# How to know when to turn off an ad

> Clear rules for when to turn off a Meta ad: how much to spend before judging, which numbers to read and when a losing ad deserves more time.
- **Author**: Marxx
- **Published**: 2026-10-21
- **Category**: Testing
- **URL**: https://marxx.ai/library-guides/how-to-know-when-to-kill-an-ad

---

Turn off an ad when it has spent two to three times your target cost per result and its cost per result is still well above target. Judging earlier wastes good ads. Judging later wastes money.

## Rule 1: spend enough before you judge

Set a test budget per ad of about 2 to 3 x your target CPA.

Example: your target cost per purchase is INR 600. Give each new ad INR 1,200 to INR 1,800 before deciding. An ad with zero sales at INR 400 spent has not had a fair test.

## Rule 2: read the numbers in order

| Check | If weak, the problem is |
|---|---|
| Hook rate (video) or CTR (static) | The opening |
| Link CTR | The message or offer |
| Landing page views vs link clicks | Load speed or a broken link |
| Conversion rate | The page or the price |
| Cost per result | The overall ad |

An ad with a strong hook and weak conversion may need a better landing page, not a new ad.

## Rule 3: three outcomes

After the test budget:

- **Cost per result at or under target:** keep it and give it more budget.
- **Within 30% over target with strong early metrics:** give it one more test budget.
- **More than 30% over target, or weak hook and CTR:** turn it off.

## Rule 4: do not judge in the learning phase alone

Meta says ad sets are less stable, and usually cost more per result, while they are in the learning phase. That phase usually ends after about 50 results in the week after the last significant edit. Do not judge on the first days alone, unless the ad is clearly broken.

## A worked example

Target CPA INR 800. A new UGC ad spends INR 2,000 with 1 sale (INR 2,000 CPA), a 34% hook rate and a 1.9% link CTR. Strong opening, strong clicks, weak sales. Before killing it, check the landing page it points to. It turns out the page shows a sold-out size. Fixing that saves the ad.

## FAQ

### Should I kill ads with high CPA but good ROAS?

Judge by the number tied to profit. If high-value orders make ROAS good, the ad may be fine despite a high CPA.

### How many ads end up winning?

No one can promise you a number. Judge each ad by its own cost per result, and keep a steady flow of new tests.

---
- [More Testing articles](https://marxx.ai/library-guides/category/testing)
- [All articles](https://marxx.ai/library-guides)