Most founders kill Meta ads at the wrong time. Two days in, before the learning phase finishes, because ROAS is still zero. Or three weeks in, out of hope, while frequency is at 5 and they're paying premium CPMs to irritate people who already decided not to buy.
Neither is a budget problem. It's a decision-making problem. And it's fixable with thresholds.
The learning phase comes first
Before evaluating any ad, check whether it's exited the learning phase. Meta needs 50 conversion events in 7 days to optimize properly. Until then, every number you see — CPA, ROAS, CTR — is volatile. The algorithm is still exploring which sub-audiences, placements, and times of day produce results. Those early numbers aren't a verdict.
Clear signals to turn off an ad
CTR below 0.5% after 2,000 impressions
If CTR (All) is under 0.5% after 2,000+ impressions in the first three days, the hook isn't stopping anyone. That's a creative problem, not an audience problem. Don't try to optimize around it by swapping the targeting. Turn the ad off and rewrite the opening.
The 0.5% threshold isn't arbitrary. Below that, you're paying for eyeballs that aren't registering the ad at all.
Spend hit 3x target CPA with zero conversions
If you've spent ฿1,500 and your target CPA is ฿500 but there's been no conversion whatsoever, something's broken upstream. It's not about giving it more time. Check the offer, the landing page, and whether the audience actually wants what you're selling.
One exception: high-ticket products (฿5,000+) with longer decision cycles. For those, use add-to-cart or initiate checkout as proxy events instead of waiting for purchase.
Frequency above 4.0 in 14 days
The same people have seen your ad more than four times in two weeks. CTR will keep dropping. CPM will keep climbing. The audience is exhausted. You're paying to irritate people who've already decided not to buy.
Don't kill the campaign entirely. Pause the specific ad set, rotate in fresh creative, or expand to a new audience segment.
CPA still 2x target for 7+ days after learning phase
Once an ad exits the learning phase, you have real data. If CPA is still double your target after 7 days with no improving trend, it's not going to optimize its way to profitability. Turn it off and audit the offer before relaunching.
Clear signals to keep running
Low ROAS but only 2–3 days in
In the first week, Meta is still exploring. It might find a better-performing sub-audience on day five. If CTR is above 1% and frequency is under 2.0, the ad's doing its job. Give it the runway.
CPA is high but trending down each day
Day 1: ฿900. Day 3: ฿700. Day 5: ฿580. That's not a failing ad. That's a learning ad. Don't panic-kill it just because day one looked rough. Check again at day seven with the full picture.
Quick reference
| Signal | Action |
|---|---|
| CTR < 0.5% after 2,000+ impressions | Kill — hook isn't working |
| Spend at 3x target CPA, zero conversions | Kill — offer or audience problem |
| Frequency > 4.0 in 14 days | Pause + refresh creative |
| Still in learning phase (< 50 conversions) | Wait — data isn't stable yet |
| CPA high but trending down daily | Wait — check at day 7 |
| CPA > 2x target for 7+ days post-learning phase | Kill — it's not optimizing |
The trap: toggling ads resets everything
If you need to test new creative, don't pause the original ad set. Duplicate it and make the change in the new copy. The original keeps its optimization data. The new one builds its own.
The decision checklist
Four questions before you touch the off switch:
- Has it exited the learning phase? If not, wait.
- Have you spent 2x your target CPA? If not, you don't have enough data yet.
- Is CTR low from day one, or did it drop recently? Low from day one is a hook problem. Recent drop is creative fatigue.
- What's the frequency? Above 4.0 means the audience is tired, not the ad.
What to do next
AdBlueprint shows learning phase status and CPA trends for every ad in the Campaign Review panel. Pull it up before you make a kill call. Sometimes what looks like a dead ad just needs 48 more hours to find its footing. And sometimes it needed to die last week.