# How Long Should You Run an A/B Test

> Stop too early and you crown false winners; run too long and you waste traffic. Learn how to set the right A/B test duration before you launch.

- Source: https://bixlow.com/en/blog/how-long-to-run-an-ab-test
- Language: en
- Published: 2026-07-13T09:00:00.000Z
- Updated: 2026-10-06T09:28:18.400Z
- Author: fernando
- Topics: A/B testing, sample size, statistical significance, CRO, conversion testing
- Publisher: Bixlow

---
One of the most common questions when launching an experiment is when to stop it. Stop too early and you get false winners; run it too long and you waste traffic and delay decisions. Getting the duration right is as important as designing the variant well, because it directly shapes how reliable your conclusion will be.

The right answer isn't a fixed number of days. It's the result of combining the sample size you need, the natural cycles of your business, and the behavior of your traffic. In this article you'll see how to estimate duration before launching and which signals tell you it's not yet time to stop the test.

## Duration depends on sample size, not the calendar

A test's runtime is a consequence of the sample size you need and the traffic you receive. If your goal requires 20,000 visitors per variant and you get 2,000 a day, the test will run at least ten days. That's why duration is calculated before you start, by estimating the effect you expect to detect and the traffic volume available. Launching without that math is flying blind.

Sample size calculators exist that, based on your current conversion rate and the minimum improvement you want to detect, tell you how many users you need. With that number and your daily traffic you get a realistic duration estimate, which you should lock in as a commitment before you look at any results.

## Respect your full business cycles

User behavior changes depending on the day of the week and the time of the month. Someone who buys on a Monday isn't always like someone who buys on a Saturday, and a test that only covers weekdays can mislead you. That's why it pays to let the experiment run across complete cycles.

- Cover at least one or two full weeks to capture the variation between weekdays and weekends.
- If your business has monthly cycles, such as payments or payroll, consider covering a full month.
- Avoid starting or stopping tests on atypical dates like sales, campaigns, or holidays that distort the data.

## Signs you shouldn't stop yet

Don't stop the test just because one version is winning: until you reach the planned sample size, that lead can be noise. Don't trust an early spike in significance either. And if your test hasn't covered at least one full week, wait even if the numbers look clear. Patience at this stage is what protects the validity of the entire investment you've made.

## Key takeaways

- Calculate duration from sample size and your daily traffic before launching.
- Let the test run across full cycles of at least one or two weeks.
- Don't stop for an early lead or a premature spike in significance.

---

## Meet our AI CRO agent

What you just read, put to work by an agent: it understands your visitors, decides what to test or personalize, builds it on your site without a developer and tells you when a result is real.

https://bixlow.com/en/ai-cro-agent