All articles

How to Scale Facebook Ads: Which Ads to Turn Off, Keep and Scale

Short answer: scale Facebook ads in two steps. First, decide which ads deserve more money: set a target cost per purchase (or ROAS) and a test spend threshold, pause ads that passed the threshold above target, and keep the ones at or under it. Second, add budget gradually, around 20% at a time a few days apart, and judge the result on your marginal cost per purchase, what the extra money bought, not on the new average.

Open Ads Manager in almost any small business account and you find the same situation: a long list of active ads, a few with purchases, many with a handful of clicks, and no clear idea which ones deserve more budget. And when one campaign finally works and gets a bigger budget, the cost per purchase climbs a few days later.

Both problems have the same solution: a rule you apply the same way to every ad, and a way of measuring what extra budget actually buys.

Before you scale: the two numbers you need

Your target cost per purchase

You cannot call an ad a winner or a loser without knowing what an acceptable cost per purchase is for your business. That number comes from your margins, not from industry benchmarks. If a product leaves you $30 of gross profit, an ad at a $45 cost per purchase loses money on every sale, however good its click-through rate looks. The Breakeven ROAS Calculator gives you your maximum affordable cost per purchase in a minute; your target should sit below it.

If your order values vary a lot, use a target ROAS instead: a $45 cost per purchase is fine on a $200 order and terrible on a $40 one. What Is ROAS? explains how to set it.

Your test spend threshold

This is how much an ad must spend before you judge it. Without it, you compare an ad that spent $8 with one that spent $300, which tells you nothing. A practical threshold is one to three times your target cost per purchase, depending on how much you can spend on testing. Applying it consistently matters more than the exact number.

Which ads to turn off, keep and scale

With those two numbers, every active ad falls into one of four groups.

Spend vs thresholdResult vs targetGroupAction
BelowAnyNot enough dataWait
AboveClearly worse, or no purchasesUnderperformerTurn off
AboveAt or betterWinnerKeep, don't edit
AboveBetter, stable for 2+ weeks, small share of budgetScaling candidateAdd budget gradually
  • Underperformers: pause them. First look at what they have in common (same angle, same format, same audience); that pattern is worth more than the individual ads.
  • Not enough data: give them time or budget until they reach the threshold. If an ad gets no delivery at all, replace it with a clearly different angle rather than a close copy; how Meta Andromeda picks ads explains why.
  • Winners: keep them running and avoid editing them. Edits can disrupt delivery on ads that are doing their job.
  • Scaling candidates: these are where extra budget should go.

Mistakes that lead to the wrong decision

  • Judging by click-through rate alone. A high CTR means people find the ad interesting, not that they buy. Judge on cost per purchase once the ad has enough spend.
  • Looking at the wrong time window. A lifetime winner may have been declining for two weeks; a 3-day loser may just have had a slow start. Check lifetime results and the last 7 days against the 7 before.
  • Waiting forever on ads Meta barely spends on. If an ad has had little delivery for a couple of weeks, Meta has made its choice. A new idea is more useful than waiting.
  • Trusting Meta's purchase count completely. Meta usually misses some purchases. If it misses 20%, an ad reporting a $36 cost per purchase against a $30 target is really closer to $29, a winner. Compare with your store's orders before judging ads close to your target.
  • Pausing everything at once. Turning off half the account in one afternoon makes the next week impossible to read. Act on the clear cases first.

How to scale without losing your results

Why results get worse when you spend more

  • Meta has to reach less responsive people. At a low budget, Meta spends on the people most likely to buy. More budget means going further, to people who are a bit less ready. Each additional purchase costs a little more than the last.
  • Large changes disrupt delivery. A big budget jump can put an ad set back into the learning phase, with less stable delivery and higher costs for a few days.
  • The same people see your ads more often. With a narrow audience, more budget mostly means more impressions to the same people. Frequency rises and the ads wear out faster.
  • The extra budget goes to weaker ads. Ads that were barely spending may now get a meaningful share, and pull the average down.
  • Duplicated campaigns compete with each other. Scaling by duplicating a winner often means two campaigns bidding for the same people.

