Should You Kill This Ad? A Simple Decision Tree for Contractors.

Sometimes, but don't count on it. A paused ad's results after a rest are unpredictable: occasionally better once the audience has forgotten it, often worse. Treat a resumed ad as a new test, and for a fatigued ad, a fresh version (different thumbnail, different opening seconds) is usually the better bet.

Your agency should be doing this automatically. Weekly creative reviews + 7-day pause thresholds should be standard. If your agency is running creatives 60+ days without pausing any, they're not optimizing — they're just running campaigns.

Stop scaling. Take a step back. The issue isn't individual ads — it's the account-level strategy (offer, audience, landing page, funnel). Use our Account Health Score tool to diagnose where the foundation is weak.

Watch your creative burn rate. If you're pausing 5+ ads per month and launching 5+ new ones, you're likely triggering learning-phase resets constantly. A healthy account runs the same 3-5 creatives for 30-45 days, with 1-2 getting refreshed monthly. If you're rotating weekly, the ads aren't the problem — your audience or offer is. Kill the right things, not everything.

Mostly yes, partially no. Meta's 'low delivery' warning is usually accurate — your ad isn't winning auctions, often due to relevance or budget vs bid mismatch. But Meta's 'we recommend you pause this ad' suggestion sometimes triggers on ads that are spending efficiently against a different goal than Meta optimizes for (e.g. high-quality leads at higher CPL is good, but Meta sees high CPL and recommends pause). Override the recommendation if your CPBJ math says the ad is profitable.

At least 7 days and about 5,000 impressions (Step 1 above). Meta's learning phase aims for about 50 optimisation events a week, which small local budgets often don't reach, so don't wait for 50 leads per ad either. Before that, you're reading noise: a new ad's first 3 days frequently overstates or understates real performance because Meta is still figuring out who to show it to. The 'kill in 48 hours' approach kills genuinely good ads while they're still learning. Apply this rule even if early data looks bad: for example, an ad costing $60 a lead on day 2 can settle much lower by day 8 once delivery stabilises. Killing everything at 48 hours means you never find out which ads would have become winners.

For marginal ads (within 20% of target CPL but not winning), use 'pause-and-wait' for 30-45 days BEFORE killing. Why: a rest sometimes lets a tired ad recover, though results after a pause are unpredictable, so treat any relaunch as a new test. Hard kill for: ads at double target CPL or more, or with frequency above 3.5 and falling CTR (Steps 2 and 4 above), ads with policy issues, ads from a campaign objective that no longer matches your strategy. The discipline: maintain a 'sleeping ad' folder of paused-but-not-deleted creatives. Resurrect them quarterly to test against fresh creative — now and then an old paused winner beats your new launches.

Often bigger than it looks. Hypothetical math: a $3K/mo campaign with 4 active ads, where one is performing at $40 CPL while the other three average $20 CPL. If you keep the bad ad running for 30 days unnecessarily, you've spent $750 on the bad ad ($25/day × 30) generating ~19 leads at $40 each — when that same $750 redirected to the winning ads would have generated ~37 leads at $20 each. Net cost: about 18 lost leads, multiplied by your close rate and average job value. Repeat that every month and the opportunity cost adds up quickly. The lesson: once an ad has had a fair test, pausing a little too quickly usually costs far less than running a loser for a month.

Use automated rules for the obvious kills, manual review for the edge cases. Set up these 3 automated rules in Ads Manager → Rules: (1) PAUSE if CPL >2x target for 7+ days AND the ad has enough results to judge (set a minimum, such as 10-20 leads, so noisy data doesn't trigger it); (2) PAUSE if frequency is above 3.5 AND CTR has dropped well below the ad's own baseline (catches fatigue automatically); (3) ALERT (don't auto-pause) if CPL trends 30%+ over baseline for 3 consecutive days — emails you for review. Automated rules cut your weekly time in Ads Manager and prevent the 'I forgot to pause that bad ad for 2 weeks' mistake. Manual review reserved for: borderline ads (within 20% of target), creative tests still in Learning Phase, ads that lost frequency war but might recover. Automated rules are free and quick to set up, and they cut the spend lost to ads left running too long.

