# A/B test duration calculator

> An A/B test's duration is the traffic it needs divided by the daily visitors entering it. Enter your visitors, your conversion rate and the lift you are after, and the calculator tells you how many days it takes, rounded to whole weeks.

- Source: https://bixlow.com/en/tools/ab-test-duration-calculator
- Language: en
- Updated: 2026-11-02
- Publisher: Bixlow

## Why whole weeks

A Monday does not buy like a Saturday. If you stop a test mid-week, one version can come out ahead just because it got more good days. That is why we always round to whole weeks and never recommend less than seven days, even when the traffic would allow it.


## What to do if it comes out too long

More than six to eight weeks is usually a bad sign: in that time campaigns, the season and even your catalogue change. You have three ways out.

- Test a bigger change, one that could move conversion by more than 20%.
- Measure a more frequent conversion: add to cart or checkout start instead of purchase.
- Drop variants. Each extra version splits the same traffic into more groups and, with more comparisons against the original, each group also needs a bigger sample.


> **Example.** A product page with 1,000 visits a day and a 5% conversion rate wants to detect a 20% lift. It needs about 16,300 visitors across both versions: 17 days of traffic, which becomes 3 weeks.


## How Bixlow does it

In Bixlow every test has a minimum sample and a minimum duration. Even if a variant is ahead on day one, no winner is named until both are met and the probability of being best reaches 95%. Until then, the panel tells you what is missing.


## FAQ

### Can I stop the test early if there is already a clear winner?

Better not. Results swing a lot in the first days, and what looks like a clear winner often fades. Wait for the duration you worked out.

### What do I enter as traffic in the test?

The share of that page's visitors you want in the test. Usually 100%; lower it if the change is risky and you would rather expose fewer people, knowing the test will take longer.

### What if my traffic swings a lot with campaigns?

Use the average of a normal month, without peaks. If you are launching a big campaign mid-test, it is better to wait until it ends before you start.

## Further reading

- [Sample size calculator](https://bixlow.com/en/tools/sample-size-calculator)
- [What is statistical significance](https://bixlow.com/en/glossary/statistical-significance)
- [What is Bayesian A/B testing](https://bixlow.com/en/glossary/bayesian-ab-testing)
- [A/B testing in Bixlow](https://bixlow.com/en/platform/ab-testing)