When to Kill an Ad Creative (Exact Thresholds)

The most expensive mistake in paid social is not launching a bad creative. It is refusing to kill it. Marketers fall in love with an ad they scripted, give it "a little more time," and quietly bleed budget while a distracted algorithm keeps serving a loser. Knowing exactly when to cut a creative is a skill, and it is mostly about pre-committing to thresholds before emotions get involved.

This is the framework for making that call with numbers instead of feelings. We will walk through the four metrics that actually diagnose a creative, the concrete kill thresholds worth using as defaults, how long to wait before you trust the data, and why changing one variable at a time is the difference between learning and guessing. Treat the numbers below as rules of thumb to calibrate against your own account, not laws.

The four metrics that matter

Most dashboards show you 40 columns. For diagnosing a single creative, four do almost all the work, and they form a funnel: hook rate, hold rate, CTR, and CPA. The power of reading them in order is that they tell you where the ad breaks, not just that it broke.

• Hook rate: 3-second video views divided by impressions. This measures whether the opening stops the scroll.
• Hold rate: viewers who reach 15 seconds or the 50 percent mark divided by impressions. This measures whether the body keeps attention.
• CTR (link click-through rate): clicks divided by impressions. This measures whether the ad creates enough desire to act.
• CPA (cost per acquisition): spend divided by conversions. This is the bottom line the other three feed into.

Read them as a chain. A weak hook rate means fix the first three seconds. A good hook rate but weak hold rate means the body loses people. Strong hold but weak CTR usually means the offer or CTA is soft. And good CTR with bad CPA points at a landing page or offer problem downstream of the ad itself.

Concrete kill thresholds

These are working defaults many direct-response buyers use as a starting point. Adjust them to your niche, price point, and platform, but having a written line in the sand is what stops the slow bleed.

• Hook rate under about 20-25 percent: the opening is failing. Recut the hook or kill it. Healthy is roughly 30 percent or higher.
• Hold rate under about 10-15 percent: the body isn't holding attention. Strong creatives often sit above 20-25 percent.
• Link CTR under about 1 percent on cold traffic: the ad isn't creating enough desire. Above 1.5-2 percent is a good sign.
• CPA at 1.5x your target with statistically meaningful spend: cut it. If it's drifting up over several days with no path back, cut it.
• No conversions after spending roughly 2-3x your target CPA: kill it rather than hoping.

Decide the kill threshold before you launch. The moment you're negotiating with a losing ad in real time, you've already lost the argument.

How long to wait before you judge

Killing too early is as costly as killing too late. Two things must be true before you trust a verdict: enough spend and enough time. On spend, a common rule of thumb is to give a creative at least one to two times your target CPA before judging conversions, and to look at hook and hold rate only after a few thousand impressions so the ratios stabilize.

On time, respect the learning phase. New creatives usually need a few days and a meaningful conversion count before the delivery system settles and the numbers mean anything. Judging a creative on its first afternoon, or toggling it off and on repeatedly, resets learning and guarantees noisy data. Set a review window, hold your nerve until it closes, then decide.

• Upper-funnel metrics (hook rate, hold rate) stabilize fast - a few thousand impressions is often enough to read them.
• Conversion metrics (CPA) need patience - wait for meaningful spend and a handful of conversions, not one.
• Avoid mid-day knee-jerk cuts; look at multi-day trends, ideally accounting for your typical attribution window.

Test one variable at a time

The reason to isolate metrics is the same reason to isolate variables: so a result teaches you something reusable. If you change the hook, the music, and the CTA all at once and CPA improves, you have no idea which change did it, and you can't apply the lesson to the next creative. Change one thing per version. Keep the body fixed and rotate hooks. Keep the hook and swap the CTA. Keep both and test a new offer.

This is where a diagnostic funnel pays off twice. When hook rate is your weak link, you test hooks in isolation and read hook rate. When hold rate lags, you test the body. You are not guessing which lever to pull, because the metrics already told you. If you want a structured place to run these tests, the AI Video Ad Kit's hook formulas and UGC scripts are built to be swapped one variable at a time so you can attribute wins cleanly.

Put the whole system together and the decision gets calm. Write your kill thresholds before launch, wait for enough spend and enough time, read the four metrics in funnel order to find the break, and test one variable at a time so every kill and every winner teaches you something. Do that and you stop mourning dead creatives and start compounding what you learn from them.