# Google tells advertisers 80% confidence is enough for small Google Ads tests

> Google's Ads and Commerce Blog repeated on 19 November 2026 an 80% confidence rule for small tests and 95% for big ones. What it means for your tests.
- **Author**: Marxx
- **Published**: 2026-11-20
- **Category**: Google
- **URL**: https://marxx.ai/library-news/google-ads-decoded-ab-tests-80-percent-confidence-november-19

---

Google says an **80% confidence level** is enough for small changes in Google Ads experiments, and 95% for big structural ones. Google's Ads & Commerce Blog repeated the rule on **19 November 2026**, after an Ads Decoded podcast episode on 18 November. [PPC Land summarised both](https://ppc.land/google-says-80-confidence-is-good-enough-for-directional-ad-tests/).

## What we know

The episode is called "A/B testing in Google Ads: How to prove what works". Google's own experts explained how campaign tests split traffic and how significance is worked out.

| Test type | Confidence Google suggests | Run time Google suggests |
|---|---|---|
| Small, directional change | 80% | 4 to 6 weeks |
| Major structural change | 95% | 4 to 6 weeks |

- **Reported by PPC Land from the episode:** budget tests run each arm on half the budget, and a shared budget test is a pilot that opened this month.
- **Also reported:** Google's help pages do not fully agree with the episode on test length and confidence levels.
- **Not stated:** which accounts have the shared budget pilot.

## What it means for your ad account

A confidence level is how sure you want to be that a result is not luck. At 80%, you accept about a one in five chance of being fooled. At 95%, one in twenty.

Say you test [AI Max for Search](/glossary/ai-max-for-search) on a INR 2,00,000 monthly budget. Google shows 83% confidence that conversions rose. Under the 80% rule that is "directional", so you can keep going. Under 95% it is not proven. Moving all your budget on 83% is a bet, so scale in steps.

Remember the tests are run by Google, inside Google's auction, on features Google sells.

## What to do this week

1. **Sort your planned changes** into small and structural before you start a test.
2. **Set the bar before you look.** Write 80% or 95% in your test notes first.
3. **Run the test 4 to 6 weeks,** as Google advises, unless the data is clearly flat.
4. **Scale in steps** after a directional win, not all at once. See [how many ads to test](/guides/how-to-calculate-how-many-ads-to-test).
5. **Keep your own record** of results outside Google Ads, so you can check them later.

## FAQ

### Is 80% confidence safe?

It is a looser bar. It suits small, easy to undo changes, not big budget moves.

### Did Google change how Experiments work?

Not in this report. It explains the existing method and a shared budget pilot.

### Does this apply to Meta tests?

No. It is Google's guidance for Google Ads experiments.

---
- [More Google articles](https://marxx.ai/library-news/category/google)
- [All articles](https://marxx.ai/library-news)