When to Kill a Facebook Ad: The Decision Framework

When to Kill a Facebook Ad: The Decision Framework

Knowing when to kill a Facebook ad is half the job of being a media buyer. Kill too early and you'll murder ads that would've worked. Kill too late and you'll bleed budget on dead weight. There's no perfect answer, but there is a framework that gets you 90% of the way there.

Most advertisers kill on emotion — they panic on day two when CPA looks high, or they let losers run for a fortnight because "it might still turn around." Both habits cost money. Let's replace them with a system.

The first rule: never kill before the learning phase ends

Meta tells you outright when an ad set is in the learning phase. There's a label in Ads Manager. While that label is on, you don't have reliable data — Meta is still figuring out who to show your ad to. Pulling the plug during learning is statistical malpractice.

The learning phase typically ends when an ad set hits 50 conversions in a 7-day window. If your CPA target is £20 and your daily budget is £30, that's 75 conversions in a week — easily enough to exit learning. If your CPA target is £20 but your daily budget is only £10, you'll never hit 50 conversions and you'll be stuck in learning forever. Either raise the budget or change your optimisation event to something cheaper (like Add to Cart instead of Purchase).

The 3x rule for individual ads

For individual ads (not ad sets), here's a simple rule: don't kill until you've spent at least 3x your target CPA.

If your target CPA is £15, give each ad at least £45 of spend before judging it. Below that, you're making a verdict on essentially zero data.

If your target CPA is £40, give each ad £120. That feels expensive, and it is — but the alternative is killing winners by accident.

The kill criteria (in order)

Once an ad has hit 3x CPA spend and exited learning, work through these checks in order:

1. CTR < 0.5%
If nobody's clicking, your creative is dead. Kill it. Don't try to "optimise the audience" — the audience isn't the problem.

2. CPA > 1.5x target with at least 5 conversions
If you're £20 over your CPA target with enough data to be confident, kill. The pattern is set. Hoping it improves wastes more money.

3. CPM 50%+ above account average AND no compensating CTR
High CPMs are sometimes fine if CTR makes up for it. But if you're paying premium for impressions and getting average click-through, the auction has decided your ad is unattractive. Kill.

4. Frequency > 4 with declining CTR
This is a fatigue signal. The ad worked, but it's worn out. Pause and refresh, don't try to revive.

5. ROAS < 60% of target after at least 7 days post-learning
For ecommerce, this is the financial backstop. If the ad set isn't producing close to target ROAS after a full settled week, it's not going to get there.

Work through these checks in this order. If the first two are fine, the rest usually are too.

When to give an ad more time

There are three signals that suggest an ad deserves more rope:

1. CTR is strong (>1%) but CPA is high. Could be a landing page or offer issue, not a creative issue. Test the funnel before killing the ad.

1. Performance is improving day over day. Sometimes ads need 5-7 days to find their audience. If yesterday was better than the day before, give it another 48 hours.

1. The ad had one bad day after a good week. Don't kill on noise. Look at 7-day rolling averages, not yesterday's spreadsheet.

When to kill an entire ad set, not just an ad

Killing an ad is different from killing an ad set. You kill an ad set when:

• All ads in it are underperforming (the audience or placement is the issue)
• The ad set has been stuck in learning for 14+ days (Meta can't find buyers)
• Frequency is high across all ads and the audience is exhausted
• You're cannibalising another ad set (audience overlap)

Killing individual ads while keeping the ad set live is your default action. Killing entire ad sets is the heavier hammer.

The 24-hour cooling-off rule

When you have the urge to kill an ad mid-day because performance is bad, wait 24 hours. The single biggest source of bad decisions in Facebook ads is intra-day reactions to noisy data.

Meta's reporting can lag for hours. Conversions trickle in. The ad you're about to kill at 3pm might look totally different at 9am tomorrow. Set a rule: no kill decisions before 24 hours of data.

What to do after you kill

Killing an ad isn't the end of the analysis. Every dead ad teaches you something. Before you move on:

1. Note why it died (low CTR, high CPM, frequency burnout, etc.)
2. Cross-reference against other ads in the same testing batch — was it a creative issue or an audience issue?
3. Update your testing notes so you don't repeat the same hypothesis next month

A kill log is one of the most valuable things a media buyer can keep. Over time it becomes a cheat sheet of what doesn't work for your specific account.

Common kill mistakes

• Killing ads in learning because you panicked. Always wait until learning ends.
• Killing ads on weekends because performance always dips on Saturdays in B2B and Sundays in some B2C niches.
• Killing ads after a single bad day when the 7-day average is still positive.
• Killing ads based on CPC when CPC isn't even your KPI. The metric you care about is CPA or ROAS, not CPC.
• Killing ads without writing down why. You'll repeat the same losing test in 8 weeks.

Letting AI make the kill calls

Phonecasting kill decisions is an emotional drain. Pix-Vu automates the kill criteria above using account-specific thresholds. It waits for statistical significance, ignores intra-day noise, and only acts when the data justifies it. £99/month flat — and it doesn't get tired or twitchy at 11pm.