Kill a Facebook ad when CPA exceeds 2x your target after the learning phase, CTR drops below 1%, or frequency climbs past 2.5 on cold audiences with declining engagement. Let it run when the ad is still in the learning phase with positive early signals, when ROAS is at or above break-even, or when CTR remains strong and CPA is trending downward. The difference between a profitable ad account and a money pit is knowing which scenario you are looking at — and acting on the data instead of your gut.
This guide gives you a complete decision framework for both sides of the coin: the specific conditions that justify killing an ad immediately, and the equally specific conditions where patience pays off. Every threshold is backed by how Meta's algorithm actually works, not guesswork.
Why Most Advertisers Get This Decision Wrong
The kill-vs-let-it-run decision is the single most consequential judgment call in Facebook advertising. Get it right and you preserve budget for winners. Get it wrong in either direction and you lose money — either by cutting a future winner too early, or by feeding a loser for weeks.
The two failure modes are equally expensive. Killing too early means you burn through creative concepts without ever letting the algorithm optimize. Not killing fast enough means you subsidize losing ads with profit from winners, dragging down your blended ROAS. Both stem from the same root cause: making decisions on emotion instead of metrics.
What makes this harder is that Meta's algorithm needs time to learn. According to Meta, an ad set typically requires around 50 optimization events within a 7-day window to exit the learning phase. During that window, performance data is unreliable. CPA can swing wildly, ROAS looks terrible, and the numbers you see on day two are almost certainly not the numbers you will see on day seven. The advertisers who understand this — and build a framework around it — are the ones who scale profitably.
When to Kill a Facebook Ad: The 6 Kill Signals
Kill decisions should be binary. If any of these conditions are true, turn the ad off. No "let's give it one more day." No "but I spent so much already." The data is the data.
1. Zero Conversions After Spending 2-3x Your Target CPA
This is the only kill signal that overrides the learning phase. If your target CPA is $30 and you have spent $60-$90 without a single conversion, the ad is not going to turn around. The audience, creative, or offer is fundamentally misaligned. Kill it immediately and redeploy the budget.
2. CPA Exceeds 2x Your Target for 5+ Days Post-Learning
Once the learning phase is over and the algorithm has stabilized, a CPA persistently above 2x your target is a clear signal. According to a widely cited framework from Madgicx, if your maximum cost per purchase is $50, set a rule to pause the ad set once it spends around 1.5x that amount without a conversion. Five consecutive days at double your target is not a blip — it is a pattern the algorithm cannot fix. For more on reading these metrics, see our guide on reading the Facebook Ads dashboard.
3. Link CTR Below 1% After 1,000+ Impressions
Low CTR means the creative is not resonating. The average CTR across Meta ads is approximately 0.90% according to industry benchmark data, and ecommerce categories frequently outperform this baseline. If your ad cannot hit at least 1% link CTR on cold traffic after sufficient impressions, no amount of budget or audience tweaking will save it. The hook is weak, the visual is not stopping the scroll, or the offer is not compelling. Kill the ad and test a new creative.
Use link CTR, not CTR (all). Facebook's default "CTR (all)" metric includes reactions, comments, shares, and video views. That number is inflated. Link CTR only counts clicks to your destination — the clicks that can convert.
| Link CTR Range | Signal | Action |
|---|---|---|
| Below 0.5% | Dead on arrival | Kill immediately |
| 0.5% – 1.0% | Underperforming | Kill after 2,000 impressions with no improvement |
| 1.0% – 2.0% | Acceptable | Monitor CPA and ROAS |
| Above 2.0% | Strong | Let it run — focus on conversion metrics |
4. Frequency Above 2.5 on Cold Audiences with Declining Performance
Ad fatigue is one of the most predictable failure modes in Facebook advertising. According to multiple industry analyses, a frequency above 2-2.5 on cold audiences is the tipping point where engagement starts declining. When someone has seen the same ad three or more times, they scroll past it — or worse, hide it. That negative feedback signal tells Meta's algorithm to deprioritize your ad, which raises CPMs and CPAs further.
High frequency plus declining CTR plus rising CPA is textbook fatigue. The fix is not more budget. More budget on a fatigued ad just accelerates the decline. The fix is new creative. Kill the fatigued ad, keep the winning elements (audience, offer), and launch a fresh variation.
