Facebook Ads Kill Rules: When to Cut, Hold or Scale

Why fixed percentage rules fail

A rule that says pause when CPA is above X ignores sample size. Say an ad set has spent $20 against a target CPA of $5. One trial start puts CPA at $20, and two trial starts put it at $10. The ad is the same in both readings, so a rule that fires at $20 acts on noise.

Delivery adds more noise. Days 1 to 3 of a new ad set are unreliable for most accounts, because the platform is still testing placements and audiences, and conversions report with a delay. A morning CPA often looks worse than the final number for the same day.

Thresholds copied from a forum carry someone else's margin. Write each rule as a multiple of a target taken from your own payback math. The next section builds that target.

Start from your break-even CPA

Break-even CPA is the most you can pay for the event you optimize on before the cohort loses money inside your payback window. Multiply revenue per payer in the window by the share of optimized events that become payers, after store or processor fees.

Take an illustrative example. Say a payer brings $30 in 60 days, already net of fees, and 20% of trial starts convert to payers. Step 1: $30 x 0.20 = $6 of revenue per trial start. Step 2: a trial start can therefore cost up to $6 at break-even. If your $30 were gross and fees took 15%, you would first multiply by 0.85, which gives $25.50 and a break-even of $5.10.

Set the target CPA below break-even. Say you pick $5. The $1 gap per trial start is your room for error and for growth, since scaling usually pushes CPA up. The rest of this article uses $6 break-even and $5 target. None of these figures is a benchmark.

Kill rules by spend stage

Sort every ad by how much it has spent relative to target CPA. Spend decides how much the data can tell you.

• Stage 1, spend below 1x target CPA (under $5): leave the ad alone. Act only on broken delivery, such as zero impressions, a disapproved ad or a dead link.
• Stage 2, spend between 2x and 3x target CPA ($10 to $15) with zero trial starts: pause. An ad running at a $5 CPA would usually have produced two or three trial starts by then.
• Stage 3, spend above 3x target CPA (over $15): compare CPA and early payer rate against target. If CPA is above 1.5x target ($7.50) once the learning window you defined in advance has passed, cut the ad.

Between 1x and 2x target, keep watching and change nothing. Add a quality check at stage 3. I read traffic quality from conversion rate by source, placement and creative to spot weak or fraudulent traffic early. A placement with cheap clicks and no downstream payers deserves a cut even when the ad set total looks acceptable. The $7.50 kill line sits above the $6 break-even, so it tolerates a loss of up to $1.50 per trial start while payer rate is still being read. Lower the multiple if you cannot carry that loss.

Hold rules: when not to touch an ad

Large budget jumps can send an ad set back into learning, which makes the next days of data hard to read. Keep each budget change to 20 to 30 percent. Platform behavior shifts over time, so check the delivery status in Ads Manager before assuming an edit was safe.

After any significant edit to budget, bid, audience or creative, hold the ad set for 3 to 4 days. The stage 1 delivery checks and the stage 2 zero-conversion pause still apply during a hold, because $15 spent with nothing back is a loss whatever you changed.

Judge creatives by the metric you pay on, such as trial start or payer. An ad with a high click-through rate and weak trial starts may attract curiosity clicks. An ad with a modest click-through rate and a low cost per payer wins on the number that funds the business.

Scale triggers and how fast to scale

Scale when CPA stays at or below target for 3 consecutive days and spend is above your minimum read size. Pick that size in advance. Illustratively, 10 trial starts at a $5 target means 10 x $5 = $50 of spend.

Step the budget up 20% every 3 to 4 days, in line with the hold window. The rollback rule below applies inside that window. Say an ad set runs at $100 a day. The first step takes it to $100 x 1.2 = $120, and the second to $120 x 1.2 = $144. For a larger jump, duplicate the ad into a new ad set at the higher budget and leave the original running. Watch CPA drift after each step.

Use a rollback rule. If CPA rises more than 25% above target within 48 hours of a scale step, go back one step. At a $5 target the line is $5 x 1.25 = $6.25, which is $0.25 above the $6 break-even. In the example above, a rollback from $144 returns the budget to $120. Treat 25% as a starting value and tune it against your own data.

Creative fatigue rules

Track frequency alongside a 3-day moving CPA. A rising CPA with rising frequency usually points to creative fatigue rather than audience saturation. If CPA jumps while frequency stays flat, check tracking and the landing page first.

Retire an ad when its 7-day CPA crosses the kill line twice, and queue the next test variant. With a $5 target and a 1.5x kill line, that line is $7.50. Log each killed ad with a reason code so patterns show up over weeks. A starter set of codes:

• Zero conversions at 2x to 3x target spend
• CPA above 1.5x target after the learning window
• Fatigue: frequency up and 3-day CPA up
• Quality: placement or source with no downstream payers
• Delivery or approval failure

Turning the rules into automation

Meta automated rules can pause ads on cost and spend conditions, but they only see events the platform reports. Payer-based logic needs revenue from your own backend. A scheduled script, or a workflow tool such as n8n or Make, can read payments and pause or flag ads through the Marketing API.

Send a notification before a rule pauses a top-spend ad, so a person confirms the cut. Set an account spending limit in billing as a backstop in case a rule or script fails, and remember that a daily budget is an average that Meta can exceed by up to 25% on a given day. Review rule decisions weekly: false kills and missed kills show you where to move thresholds.

Start this week with four steps:

• Calculate break-even and target CPA from your own revenue per payer and payer rate.
• Write the stage 1, 2 and 3 thresholds, the hold window and the scale and rollback triggers in one sheet.
• Turn on the stage 2 pause as a notify-only rule and compare its calls to your own for a week.
• Review false kills and missed kills weekly, then move one threshold at a time.