Killing ad campaigns is where most paid budgets get wasted.
Either you cut too early and miss the campaigns that would have worked. Or you let losing campaigns drag on for weeks burning budget that should have moved to winners.
So in this guide, I'll walk you through the actual decision framework we use with our scaling clients. The four signals that mean kill, the three that mean keep going, and the timing windows for each.
The single biggest mistake: killing too early
Across the audits we've run on scaling brands, the most common paid mistake is pulling the plug at day 3 or day 4 of a new campaign.
Here's why that breaks performance. Meta's machine learning needs roughly 50 conversions per ad set per week to optimise properly. Below that threshold, performance is essentially random.
If you launch a campaign on Monday and check it Wednesday with 8 conversions, you have noise. Not signal.
The minimum reasonable assessment window for any new campaign:
- Meta Advantage+ Shopping: 14 days, minimum £100/day spend
- Manual Meta campaigns: 7 days, minimum £50/day per ad set
- Google Search: 7 days for branded, 14 days for non-branded
- Google PMax: 14 to 21 days (PMax has a longer learning curve)
- YouTube: 14 days (audience-driven optimisation)
Below these windows, you don't have enough data to make a kill decision.
The four signals that mean cut
1. CAC drift trending up week-over-week for 3+ weeks
One bad week isn't a signal. Three consecutive weeks of CAC climbing 15%+ is.
This pattern usually means audience fatigue, creative fatigue, or competitor pressure (someone else bidding into your audience). Killing the campaign and re-launching with refreshed creative and a new audience often resets performance.
2. Frequency over 4 with declining CTR
Frequency above 4 means each person in your audience has seen your ad 4+ times in the period. CTR halving while frequency climbs is a textbook fatigue signal.
The fix usually isn't to push harder. It's to refresh creative and reset audience targeting.
3. Creative-fatigued ads where new creative is performing
If you've shipped fresh creative that's outperforming the old by 30%+, the old creative isn't worth keeping alive. Reallocate budget to the new winners.
The mistake we see: brands keeping low-performing legacy creative "just in case" and splitting budget too thin across 8-10 underperforming ad sets.
4. Cohort LTV from this channel is meaningfully worse than blended
This is the slow-killer signal most paid managers miss.
If a channel reports £80 CAC with £200 first-purchase value, looks profitable. But if customers from THAT channel have 30% lower 12-month LTV than blended, the actual unit economics are worse than they look.
Pull the cohort report monthly. If channel-specific LTV underperforms blended by 20%+, the channel is acquiring lower-quality customers and warrants a cut even if surface metrics look fine.
The three signals that mean keep going (even when CFO is pushing to cut)
1. Conversion volume is hitting the learning threshold but CAC is still settling
The first 2-3 weeks of a new campaign on Meta produce noisy CAC numbers. If you're still under 50 conversions/week and CAC has dropped 20%+ since week 1, the algorithm is still learning and the trend is in your favour.
Cut at this stage and you've kneecapped a campaign that was 1-2 weeks from profitability.
2. Top-funnel awareness campaigns showing brand-search lift
Awareness campaigns rarely show direct ROAS in last-click attribution. The real return shows up in:
- Brand search volume increasing 30%+
- Direct traffic increasing
- Paid search efficiency improving (because awareness lifted brand recognition)
If your top-funnel campaign isn't producing direct conversions but IS driving brand-search lift, it's working. Don't cut it because of last-click attribution alone.
3. Audience expansion campaigns with high view-through but low click-through
Some campaigns work via view-through impact, not click-through. Especially YouTube and TikTok awareness campaigns.
If your campaign reports low CTR but the campaign window correlates with measurable lift in branded search, direct traffic, or organic conversions, the channel is working. Just measure it correctly.
The decision framework
Every Friday, we review every active campaign for our clients with this 4-question checklist:
- Has the campaign run for the minimum assessment window? If no, hold.
- Is CAC trending up, flat, or down over the past 3 weeks? Up trend = warning. Flat = monitor. Down = scale.
- Is frequency over 4 with declining CTR? If yes, refresh creative.
- Is cohort LTV from this channel underperforming blended? If yes, cut even with surface profitability.
This 4-question filter prevents 80% of premature kills and 80% of zombie campaigns.
The trap most teams fall into
The hardest discipline in paid: killing campaigns that show "acceptable" CAC but are actually pulling worse-quality customers than blended.
These campaigns LOOK profitable in the platform dashboard. They survive monthly reviews. They consume budget for months.
Meanwhile, the high-quality channels (often slower-CAC organic and brand-led acquisition) get under-funded because the data dashboard says paid is working.
The fix: cohort-level reporting. Channel performance isn't the channel's CAC. It's the LTV of the cohort it produces.
Once you measure that way, the kill decisions get obvious.
Putting it together
Most paid budget waste comes from two errors:
- Killing winning campaigns 2 weeks too early (before the algorithm finished learning)
- Letting zombie campaigns drag on (because they look acceptable in last-click attribution)
The fix is process, not instinct. Define your minimum assessment window before launching. Run the 4-question filter weekly. Use cohort LTV as the ultimate measure of channel quality.
Run this discipline and your paid budget compounds. Don't, and you'll keep paying for the same audience optimisation lessons every quarter.
