad set → 1 ad → $50/day)
Launch CBO campaign (1 campaign → 1 ad set → 5–10 ads → $50/day)
Do NOT touch anything after launch. Let Meta enter its learning phase.
1
Day 1 — 24 Hours
First Glance. Only Kill if Catastrophic.
Check CPC — is it extremely high? Check CTR — is it near zero?
Kill ONLY if CPC or CTR is catastrophically bad. Otherwise leave it running.
New accounts: wait the full 48h — Meta needs more time to exit learning phase.
2
Day 2 — 48 Hours
Main Decision Point.
$30–$40 spent + 0 ATC → Kill immediately. No buyer intent. Move on.
ATC present → Continue running. Check CPA vs break-even.
Some ATC but no purchases → Wait. Check CPA trajectory over next 24h.
3–4
Days 3–4
Identify Potential Winners.
Identify which creatives are consistently driving ATC and purchases.
Monitor CPA consistency. A spike is not a winner — consistent performance is.
Kill any creative that hits 1.5x break-even CPA with no purchase.
5+
Day 5+ — Scaling Phase
Scale Only If Criteria Are Met.
Profitable or at break-even with consistent results → Begin vertical scaling.
1 sale, inconsistent results, or spike performance → Do NOT scale yet.
Hard Rules — Non-Negotiable
Don't Edit Campaigns
Editing resets the learning phase. Wait until the test is done.
Don't Use Gut Feeling
Every decision must be backed by data. Emotions kill campaigns.
Ignore Meta Suggestions
Turn off all Meta auto-enhancements and recommendations.
No Detailed Targeting
Always broad. Interests limit the algorithm's reach.
No Weekend Scaling
Normal operation: don't scale on weekends. Q4 exception only.
Optimize for Purchase
Never ATC, never engagement. Purchase only, always.
Performance Metrics
Three numbers run the system. Know what each means, and when to act.
CPC
Cost Per Click
How much you pay every time someone clicks your ad. High CPC signals a weak creative — your hook isn't stopping the scroll or resonating with your audience. Always read CPC together with CTR.
⚠️ Warning signal$1.50+ CPC — depends on AOV & margins
CTR
Click-Through Rate
The percentage of people who see your ad and click it. A low CTR means people are scrolling past your ad. Evaluate CTR together with CPC — high CTR + high CPC can still mean a problem.
✅ Healthy range1%+ is acceptable. Higher = stronger creative.
ATC
Add To Cart — Primary Signal
The most important early metric. Someone clicked AND showed purchase intent by adding to cart. This is the first real signal of buyer behavior. No ATCs = no interest in the product or offer.
🚫 Kill signal$30–40 spend + 0 ATC → turn off immediately
Break-Even CPA
Cost Per Acquisition · Your Kill Ceiling
Your break-even CPA equals your net profit per sale. This is the maximum you can spend to acquire a customer before you start losing money. Every ad has a kill point based on this number.
💀 Kill ruleSpend reaches 1.5× break-even CPA → Kill
✅ Scale ruleCPA ≤ break-even or profitable → Scale
⏳ Wait ruleCPA approaching but not at limit → Monitor
Kill / Keep / Scale Table
| Situation | Action | Reasoning |
|---|---|---|
| High CPC + no ATC after $30–40 spend | Kill | No creative pull, no buyer intent |
| No ATC after $30–40 spend | Kill | Zero purchase intent signal |
| Spend reaches 1.5× break-even CPA | Kill | Math doesn't work, cut losses |
| Some ATC but no sales yet | Wait | Intent is there, give it time |
| Inconsistent performance / spikes | Wait | Not enough data to conclude |
| Stable CPA at break-even | Continue | You're learning, not losing |
| CPA below break-even (profitable) | Scale | Winner found — increase budget |
Decision Tree
Follow this flowchart exactly. Every decision point has one correct answer based on data.
🚀 Launch Campaigns (ABO + CBO)
⏳ Wait 24–48 hours
(48h for new accounts)
(48h for new accounts)
💸 Did you spend $30–40?
✗ NO ATC
🚫 Kill Campaign
No buyer intent. Test new creative or angle.
No buyer intent. Test new creative or angle.
✓ ATC EXISTS
📊 Check CPA vs Break-Even
CPA > 1.5× break-even
🚫 Kill
Math doesn't work
Math doesn't work
Inconsistent / 1 sale
⏳ Wait
Monitor 24–48h more
Monitor 24–48h more
Profitable / break-even
✅ Scale
Increase budget 20%
Increase budget 20%
🏆 Repeat. Test new creatives. Scale winners.
System Flow — The Full Loop
Step 1
Launch
→
Step 2
Wait 48h
→
Step 3
Check ATC
→
No ATC
Kill
or
Has ATC
Check CPA
→
Profitable
Scale
Scaling System
Only scale when the data says so. Never chase a spike.
Vertical Scaling
Increase Budget Gradually
Raise budget by ~20% every 2 days. Never jump dramatically — it disrupts the learning phase. After reaching $200–250/day, switch to $50 increments.
$50→Starting budgetDay 0
$60→+20% increaseWait 2 days
$75→+20% increaseWait 2 days
$90→+20% increaseWait 2 days
$250+→Switch to $50 stepsEvery 2 days
Horizontal Scaling
Duplicate Winners
Duplicate winning campaigns into new campaigns at the same or increased budget. This spreads risk and allows the algorithm to find new audiences without disrupting your winning campaigns.
Creative Scaling
Duplicate & Refresh Creatives
When a winning creative starts to fatigue, duplicate it and test new hooks, angles, or messaging variants. The offer stays the same — only the hook changes.
✅ When to Scale
✓ Campaign is profitable
✓ CPA at or below break-even
✓ Results are consistent (not a spike)
✓ At least 3–5 sales, not just 1
Note on break-even scaling: Scaling at break-even is intentionally allowed. You're not losing money — and you're buying more data to optimize with. The algorithm gets more signal, which typically improves performance over time.
🚫 Do NOT Scale If...
Only 1 sale in the campaign
Results are inconsistent
Performance is spike-based
CPA is above break-even
It's the weekend (non-Q4)
Learning phase not complete
Common Beginner Mistakes
01
Killing too early
Turning off ads before 48h or before $30–40 is spent
02
Editing live campaigns
Resets the learning phase and destroys accumulated data
03
Testing variables
Changing too many things at once — you can't read the data
04
Detailed targeting
Interests and lookalikes restrict Meta's learning capacity
05
Trusting Meta tips
Meta recommendations benefit Meta's revenue, not yours
06
Wrong optimization
Optimizing for engagement or ATC instead of Purchase
07
Gut-feeling decisions
Every action must be justified by a data point, not a feeling
08
Scaling a spike
One good day is not consistency — wait for a pattern first