When to Kill a Facebook Ad: The Exact Numbers That Decide

Kill a Facebook ad when its ROAS is below half your breakeven ROAS after at least 3 days and $100 of spend. Breakeven ROAS is 1 ÷ your gross margin. At a 50 percent margin you need 2.0x to break even, so anything under 1.0x is a kill. An ad with zero revenue after $100 across 3 or more days is also a kill.

If you just want the verdict on a specific ad, run your ad's numbers through the free calculator in 60 seconds. The rest of this post is the framework behind it: where those thresholds come from, the sample gates that stop you from reading noise, what to check before you turn off an ad, and a worked example with real dollar figures.

#Table of Contents

• Why do founders kill ads too early?
• Why do founders kill ads too late?
• The four-step framework for killing a losing ad
  • Step 1: Know your breakeven ROAS
  • Step 2: Pass the minimum sample gates
  • Step 3: Apply the kill, iterate, scale matrix
  • Step 4: Check four things before you kill
• How many creatives should a $1M to $10M brand be testing?
• A worked example: one ad through the framework
• FAQ

#Why do founders kill ads too early?

You launch a new creative Monday afternoon. Tuesday morning it has spent $38 with no sales, and Ads Manager shows 0.0x. You kill it and tell yourself you were being disciplined.

You were flipping a coin. At $38 of spend and a few hundred impressions, one $60 order was the difference between a 0.0x read and a 1.6x read. You did not learn anything about the creative. You paid $38 to generate noise, then acted on the noise.

There is a mechanical reason day-one reads mislead. Meta runs every new or significantly edited ad set through a learning phase, and Meta's own documentation says an ad set needs roughly 50 optimization events within 7 days of its last significant edit before delivery stabilizes. During learning, the system is still hunting for the pocket of your audience that converts, so cost per result swings hard day to day. Judging an ad on day one means judging the algorithm mid-search.

Now put a dollar figure on the habit. Say you test eight creatives a month and kill anything without a purchase in the first 24 hours, roughly $40 per ad. That is $320 a month spent to earn zero decisions, because none of those tests reached $100 of spend or 1,000 impressions, the floor where a read starts to mean something. Across the Shopify stores Arlo analyzes, it is common for a creative that ends up a durable winner to show its first purchase on day 2 or 3, not hour 6. Impatient founders kill the winner along with the losers, then conclude paid social does not work for their brand.

#Why do founders kill ads too late?

The opposite failure is quieter and costs more. An ad that produced last quarter starts slipping, and you keep it alive because it used to work, because you paid an editor for the footage, or because pausing it feels like admitting the account is in trouble. That is sunk cost, and Meta will happily keep spending your budget while you work through it.

Here is what a slow kill costs. Your gross margin is 50 percent, so your breakeven ROAS is 2.0x. The ad is returning 0.9x on $50 a day. Each day it brings in $45 of revenue, which is $22.50 of gross profit against $50 of ad spend. You are losing $27.50 a day, about $825 a month, on one ad set. Under the framework this is not even a judgment call: 0.9x is below half of your 2.0x breakeven, which makes it a same-day kill.

Blended reporting makes late kills easier to rationalize. If your account-level numbers look acceptable, a single bleeding ad set hides inside the average. That is one reason to track MER alongside ROAS and ROI: account-level efficiency tells you the account is drifting, but only per-ad math against breakeven tells you which ad to turn off.

#The four-step framework for killing a losing ad

Four steps, in order. Skipping one is how the too-early and too-late mistakes happen.

#Step 1: Know your breakeven ROAS

Meta shows you ROAS. It cannot show you whether that ROAS is any good, because it does not know your gross margin. The conversion is one division:

Breakeven ROAS = 1 ÷ gross margin

Gross margin here is revenue minus cost of goods, shipping to the customer, and transaction fees, divided by revenue. Use the real number, not the one you quote at dinner parties.

This is why "what's a good ROAS?" has no universal answer. A 2.2x ROAS is profitable for a 70 percent margin skincare brand and a slow leak for a 40 percent margin apparel brand. Remember also that ROAS measures revenue efficiency, not profit. If the distinction is fuzzy, read ROAS vs ROI before you scale anything.

