When to kill an ad: a practical guide to creative fatigue

Killing ads too late is the most common cause of monthly ROAS slumps. Killing them too early is the most common cause of stalled accounts. Here’s the framework we use to make the call.

The signals, in order

1. Frequency

At frequency < 1.5, you’re still in expansion. At > 2.5, the same humans have seen this ad 3+ times. Past 3.5, you’re paying Meta to remind people they aren’t buying.

2. CPM trend

Compare today’s CPM to last week’s, on the same audience. If CPM is up > 25% and frequency is up, the audience pocket is full.

3. Hold rate / 3-second video views

The earliest fatigue indicator. If 3-second view rate falls 10% over 5 days while spend is held flat, your hook is tiring — even if ROAS still looks fine. This is the leading indicator. Most others are lagging.

4. CTR

CTR drops are the noisiest signal. Don’t use it alone. Combined with frequency and CPM, it confirms.

5. ROAS

Lagging. By the time ROAS visibly crashed, you lost a week of spend. Don’t wait for this one.

When NOT to kill

• First 3–5 days. Learning phase. Bad numbers don’t mean bad ad.
• Spend < ₹3K total. Statistical noise.
• Right after a budget jump. Meta re-learns. Wait 48 hours.
• During a category-wide spike (Black Friday, Diwali). Everyone’s CPMs are up; yours rising isn’t a fatigue signal.

Killing without losing the win

When you kill a winner that’s fatigued, archive — don’t delete. Most fatigued ads come back to life after 4–8 weeks of rest. Re-running an old winner with refreshed targeting is one of the cheapest tests in the playbook.

The replacement rule

Don’t replace 1 killed ad with 1 new ad. The variance bus you’re on is creative diversity, not 1:1 substitution. Kill 1, ship 3 — ideally in different formats. The algorithm will pick the replacement.