5. ROAS Below Break-Even for 7+ Days Post-Learning
ROAS is the most definitive kill metric because it directly measures profitability. If your ad is below your break-even ROAS for a full week after exiting the learning phase, it is losing money on every dollar spent. Your break-even ROAS depends on your gross margin — a store with 50% margins breaks even at 2:1 ROAS, while a store with 25% margins needs 4:1 just to cover costs.
| Gross Margin | Break-Even ROAS | Kill If Below For 7+ Days |
|---|---|---|
| 20% | 5:1 | Below 5:1 |
| 30% | 3.3:1 | Below 3.3:1 |
| 50% | 2:1 | Below 2:1 |
| 65% | 1.5:1 | Below 1.5:1 |
6. Two or More Relevance Diagnostics Ranked "Below Average"
Meta replaced the single relevance score with three diagnostics: Quality Ranking, Engagement Rate Ranking, and Conversion Rate Ranking. Two or more "Below Average" rankings mean your ad is being outperformed by competitors targeting the same audience on multiple dimensions. The algorithm will charge you more to compensate — higher CPMs, higher CPAs, lower delivery. Kill it and rethink the creative angle or audience.
Know your break-even ROAS before you kill or keep a single ad.
Every kill decision in this guide depends on your break-even threshold. Plug in your margins and get the exact number in under 60 seconds.
Calculate Your ROAS →When to Let a Facebook Ad Run: The 5 "Hold" Signals
Patience is just as important as decisiveness. These are the scenarios where cutting an ad would cost you a potential winner.
1. The Ad Is Still in the Learning Phase with Positive Early Signals
If the Delivery column says "Learning" and your link CTR is above 1%, let it run. Meta's algorithm needs approximately 50 optimization events to stabilize, though Meta has recently lowered this requirement to as few as 10 events for some Purchase-optimized and App Install campaigns. During this window, CPA and ROAS numbers are unreliable. A $75 CPA on day two based on 3 conversions is noise, not signal. That same ad might settle at $28 by day seven.
The key qualifier is "positive early signals." If CTR is above 1% and you are seeing add-to-carts, the conversion funnel is working — the algorithm just needs time to find more of those people. If CTR is below 0.5% with zero engagement on day three, the creative is the problem and waiting will not fix it.
2. CPA Is Above Target But Trending Downward
A CPA of $45 when your target is $30 is concerning — unless it was $60 two days ago and $52 yesterday. A clear downward trend in CPA means the algorithm is learning and optimizing. Give it another 3-5 days to see if it reaches your target. The trajectory matters more than any single day's number.
3. ROAS Is at or Above Break-Even
An ad that is hitting break-even ROAS is not a loser — it is an ad that is covering its own cost and potentially driving brand awareness and remarketing audiences at no net cost. Not every ad needs to be a home run. Ads that break even on the front end often generate long-term profit through repeat purchases and increased customer lifetime value.
4. The Ad Has High Engagement and Social Proof
An ad with hundreds of likes, comments, and shares has accumulated social proof that makes it more effective over time. Even if ROAS has dipped slightly, that social proof reduces the cost of future impressions and increases trust with new audiences. Before killing a high-engagement ad, duplicate it into a new ad set to preserve the social proof while testing a different audience.
5. Seasonal or External Factors Are Temporarily Affecting Results
CPMs and CPAs spike during major events — Black Friday, Christmas, political election cycles, and other periods of intense advertiser competition. If your ad was performing well and suddenly got more expensive during a known high-competition period, the ad itself might not be the problem. Check whether CPM has risen across your entire account, not just the specific ad. If it has, hold steady and wait for costs to normalize.
The Complete Kill vs Let Run Decision Framework
Run through this table every time you evaluate an ad. Start at step 1. The first "Yes" gives you your answer.
| Step | Question | If Yes | If No |
|---|---|---|---|
| 1 | Spent 2-3x target CPA with zero conversions? | Kill now | Go to step 2 |
| 2 | Still in learning phase with CTR above 1%? | Let it run — re-check in 3 days | Go to step 3 |
| 3 | Link CTR below 1% after 1,000+ impressions? | Kill — creative is the problem | Go to step 4 |
| 4 | CPA above 2x target for 5+ days post-learning? | Kill — audience or offer mismatch | Go to step 5 |
| 5 | Frequency above 2.5 with CTR declining? | Kill — ad fatigue, launch new creative | Go to step 6 |
| 6 | ROAS below break-even for 7+ days? | Kill — not profitable | Go to step 7 |
| 7 | CPA trending downward toward target? | Let it run — re-check in 3 days | Go to step 8 |
| 8 | 2+ relevance diagnostics "Below Average"? | Kill — outcompeted on quality | Let it run |
If an ad passes all eight checks, it is performing well enough to keep running. Re-evaluate every 3-4 days using the same framework. This eliminates emotional decision-making and keeps your ad account lean.
What to Do After You Kill an Ad
Killing an ad is not the end — it is the start of your next test. The data from a dead ad is valuable if you use it correctly:
- Diagnose the failure mode. Was it the creative (low CTR), the audience (high CPA despite strong CTR), or the offer (low conversion rate on the landing page)? Each failure mode has a different fix. Get specific.
