When to Turn Off a Meta Ad: The 1.5x CPA Rule and Other Kill Points

Here is the rule we run across Spark client accounts: in a creative testing ad set, turn off a Meta ad once it has spent 1.5x your target CPA without a single conversion. Target CPA $40, spend hits $60, zero purchases: kill it and move the budget to the next test. That is the short answer to when to turn off a Meta ad. The longer answer is that the 1.5x rule applies to exactly one situation (a fresh creative in a test set), and applying it anywhere else, to retargeting ads or freshly scaled winners, is how operators kill ads that were about to pay for the quarter.

This post lays out every kill point as a decision table, works the maths behind each threshold, then covers the four situations where the right move is to sit on your hands.

The kill point decision table

Seven situations cover almost every "should I pause this?" moment you will hit in Ads Manager. The table is the verdict. The sections after it are the reasoning.

Two clocks run through that table, and they answer different questions. The 1.5x rule is a test set rule: it exists to triage new creatives fast and cap the cost of each loser. Scaling decisions run on a slower clock: day 10 or later, 20 to 30 conversions, trends over snapshots. Most bad pausing decisions come from applying the fast clock to the slow question.

What is the 1.5x CPA kill rule?

A new creative in your testing ad set gets a budget of 1.5x your target CPA to produce one conversion. If it converts, it earns a longer look. If it spends through 1.5x with nothing, it dies. No debate, no "one more day".

Worked example. Your target CPA is $40. A new UGC ad goes live in the test set on Monday. By Thursday it has spent $61 with zero purchases. It is dead. Not "let it finish the week" dead. Dead now, and the next creative in the queue takes its slot.

The threshold is not arbitrary. At 1.5x spend with zero conversions, the best remaining case is already grim: if the ad converted on its very next impression, its test CPA would land at $60 against a $40 target, a 50% miss. Every dollar past the kill point buys a worse version of an answer you already have. The rule caps your downside at 1.5x target CPA per losing creative, and that cap is what lets you test aggressively without torching the account. The upside half of the same maths, how much to give each creative for a fair read, is in our testing budget breakdown: 2x to 3x target CPA per creative, with a $100 to $150 floor.

One process note from running this across client accounts: we write the kill point into Meta's automated rules at launch, before the ad spends a penny. Not because the maths is hard. Because the moment of decision is exactly when operators decide worst. At $58 of spend against a $40 target, every buyer alive can talk themselves into "it feels close". The rule exists so nobody has that conversation with the dashboard at 9pm.

The 1.5x rule is a triage rule for the test set. It caps what a losing creative can cost you. It was never meant to judge a scaled campaign.

How long should you let a Meta ad run before pausing it?

For a test creative: until it converts or hits 1.5x target CPA in spend, whichever comes first. For anything already scaled, the honest answer is longer than feels comfortable.

Days 1 and 2 are noise, for two reasons. The first is attribution lag. Meta reports conversions against the day of the click, not the day of the purchase, so an ad's early numbers keep getting revised upward for days after the fact. Five Nine Strategy's breakdown of attribution lag covers how campaigns that look like disasters at 48 hours routinely backfill into winners by day 7. Kill on day 2 CPA and you are grading an exam that is still being marked.

The second is the learning phase. Meta's own learning phase documentation says an ad set needs roughly 50 optimisation events within 7 days to exit learning, and that performance is expected to be volatile until it does. A scaled ad set judged mid-learning is being judged on delivery Meta itself considers unstable.

So put numbers on the slow clock: wait until day 5 to 10 and 20 to 30 conversions before making a scaling read, and closer to day 14 before final cull decisions on ad sets that carry real volume. Coinis puts it well in their kill framework: day 2 CPA is a draft, not a verdict. But once an ad has 10+ days and real conversion volume behind it, the number is the ad. CPA still sitting 50% or more above target at that point is no longer noise.

Not everyone runs 1.5x on the test set, to be fair. Roaspy's 2026 scaling timeline recommends 2x target CPA with zero conversions as the phase 1 kill line, plus holding all kills inside the first 72 hours unless a hard threshold trips. That is a defensible, more patient version of the same idea. We run 1.5x because the extra 0.5x rarely changes the verdict and always changes the bill: on a $40 target, killing 8 losers a month at 1.5x instead of 2x saves $160, which funds one entire extra creative test at the $120 floor.

The other kill points, unpacked

3x target CPA with zero purchases is not a creative problem

One creative dying at 1.5x is Tuesday. A whole batch blowing through 3x target CPA with zero purchases is a different diagnosis. The traffic existed, the clicks happened, and nothing converted, which points past the ads to the machine behind them: landing page speed, offer strength, price shock at checkout, a broken pixel, a form that dies on mobile. Pause the spend, yes. But do not brief another round of creative until the page and the offer have been audited, because the next batch will hit the same wall. Median ecommerce landing pages convert around 1.6% of clicks (the full breakdown is in our 2026 Meta benchmarks by business type); if yours is at zero across hundreds of clicks, the ad was never the variable.

