All answers

The Prevailing
Answers for
Business model:
Business model, pick up to two

When should we turn off a Meta campaign that isn't working, and how long do we give it first?

The PrevailingPublished Nov 17, 2026

Our answer

Write the stop rule before launch. Give a new campaign at least a week and 3 times your target cost per purchase, then turn it off if it has one purchase or none. Stop sooner only for a broken setup.

Confidence
Medium
Strongest proof
3 Meta rules
4 practitioners, 4 opinions or claims
Sources
8 sources
11 statements

New here? Take the 1-minute tour

Why this is the answer

4 reasons and our judgment

Early results mislead. Meta says ad sets in learning are less stable and usually cost more per result, and it advises waiting to edit until learning ends. Coinis, an ad tool maker, argues that a kill decision before day 5 is a reaction.

Spend against your target, not days live, decides the read. Flighted spends at least 3 times the target cost per purchase, then cuts with one purchase or none. The Social Outline uses the same floor, and BrandRocket goes up to 5 times.

Judge the campaign, not its weakest ad. BrandRocket reads 7, 14 and 30 days and waits out a bad week inside a good month. It changes a campaign for broken structure or economics, such as a margin too thin for what a customer costs.

Switching off has a cost. Pausing an ad set for 7 days or longer restarts learning (Meta). Ecommerce Alley warns that cutting campaigns on a bad Meta day can stretch it into a week-long dip.

Our judgment

Give a new campaign at least a week, because Meta’s learning window is a week, Flighted’s standard test runs 7 days and Coinis makes its first real read from day 7. Apply the spend floor to the whole campaign, since BrandRocket puts the decision at the ad set and campaign level while the others write their floors for new ads. Use 3 times the target as the floor because Flighted, Coinis and The Social Outline share it and it is the low end of BrandRocket’s range, and AdManage.ai’s model puts the chance of zero purchases at that spend at about 5% if the target cost were true.

Our judgment, built from Meta, Flighted, Coinis, The Social Outline, BrandRocket and AdManage.ai: our reading of these sources, not a finding from any one of them.

What the sources say

11 statements from 8 sources
  • MetaUndatedSupports

    Meta says ad sets exit the learning phase after about 50 results in a week, warns that many ads and ad sets slow learning, and recommends combining similar ad sets.

    Read it on facebook.com

  • AdManage.aiOct 31, 2026Disagrees

    AdManage.ai, an ad rules tool seller, says no universal 3-day or 3 times cost per acquisition rule exists, so set a loss limit before launch. Its simple model gives a 5% chance of zero sales at 3 times the assumed cost.

    “There is no universal three-day or 3×-CPA rule that answers every case.”

    Read it on admanage.ai

  • MetaOct 15, 2026Supports

    Meta says ad sets in the learning phase are less stable and usually cost more per result, so wait to edit until learning ends and avoid frequent budget changes, but do not try to avoid learning entirely.

    “During the learning phase, ad sets are less stable and usually have a higher CPA.”

    Read it on facebook.com

  • MetaOct 15, 2026Caveat

    Meta says targeting, creative or optimization event changes and adding a new ad are always significant edits, while a budget change may or may not restart learning, depending on its size. No safe percentage is given.

    “if you increase your budget from $100 to $101, that isn't likely to cause one or more ad sets to reenter the learning phase.”

    Read it on facebook.com

  • BrandRocketAug 19, 2026Supports

    BrandRocket, an ad agency, says act on an ad at about 3 to 5 times target cost per result with no results, and read results over 7, 14 and 30 days before deciding.

    “Make the three-window read a habit. Seven, fourteen, thirty.”

    Read it on brandrocket.net

  • The Social OutlineAug 14, 2026Supports

    The Social Outline says judge a Meta ad only after it spends about three times target cost per result over at least three to five days, reading a seven-day window. No account data.

    “Never kill anything on a single day of data”

    Read it on thesocialoutline.com

  • Common Thread CollectiveAug 6, 2026Caveat

    Taylor Holiday says ad-level on/off decisions had no discernible impact in Common Thread Collective's research. An internal rule to switch off ads at 5x CPA with zero purchases did not work and hurt account structure.

    “doesn't work”

    Read it on youtube.com

  • FlightedJul 1, 2026Supports

    A practitioner with no data says spend 3x target CPA before judging an ad, and set test budget as target conversions x CPA / test days. He calls Meta's 50-in-7-days benchmark impractical at a $100 CPA.

    Read it on flighted.co

  • FlightedJul 1, 2026Supports

    Flighted, an agency, judges an ad once it spends three times its target cost per acquisition: 3 or more conversions keep it, 2 get more spend, 0 or 1 cut. It kills ads early only for no delivery or rejection.

    Read it on flighted.co

  • Ecommerce AlleyMay 21, 2026Supports

    The Ecommerce Alley podcast says rash changes on a bad Meta day, like cutting budget or pausing campaigns, can turn a one or two day dip into a week or more.

    “people would kind of freak out and make like rash decisions that then extended a one or two day bad day into like a week plus”

    Read it on feeds.buzzsprout.com

  • CoinisMar 16, 2026Supports

    Coinis, an ad tool vendor, says to kill a Meta ad that has spent three times Target CPA with no purchases, but otherwise not to judge before day five, with the first real read at days 7 to 14.

    “If you’re making kill decisions before day five, you’re not optimizing, you’re reacting.”

    Read it on coinis.com

In the full answer

For readers who sign in
  • What to do this week3 steps
  • The numbers behind it3 figures from cited sources
  • Where each source standsA chart placing each source on the question
  • The strongest disagreementThe best case against the answer, from a cited source
  • How sure we areWhat its confidence rests on, and what would change the answer

Read next

More in Bidding and optimization