- Reallocate budget to winners. Move the freed-up spend to your best-performing ad sets. If you are already scaling without killing ROAS, increase winner budgets by 15-20%.
- Test one variable at a time. If the creative failed, test a new creative with the same audience. If the audience was the problem, test the same creative with a different audience. Changing both at once means you will never know which fix worked. See our full guide on how to test Facebook ads for the complete framework.
- Preserve winning elements. A dead ad often has one strong component. Maybe the hook worked but the offer was weak. Maybe the audience was right but the format was wrong. Extract what worked and build your next test around it.
- Set a testing timeline. Launch the new variation and give it the same evaluation window — the full learning phase or 7 days. Run it through the same kill decision framework. No shortcuts.
Setting Up Automated Kill Rules in Ads Manager
You do not need to babysit every ad 24/7. Meta's automated rules can act as a safety net between your manual reviews. Here are the rules worth setting up:
| Rule Name | Condition | Action | Time Window |
|---|---|---|---|
| Zero-Conversion Kill | Spend > 2x target CPA AND purchases = 0 | Pause ad set | Last 3 days |
| CPA Ceiling | Cost per purchase > 2x target CPA | Pause ad set | Last 5 days |
| CTR Floor | Link CTR < 0.8% AND impressions > 1,500 | Pause ad | Last 3 days |
| Winner Scale | ROAS > 3x break-even AND spend > $50 | Increase budget 15% | Last 3 days |
Automated rules handle the extremes — the clearly dead and the clearly winning. The ambiguous middle ground still requires your judgment using the decision framework above. Check manually every 2-3 days during the first two weeks of a campaign, then shift to weekly reviews once performance stabilizes.
The Real Cost of Getting It Wrong
Every day you let a losing ad run is a day you are choosing to burn money. Every winner you kill on day two is a winner you will never find. The compound cost of bad kill decisions shows up in your True Margin — the real profit left after ad spend, COGS, shipping, and everything else your revenue dashboard does not show you.
The advertisers who grow profitably are not the ones who never launch a bad ad. They are the ones who recognize bad ads fast, cut them decisively, give promising ads the time they need, and redeploy budget intelligently. That discipline — applied consistently with a framework, not feelings — is what separates successful Facebook advertisers from the ones who keep wondering why their ad account is "not working."
Stop guessing. Calculate your real break-even ROAS.
True Margin's free ROAS calculator shows your exact break-even threshold so every kill-or-keep decision is backed by real math.
Calculate Your ROAS →Frequently Asked Questions
How do I know if my Facebook ad is in the learning phase?
Check the Delivery column in Ads Manager. If it says "Learning," the ad set has not yet collected enough optimization events to stabilize performance. Meta typically requires around 50 conversion events within a 7-day window to exit the learning phase, though Meta has lowered this requirement to as few as 10 events for some Purchase-optimized and App Install campaigns. Do not make kill decisions while the status reads "Learning" unless the ad set has spent 2-3x your target CPA with zero conversions.
Should I kill a Facebook ad that used to perform well but is now declining?
Yes, if the decline has lasted 5-7 days and the ad shows rising frequency alongside falling CTR. This pattern is ad fatigue — the audience has seen the creative too many times. A frequency above 2.5 on cold audiences with a corresponding CTR drop is the clearest fatigue signal. Kill the fatigued ad, document what worked, and launch a fresh variation. Industry data suggests refreshing creatives every 4-6 weeks for broad audiences, and as often as weekly for smaller, niche audiences.
What is the minimum budget I should spend before killing a Facebook ad?
Spend at least 2-3x your target CPA per ad set before making a kill decision. If your target CPA is $30, spend at least $60-$90 before judging. This gives Meta enough data to optimize delivery. If you spend that amount with zero conversions, kill immediately. If you have some conversions but CPA is high, let it run through the full learning phase before deciding. Use our ROAS calculator to determine what CPA you can actually afford.
Can I pause a Facebook ad instead of killing it permanently?
You can, but pausing and restarting an ad set resets the learning phase. The algorithm has to re-learn who to show the ad to, and performance will be volatile again during that period. If you pause for more than a few days, treat it as a new ad set when you restart. For most advertisers, killing and launching a new variation is more effective than pausing and restarting the same creative.
How often should I check my Facebook ads for kill signals?
Check every 2-3 days during the first two weeks, then shift to weekly reviews once ads are stable. Daily checks lead to premature kill decisions based on normal fluctuations. Set up automated rules in Ads Manager to pause ad sets that exceed your CPA threshold, so you have a safety net between manual reviews. See our Ads Manager dashboard guide for how to set up the right columns and views.