# How to calculate how many ads to test

> Work out how many ads to test by taking the smallest of three numbers: what your budget can judge, what Meta's learning rule allows, and what you can make.
- **Author**: Marxx
- **Published**: 2026-11-17
- **Category**: Testing
- **URL**: https://marxx.ai/library-guides/how-to-calculate-how-many-ads-to-test

---

Test the smallest of three numbers: how many ads your budget can judge, how many your ad set can learn from, and how many truly different ads you can make this week. The formulas below are ours. The 50-result rule inside them is Meta's.

## The three caps

**Cap 1: money.** Each ad needs enough spend before you can judge it. Our [turn-off guide](/guides/how-to-know-when-to-kill-an-ad) says 2 to 3 times your target cost per result. Use 3 to be safe.

**Ads you can afford = weekly test budget / (3 x target cost per result)**

**Cap 2: learning.** Meta says an ad set usually exits the learning phase after about 50 results in the week after its last significant edit ([Meta Help Centre](https://www.facebook.com/business/help/112167992830700), checked 2026-10-20). Meta also says that when you run too many ads at once, each one delivers less often, so fewer exit learning ([Meta Help Centre](https://www.facebook.com/business/help/2720085414702598), checked 2026-10-20). So one ad set needs a weekly budget of about 50 x cost per result.

**Ad sets you can fund = weekly test budget / (50 x cost per result), rounded down**

Then allow up to 5 ads per ad set. That limit is our rule of thumb, not Meta's.

If this comes to 0, your budget is too small for that result. One of Meta's fixes for learning limited is to optimise for a more frequent event, such as add to cart ([Meta Help Centre](https://www.facebook.com/business/help/269269737396981), checked 2026-10-20). See [how to exit learning limited](/guides/how-to-exit-learning-limited).

**Cap 3: production.** Count the truly different ads you can make this week. A new angle, person or hook counts. A new button colour does not. See [creative diversity](/glossary/creative-diversity).

**Ads to test = the smallest of cap 1, cap 2 x 5, and cap 3.**

## A worked example

Say your weekly test budget is INR 30,000 and your target cost per purchase is INR 400.

- **Cap 1:** 30,000 / (3 x 400) = 30,000 / 1,200 = **25 ads**.
- **Cap 2:** 30,000 / (50 x 400) = 30,000 / 20,000 = 1.5, rounded down to 1 ad set. One ad set x 5 ads = **5 ads**.
- **Cap 3:** you can make **4** truly different ads this week.

The smallest is 4. Test four ads in one ad set. Your budget could fund more, so the limit is ideas, not money. Spend the next day on new concepts, not a bigger budget. The numbers are made up.

## After you pick the number

1. **Put the ads in one ad set** with the same audience and budget.
2. **Run the test for at least 7 days.** Meta recommends a minimum of 7 days for A/B tests, and says shorter tests can be inconclusive ([Meta Help Centre](https://www.facebook.com/business/help/290009911394576), checked 2026-10-20).
3. **Do not edit during the test.** Significant edits can reset learning.
4. **Turn off losers by your rule,** not by feel. See the [creative testing budget guide](/guides/how-to-build-a-creative-testing-budget).
5. **Write down the result,** with the angle, hook and format, so the next test starts smarter. The [ad testing hypothesis template](/templates/ad-testing-hypothesis-template) fits here.

## Mistakes to avoid

- Testing 20 ads because the budget "allows" it
- Judging an ad before it has spent 3 times your target cost per result
- Counting small tweaks as separate tests
- Testing with a budget so small that no ad set can learn

We make Marxx. It tags creative by hook, angle, format and offer, so you can see what you have already tested and what you have not. See [creative intelligence](/features/creative-intelligence).

## FAQ

### Can I test 20 ads at once?

Only if all three caps allow it. Meta warns that too many ads at once make each one deliver less often.

### What if my budget is tiny?

If cap 2 comes to 0, test on a more frequent event such as add to cart, or test one new ad at a time against your best ad.

### How long should an ad test run?

At least 7 days, per Meta's A/B test advice (checked 2026-10-20). Run longer if your buyers take more than a week to convert.

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