CPA 50% over target after day 10 on a scaled ad

This is the scaled equivalent of the 1.5x rule. Once an ad has 20 to 30 conversions and 10+ days of delivery behind it, the excuse column is empty: it has had every chance to optimise, and what you see is what it costs. An ad set delivering $60 CPAs against a $40 target at that volume gets killed, or, if it carries volume you cannot instantly replace, gets its budget stepped down 20% while a replacement tests.

Ads Meta has already stopped spending on

Inside a live ad set, Meta does not split budget evenly. It runs an auction-level tournament and concentrates spend on whichever ad it predicts will win, a behaviour that has only sharpened under Andromeda. Which produces our second house rule: in a scaled ad set, only pause the ads Meta has already stopped spending on. An ad at near-zero delivery for a week has been killed by the auction; pausing it is admin, not a decision. The reverse also holds. An ad still pulling daily spend in a healthy ad set has Meta's vote, and manually pausing it forces a budget redistribution the ad set did not ask for.

When should you not turn off a Meta ad?

Four situations where the pause button is the mistake.

Retargeting ads with poor ROAS. Never kill retargeting on ROAS alone. A retargeting ad sits at the end of a journey your prospecting ads started, so its reported ROAS is a function of attribution windows as much as performance, and the ad may be warming an audience that converts through another touchpoint entirely. Judge retargeting on frequency (past 4 to 6 per week, fatigue is doing damage), audience freshness, and your blended account CPA. Kill it when the audience is exhausted, not when one ROAS number dips.

Anything in its first 48 hours. Attribution lag plus learning volatility make day 1 and 2 numbers close to worthless for kill decisions. The one exception is the hard kill: a test creative that has already spent through 1.5x target CPA. The rule outranks the calendar.

An ad still taking spend in a winning ad set. Covered above. Meta's budget allocation is a vote, and it has more data than your gut does.

A winner you want to "rest". Pausing a fatigued winner to bring it back fresh later sounds tidy, but Meta's guidance is that ad sets paused for 7 days or more re-enter the learning phase on resume. You will not get the same ad back; you will get a reset. If a winner is fatiguing, the play is iteration (a new hook on the proven body, a static cut of the winning video), not a nap.

Edge cases worth knowing

You do not have a target CPA yet. New account, new product, no history. Set a provisional target from unit economics (contribution margin per order minus the profit you want to keep) or, failing that, from category benchmarks, and enforce the rule against that. A soft target with a hard rule beats a hard target you never act on.

High-ticket targets. If your target CPA is $200, the 1.5x rule prices each losing creative at up to $300. That is not a reason to abandon the rule. It is the real cost of testing at that conversion event. If the budget cannot carry it, optimise for a shallower event (qualified lead, booked call) and set the kill point against that target instead.

Lead gen with an offline close. The rule works on cost per lead, with one caveat: audit lead quality weekly. A $15 CPL ad filling the pipeline with junk is worse than the $32 CPL ad it beat, and no ad-level kill rule will tell you that. The CPL kill point protects your budget; only your CRM protects your close rate.

Tiny budgets. At $50 a day the rule still holds, you just run one or two tests at a time and each verdict takes longer to arrive. Slower reads are fine. Reads without a kill point are how small accounts die.

Kill rules only work if you can reload

Here is the part that never makes it into the threshold threads. A kill rule is only as good as your willingness to pull the trigger, and operators hesitate for one reason above all: nothing ready to replace the dead ad. When killing a creative leaves an empty slot in the test set, the rule quietly gets renegotiated. One more day becomes one more week, and dead spend piles up in the name of patience.

The accounts where the 1.5x rule actually gets enforced are the ones with a creative system feeding the test set: the next batch briefed before this batch reads out, replacements sitting in the queue. That pipeline of test creatives, across AI and real creators, is the thing we build for brands, priced so that testing at real volume stays affordable (pricing here, and the rest of our operator playbooks live in the resources library).

If your kill decisions keep sliding because the test set has no bench, start a project. We will keep the queue full enough that killing an ad costs you nothing but the click.

Frequently asked questions

Does pausing a Meta ad reset the learning phase?

Short pauses usually do not. Meta's guidance is that an ad set paused for 7 days or more re-enters the learning phase when it resumes, while shorter pauses generally pick up where they left off, with some volatility either way. Pausing a proven winner is rarely free.

Should I turn off a retargeting ad with poor ROAS?

Not on ROAS alone. Retargeting sits at the end of a journey your prospecting ads started, so its reported ROAS reflects attribution as much as performance, and the ad may be warming an audience that converts elsewhere. Judge it on frequency, audience freshness and blended account CPA, and kill it when the audience is exhausted, not when one ROAS number dips.

How many conversions do I need before trusting CPA?

For a scaled ad, wait for 20 to 30 conversions and at least 5 to 10 days of delivery before making scaling or kill decisions. Meta's learning phase wants roughly 50 optimisation events in 7 days before delivery stabilises. Test creatives are the exception: 1.5x target CPA in spend with zero conversions is enough evidence to act.

What if every ad in my test batch hits the kill point?

A whole batch spending past 3x target CPA with zero purchases points to a structural problem, not a creative one. Audit the landing page, offer, price and pixel before briefing more creative, because the next batch will hit the same wall.