When to Kill an Ad: A Practical Decision Framework

Kill an ad when it has spent enough to be statistically meaningful for your account (not a fixed dollar figure — that depends entirely on your margin, average order value, and typical conversion volume) and it is still underperforming your existing baseline on the metric that matters most to you, whether that's ROAS, cost per purchase, or hook rate. The exact number is yours to define; the discipline of defining it in advance and sticking to it is what actually saves budget.

What signals actually matter?

Three signals tend to matter more than any single dashboard number. First, is the ad getting enough impressions and clicks to trust the data, or is the sample still too thin to mean anything? Second, is the hook rate low relative to your other creative — a weak hook means people are dropping before your offer even lands, which points to a creative problem rather than a targeting or pricing problem. Third, is your effective ROAS below the breakeven point for your margin structure, and has it stayed there across enough spend to rule out normal auction noise.

How much spend is enough to judge an ad?

There's no universal threshold, because it depends on your price point, your typical conversion rate, and how volatile your niche's auction is. A guideline that holds across accounts: don't judge an ad until it has had a real chance to exit the learning phase and accumulate a sample size where a handful of extra conversions wouldn't flip your conclusion. If you're killing ads after a few dollars of spend, you're probably reacting to noise, not signal.

Is a low hook rate always the reason?

Not always, but it's the first thing worth checking. A weak hook rate points squarely at the first two to three seconds of the creative — the wrong opening line, a slow visual, or an unclear subject. If your hook rate is healthy but conversions still lag, the issue is more likely the offer, the landing experience, or targeting rather than the ad itself. Testing hook variations in isolation with a tool like the hook generator can help you separate a hook problem from an offer problem.

Should you kill or just pause and iterate?

If the underlying concept has promise but the execution is weak — say the offer resonates but the hook is flat — pausing to iterate a new variant is often smarter than killing the whole concept. This is where creative volume changes the math: when producing a new UGC-style variant is fast and cheap, you can afford to test three or four hook and pacing variations on a promising concept instead of abandoning it after one weak cut. Polaris is built for exactly this — you describe the variant you want directly in a Claude chat, attach your product photo, and get a rendered ad back in under three minutes, so iterating on a borderline ad costs you a conversation, not a production cycle.

How do you avoid killing ads too early?

Set your kill criteria before you launch, not while you're staring at a dashboard mid-flight. Decide your minimum spend threshold and your target metric in advance, and don't move the goalposts based on how the first few hours look. Run your numbers through a ROAS calculator to know your real breakeven point given your margin, so "underperforming" is a defined number, not a gut feeling.

What does this mean for your testing pace?

The accounts that consistently find winners are the ones testing more concepts, not the ones agonizing longer over each one. If you can only afford to make one new ad a week, every kill decision is high-stakes and painful. Read more on generating ads inside Claude via MCP, or see how AI UGC ads fit into a volume-based testing approach.