Diagnose WHERE the funnel is breaking, then decide. High CTR + low conversion = the AD is working but something downstream is broken. Three checks: (1) AD-TO-LP MISMATCH — does the ad promise something the landing page doesn't deliver? Click 'View Ad' in Ads Manager, then click through to your LP yourself; read both copies side-by-side. If the promises differ, the LP is your bottleneck (don't kill the ad — fix the LP); (2) WRONG CTA — ad drives clicks but the LP CTA is unclear or buried. Check LP CTA prominence; consider rewriting; (3) LANDING PAGE LOAD speed — high CTR with low conversion sometimes means visitors are clicking but the LP loads too slowly + they bounce. Test mobile load on a 4G connection. Don't kill an ad with high CTR — those are precious. The ad earned attention; you just need to convert it. Fix the downstream layer first; only kill the ad if you've verified the LP + funnel are clean and the ad still doesn't convert.

Reframe the cost as already-spent. Three mental tools: (1) SUNK COST is invisible to the algorithm — Meta doesn't care that your video cost $3,000 to produce; it only cares about current performance. Treating an ad as 'unkillable because of investment' is how contractors lose another $3K running a $3K creative for a year past its peak; (2) ARCHIVE, don't delete — paused ads don't disappear. The footage + ad copy + targeting are still in your library. You can revive any time. The 'kill' isn't permanent loss; it's tactical pause; (3) THINK in PORTFOLIO — like a stock portfolio, you keep what's appreciating + cut what's dropping. Even if the cut feels painful, the alternative (running a losing ad) costs MORE money over time than the original production. It's easy to keep a mediocre ad alive well past the point the numbers said stop, because of attachment to the production effort. The emotional cost of pausing is felt; the financial cost of NOT pausing is invisible. Train yourself to act on the math, not the emotion.

Default to AD level for under-performers; AD-SET level only when broader signals are off. AD-LEVEL pause: kills one creative within an ad set. The ad set keeps running with remaining creatives; the algorithm keeps its accumulated learning + audience signal. Use this when one specific creative is fatigued but others in the set are performing. AD-SET LEVEL pause: kills the entire targeting+creative+budget combination. The algorithm loses ALL its accumulated learning for that audience; restarting later means a fresh learning phase, which small budgets can take weeks to exit. Use this only when: (a) the audience itself is broken (wrong targeting); (b) the entire creative set is fatigued + you're rebuilding with all-new creative; (c) the ad set objective is fundamentally wrong. Killing at the ad set level throws away delivery history that took weeks to build; kill ads first. Better default: pause one ad, leave the ad set running, swap in fresh creative. Surgical kills preserve the value of what's working; broad kills throw the baby out with the bathwater.

Keep a 'Killed Ads Log' spreadsheet — 5 columns, simple structure: (1) AD NAME + launch date; (2) DATE KILLED; (3) PERFORMANCE at kill (CPL, frequency, CTR); (4) REASON (creative fatigue / wrong audience / policy issue / replaced by winner / etc.); (5) LESSON LEARNED in 1-2 sentences. Review the log quarterly: pattern-match across kill reasons. Common findings: 'I kill ads too early when frequency is still under 3' (over-killing pattern); 'Most of my UGC fails after 6 weeks; my owner-on-camera lasts 10-12 weeks' (creative-format learning); 'Audience X often underperforms after 30 days' (audience saturation pattern). Without the log, every kill feels like a one-time decision; with it, patterns emerge that guide better future decisions. Document kill decisions so you don't repeat the same mistakes. The log compounds learning over time — by year 2 of running ads, you have a written playbook of what works + what doesn't in YOUR specific market.

Don't kill all ads at once. Three-step diagnostic + recovery sequence: (1) PAUSE the worst 50% of ads first (ones at double target CPL or more, or with frequency above 3.5); leaves the better 50% running at reduced budget while you investigate; (2) AUDIT the foundation — Pixel + CAPI firing? Audience overlap? Landing page broken? Use the Account Health Score tool. If you find foundation issues, fix them before any creative changes; (3) LAUNCH 2-3 fresh creatives in a SATELLITE ad set with $20-30/day; let them run 14 days; if they outperform the surviving ads, gradually shift budget. If even the satellite tests fail, the foundation issue is bigger (offer-market mismatch, broken targeting). Worst response: full kill + restart. Throws away algorithm learning and restarts the learning phase, which small budgets can take weeks to exit. Better: surgical pruning + diagnostic before rebuilding.