ROAS, Decision Rules & Kill/Scale Engine
The single most re-asked question in ecom: "Should I kill this or wait?" Use the interactive engine below — enter your live numbers and get a verdict with specific action steps, right now.
The Kill/Scale engine removes emotion from campaign decisions. Enter your live numbers and get a verdict. This is the module you'll return to most often — bookmark it separately if you can.
Bookmark this page. You'll return here every time you're unsure whether to kill or scale a campaign. The interactive engine at the top takes 60 seconds to run. The reference table below handles edge cases. The FAQ at the bottom covers the questions nobody else answers clearly.
Interactive Kill/Scale Decision Engine
Enter your live campaign numbers. Get an instant verdict and specific action plan.
The Kill/Scale Framework — Printable Reference
| Signal | What It Means | Action |
|---|---|---|
| CTR below 0.5% after $30 spend | Hook failure — creative, not product | Kill creatives. New hooks only. |
| ATC below 3%, spend below 4× CPA | Weak interest. Page problem. | Investigate page. Don't kill yet. |
| ATC below 3%, spend above 4× CPA, zero purchases | No purchase signal. Dead. | Kill product. Move on. |
| ATC 5%+, ROAS near break-even | Sales happening. Math needs fixing. | Fix offer: price, bundle, upsell. |
| ROAS 100%+ above KPI for 5+ days | Clear winner. Scale immediately. | Double budget. Add creatives. |
Critical FAQ — New Campaign vs New Ad Set
This is the most consequential structural decision in campaign management, and the wrong choice costs you weeks of learning data.
Add new ads to the existing winning campaign/ad set. Creating a new campaign resets the learning phase entirely and loses all the pixel data Meta collected on your original campaign. New ads within an existing campaign inherit the audience intelligence Meta has already built.
The exact rules:
- New creatives (same angle, same product): Add to the existing ad set. Never create a new campaign.
- New angle (different avatar, different pain point): Create a new ad set within the same CBO campaign. Same campaign, new ad set — preserves some pixel data while isolating the angle test.
- New product entirely: New campaign. A new product needs a clean data environment — don't contaminate your winning product's data with an unrelated SKU.
- Scaling a winner: Never create a new campaign to "scale." Duplicate the winning ad set within the same campaign and increase budget incrementally.
The core logic: every new campaign starts with zero audience data. Every new ad in an existing campaign starts with the benefit of all purchases Meta has already recorded. The more purchase data Meta has for a campaign, the cheaper your future conversions. Throwing that away to "start fresh" is almost always wrong.
The Most Common Kill Mistake
Day 1 and day 2 data is almost always bad. Meta is spending your budget on exploration — finding out who might be interested. CPMs are higher, delivery is less efficient, and ROAS looks terrible. Operators kill 80% of their campaigns before they have enough data to make any decision at all. The rule: do not make a kill decision until you've spent at least 2× your target CPA, and wait 4× CPA for a definitive verdict.
Bad Day vs. Dead Product — The Critical Distinction
On budgets below $100/day, just 2–3 people deciding not to buy can destroy your daily ROAS. One bad day is noise. Two consecutive bad days after a run of good days is still noise. The signal only becomes reliable when you've used 7-day rolling data, not daily swings.
- Dead product signals (across 4× CPA spend): High CPC consistently, CTR below 0.7% across all creatives, zero ATCs regardless of ad set, no purchase signal across multiple angles.
- Bad day signals (don't kill yet): One or two days of poor ROAS after several profitable days, ATC present but no purchase conversion, single ad set underperforming while others hold.
- The 7-day rule: If your 7-day average ROAS is above break-even and trending flat or up — keep running. One bad day inside a profitable 7-day window means nothing.
Testing Enough Angles Before Quitting the Product
The most common mistake operators make is killing a product after testing one or two angles. A product is not dead until you've tested 4–8 distinct angles with proper budget. Most winning products looked dead after the first angle.
| Angles Tested | Spend Threshold | Verdict |
|---|---|---|
| 1–2 angles | Any amount | No verdict. Test more angles before any decision. |
| 3–4 angles, zero ATC | 4× CPA each | Kill product. Problem is product-market fit, not creative. |
| 4–8 angles, ATC present, no purchase | 4× CPA each | Price or page problem. Fix offer, not creative. |
| 8+ angles, zero traction | 4× CPA each | Kill product. Exhausted all angles. Move to next product. |
| Any angle, ROAS 100%+ KPI, 5+ days | Any | Scale. You have a winner. Double budget and add creatives. |
Spent $300 and got nothing? That is not failure — that's research. The operators who quit at $300 never find out they were two angle iterations from a winner. Cut what the data says to cut. Test what the framework says to test. Keep moving.
The Mindset Behind the Numbers
Every operator hits this moment. You've followed the framework. You've tested angles. You've spent real money. You have nothing to show. Here's the honest reframe: $300 of learning data is not $300 lost — it's $300 invested in understanding what the market doesn't want. That's not failure. That's research. The operators who quit at $300 never find out they were two angle iterations away from a winner.
The kill/scale framework removes emotion from the decision. You're not giving up — you're executing the plan. Cut what the data says to cut. Test what the framework says to test. The operators who stay in this business long enough to win are the ones who stop treating every kill decision as a personal failure and start treating it as a professional signal.
Post-Andromeda reality: Small iterations on a losing angle won't save it. When a concept isn't working, you need an entirely new angle and concept — not tweaked copy on the same hook. The market has seen the same hooks. Give it something new.
Grey-zone decisions — where your data is mixed and neither kill nor scale is obvious — are exactly when operator experience matters most. Post the numbers in the Discord. Someone who's been in that exact grey zone will give you a real read instead of a guess.
Quick reference — the kill / tune / scale decision
Once a product has had a fair test, this is the decision in one view.
🧮 Use the Break-Even ROAS Calculator for this step →
Not sure if you're actually ready for this stage?
Get your free Readiness Score — 10 questions, 90 seconds, tells you exactly which gaps to close first.