#Step 2: Pass the minimum sample gates

Before you judge any ad, it must clear all of these:

• At least 3 days running. Fewer than that and you are reading day-of-week and learning-phase variance, not performance.
• At least $100 of spend. Below this, single orders swing the ROAS reading too much to trust.
• Ideally 1,000 or more impressions. A cheap proxy for whether Meta has actually shown the ad to enough people.

If the ad has not cleared the gates, the verdict is wait, not kill. Hold the spend, give it the rest of the week, and check back. These are floors, not targets: an established brand spending $200 a day per ad set clears them in a day or two, but a $30-a-day test cell needs the full window.

#Step 3: Apply the kill, iterate, scale matrix

Once the gates are passed, compare current ROAS to breakeven ROAS. The verdict falls out of where it lands. The example column uses a 50 percent margin, so breakeven is 2.0x.

Two things about this matrix. First, the kill line is half of breakeven, not breakeven itself. An ad between half of breakeven and breakeven is losing money but might be one fix away from working, which is why it earns a pause and an iteration rather than a delete. Second, the scale bar is deliberately high. Scaling an ad that sits at 2.1x against a 2.0x breakeven means funding an ad with no headroom, and Meta performance usually degrades slightly as budgets rise.

#Step 4: Check four things before you kill

A kill or iterate verdict from the matrix is a strong prior, not a reflex. Before you act, spend five minutes ruling out causes that killing the ad will not fix.

Frequency and fatigue. A pattern Arlo sees constantly across the Shopify stores it analyzes: an ad clears breakeven for weeks, then slides while frequency climbs past 3 to 4 and CTR sags. That is audience fatigue, not a broken concept. The fix is a fresh variation of the same winning angle, not abandoning the angle.

The landing page. If CTR is healthy but conversion rate on the destination page is well below your site average, the ad is doing its job and losing the sale after the click. Killing it punishes the wrong asset. Check page speed, offer match, and whether the page shows what the ad promised.

Seasonality and account-wide movement. If every ad set dipped the same week, you have a demand or tracking issue, not one bad ad. Compare the ad against the account, not just against its own history.

Attribution drift. Meta's reported ROAS is modeled, and it can drift meaningfully from what your Shopify order data shows. Before killing a borderline ad, sanity-check platform revenue against actual orders. If your numbers never seem to reconcile, read why Meta's reported ROAS misleads Shopify brands.

If none of the four explains the miss, trust the matrix and act.

#How many creatives should a $1M to $10M brand be testing?

At $1M to $10M in revenue you are probably spending $20,000 to $150,000 a month on paid. At that level, killing losers is only half the job. The other half is a testing cadence that keeps producing candidates, because every winner fatigues eventually.

The pattern across stores Arlo analyzes in this revenue band looks like this:

• Reserve 10 to 20 percent of ad budget for testing. The rest funds proven winners. Brands that skip a testing budget end up over-scaling one tired creative because there is nothing on the bench.
• Test 2 to 4 new concepts a month, with 2 to 3 variations each. A concept is a genuinely different angle: new hook, new format, new proof. A variation is the same angle with a different opening three seconds or thumbnail.
• Give every variation its own sample. Each one needs the Step 2 gates, at least $100 to $150 of spend and 3 days, before it gets a verdict. Do not let Meta starve half your variants of delivery and then judge them on 12 impressions.

One more distinction saves a lot of wasted iterations: knowing when a string of losers is not a creative problem. If four or five consecutive concepts all die below half of breakeven, stop testing creative and look downstream. Healthy CTRs with weak conversion point at the offer, the price, or the landing page. Weak CTRs across every concept point at the hook or the audience, and that is a copy and angle problem; the fixes in our Facebook ad copywriting guide are the right tool there. Creative testing cannot rescue an offer nobody wants.

This framework is built for prospecting on Meta and TikTok, where the ad creates the demand. On Google, branded search will post spectacular ROAS because it harvests demand you already created, so judging it against the same thresholds flatters it unfairly. See our guide to Google Ads for Shopify for how to segment that spend before applying kill rules to it.