The number that matters: marginal cost per purchase

The average after scaling hides what the extra money did. What you want to know is how much each additional purchase cost:

Marginal cost per purchase = (spend after – spend before) ÷ (purchases after – purchases before)

An example. At $100 a day you got 5 purchases a day, a $20 cost per purchase. You doubled to $200 a day and now get 8 a day. The average is $25, which looks fine. But the extra $100 bought only 3 extra purchases, about $33 each. If your maximum affordable cost per purchase is $30, the extra budget is losing money, even though the average looks healthy.

Use whole days, leave out the last two or three (Meta keeps attributing purchases for days after the click), and check purchase counts against your store's orders: with a few purchases a day, one missing order changes the result a lot.

How to scale step by step

  1. Scale the right campaign: the one whose winners are stable and not yet getting most of the budget.
  2. Increase in small steps: around 20% at a time, a few days apart. Small steps are easier to judge and less likely to restart learning.
  3. After each step, calculate the marginal cost per purchase against your breakeven.
  4. Check where the extra spend went: to your winners, or to weaker ads that started spending.
  5. Watch frequency and click-through rate: rising frequency with a falling CTR means the audience or the creative is reaching its limit.

See which ads to pause and where extra budget went. Our Meta Ads Analyzer sorts every ad into underperformers, winners, scaling candidates and fading ads against your own target and test spend threshold, compares periods at every level and corrects for the purchases Meta misses. Start free

What to do if scaling hurt your results

  • Marginal cost per purchase above breakeven: step back to the last profitable budget. Scaling further only makes it worse.
  • Extra spend went to weaker ads: pause the ones past your threshold without reaching target.
  • Your winners are wearing out: add new creative angles before the next increase; scaling with the same two ads runs into the same wall. See Facebook Ad Fatigue. Our AI Ad Generator can build those new angles from the ads that already work for you.
  • The audience is saturating: broaden it, or give Meta more room with fewer targeting restrictions.

If you scale automated campaigns, Advantage+ Shopping Campaigns Not Performing covers the specific issues there.

Frequently asked questions

How much should I increase my Facebook ads budget at once?

Smaller steps spaced a few days apart are easier to judge and less disruptive than one big jump. A common rule of thumb is around 20% at a time, waiting a few days before the next step.

How long should I let a Facebook ad run before turning it off?

Judge by spend, not days. Let an ad reach your test spend threshold, typically one to three times your target cost per purchase, before you decide. With a small budget that can take a week or more.

Is it normal for ROAS to drop when scaling?

Some drop is normal, because each additional purchase costs a bit more than the previous one. What matters is whether the extra purchases are still profitable: compare your marginal cost per purchase with your breakeven.

Should I duplicate a winning campaign or increase its budget?

Increasing the budget keeps what the campaign has learned. Duplicating starts from scratch and can make the two campaigns compete for the same people. Duplicate to test a different audience or setting, not as the default way to scale.

Should I turn off an ad with a high CTR but no sales?

If it has passed your spend threshold without purchases, yes. Check first whether other ads convert on the same page; if none do, the problem is after the click. See Facebook Ads Not Working?

Is it better to pause an ad or delete it?

Pause it. A paused ad keeps its history, so you can later see which angles you already tried.

How many ads should I run per ad set?

Enough that Meta has real options, few enough that each can reach your spend threshold. With a small budget, three to six clearly different ads usually beat a dozen that never get enough spend to judge.


Stop deciding by feel. Connect your ad account, set your target, and see which ads to pause, which to keep and where to add budget. Try our Meta Ads Analyzer for free

Related: Facebook Ads Not Working? How to Find the Real Cause

More articles

Find out what's costing you money.

Tell us about your business and current setup. We'll tell you what we'd fix first and what it would cost. No pressure, no obligation.

Book a free consultation