#A worked example: one ad through the framework

Alder & Flame is a fictional candle brand doing $2.4M a year on Shopify. Gross margin after COGS, shipping, and fees is 55 percent. One prospecting ad set, a UGC video hook, has been live for 9 days at $60 a day.

Step 1. Breakeven ROAS = 1 ÷ 0.55 = 1.82x. Half of breakeven is 0.91x. The scale line is 1.82 × 1.5 = 2.73x.

Step 2. Nine days and $540 of spend clear every gate. The read is trustworthy.

Step 3. A 1.5x ROAS sits between 0.91x and 1.82x, about 18 percent below breakeven. Verdict: iterate. Not a kill, but it is losing money on every order. The daily contribution is $60 × (1.5 × 0.55 − 1) = negative $10.50 a day, roughly a $315 loss over the next 30 days if nothing changes.

Step 4. Frequency is 2.1, no fatigue. CTR is 1.9 percent, healthy. But the landing page converts at 1.1 percent against a site average of 2.4 percent, and the ad promises a three-candle discovery set while the link goes to a single-candle product page. The ad is fine. The click is being wasted.

The founder pauses the ad set, points the same creative at the discovery-set bundle page, and relaunches. After the new variant clears the gates, it reads 2.9x, which is 59 percent above the 1.82x breakeven. Verdict: scale. Budget goes from $60 to $75 a day, a 25 percent bump, with a recheck in 3 to 4 days.

Notice both failure modes this avoided. Killing at day 2 would have thrown away a working creative with a fixable landing page problem. Letting it run untouched would have burned $315 a month indefinitely. The framework took about ten minutes to apply, and you can compress those ten minutes further: drop your own ad's five numbers into the free calculator and get the verdict, your breakeven gap, and the 30-day projection instantly.

#FAQ

#How long should I let a Facebook ad run before judging it?

At least 3 days, with at least $100 of spend and ideally 1,000 or more impressions. Below those floors, day-to-day variance swamps the signal: a single order can swing your measured ROAS from 0.0x to 1.6x. Meta's delivery system is also still in its learning phase early on, so performance is volatile by design. Hold the spend until the gates are cleared, then judge.

#Is a 2x ROAS good?

It depends entirely on your gross margin. At a 50 percent margin, 2x is exactly breakeven: every order pays for itself and nothing more. At a 33 percent margin, 2x is 33 percent below breakeven and the ad loses money on every order. At a 70 percent margin, 2x is comfortably profitable. Always compare ROAS to your breakeven, which is 1 ÷ gross margin, never to a generic benchmark.

#What ROAS means I should kill a Facebook ad immediately?

Kill immediately when ROAS is below half your breakeven after the minimum sample of 3 days and $100 of spend. At a 50 percent margin that is anything under 1.0x. Also kill any ad showing zero revenue after $100 of spend across 3 or more days. In both cases the loss rate is too high for a creative tweak to fix, and every extra day is a measurable loss.

#Should I turn off the ad, the ad set, or the campaign?

Start at the level where the data is conclusive. If one creative inside an otherwise profitable ad set drags ROAS below breakeven, turn off that ad. If every creative in the ad set sits below half of breakeven after the sample gates, kill the ad set. Only kill a campaign when multiple ad sets fail the same test, which usually signals an offer or audience problem rather than a creative one.

#When should I scale a winning ad instead?

Scale when ROAS is at least 50 percent above your breakeven ROAS after at least 3 days and $100 of spend. Increase the budget by 20 to 30 percent, then recheck in 3 to 4 days. Bigger jumps can push the ad set back into Meta's learning phase and destabilize delivery. Scaling an ad that merely sits at breakeven means funding a marginal ad with no headroom.

---

Running this framework once is easy. Running it every week, across every live ad, against your real margins, while you also run the rest of the company, is where it breaks down. Arlo does the watching for you: it connects to your Shopify store and ad accounts, reads your live data every week, computes your actual breakeven, and tells you in plain English which ad to kill, hold, or scale. It costs $47/month with a 14-day free trial. Install Arlo on Shopify.