Enter your daily traffic, conversion rate, number of variations, and MDE to calculate how long to run your A/B test.
Running an A/B test for the right length of time is one of the most consequential and most overlooked steps in the process. Stop too early and you risk acting on noise — a variant that looks like a winner on day three may revert to average by day fourteen as your sample stabilises and the novelty effect fades. Run too long and you accumulate seasonal bias, waste traffic, and delay the next experiment in your queue.
The correct test duration is not a gut feeling or a generic "two-week rule". It depends on four measurable inputs: your daily traffic, your baseline conversion rate, the size of the improvement you want to detect, and the confidence level you require. This free A/B test duration calculator solves for the number of days using those four inputs so you can plan before you launch — not guess after.
The duration estimate is the required total visitor count divided by your daily traffic. The required total is derived from your baseline conversion rate, your MDE, and your confidence level. The core formula is:
Days = Total Required Visitors ÷ Daily Traffic
Total Visitors = Variations × (4 × σ / effect_size)²
σ = √(CR × (1 − CR)) effect_size = CR × (MDE / 100)
Enter your numbers above and the calculator applies this formula instantly. The result tells you how many days to run your test to detect the effect you specified at 95% statistical confidence.
Your baseline CR determines the natural variance in your data. A page converting at 2% has much higher variance per visitor than one at 20%, because each conversion is rarer. Higher variance means you need more observations to separate a real signal from random noise — which extends required test duration significantly. Low-CR pages almost always need longer tests than teams expect.
MDE is the smallest relative lift you want to reliably detect. Set it to 5% and the test is designed to catch a 5% relative improvement in conversion rate. The catch: smaller MDEs require exponentially more data. Chasing a 2% relative lift needs roughly 25× more visitors than detecting a 10% lift. Set your MDE to the smallest improvement that is genuinely worth shipping — not the smallest that is theoretically possible.
This calculator targets 95% statistical significance (5% false-positive rate) and 80% statistical power (20% false-negative rate) — the industry standard for A/B testing. Raising the threshold to 99% confidence will increase your required duration considerably. For most marketing-page experiments, 95% is the right balance between rigour and speed. Never lower the threshold retroactively once you have seen results.
Traffic is the multiplier on everything else. A test that needs 10,000 visitors takes 10 days at 1,000 visitors per day and 100 days at 100 visitors per day. If your daily traffic is low, you have two practical options: set a larger MDE (detect only bigger effects), or accept a lower confidence threshold. Never inflate traffic by including visitors who did not actually qualify for the test.
The most expensive mistake is stopping a test early because interim results look good. This is called peeking and it dramatically inflates your false-positive rate — a result that looks significant on day three is often noise. Commit to the estimated duration before the test goes live.
Setting an MDE that is too small for your traffic level is equally common. If you have 500 daily visitors and a 3% base CR, a 2% MDE would require more than a year of testing. Be realistic about what your traffic can support within a two-to-four week window.
Running tests across promotions, holiday periods, or major traffic spikes without flagging those windows introduces audience bias that the duration formula cannot account for. The estimate assumes stable, representative traffic — if your audience changes mid-test, the results are unreliable regardless of duration.
Duration is one part of experiment planning. Use these tools alongside this calculator for a complete testing workflow:
Running a test for the right length of time is critical. A calculator gives you a data-driven estimate based on your traffic volume, current conversion rate, and the size of improvement you want to detect — so you stop testing at the right moment.
Stopping early is the single most common A/B testing mistake. Early results fluctuate heavily due to small sample sizes, so a variant that looks like a winner at day 3 may be completely average by day 14. Always run the test to the duration the calculator recommends.
Yes, but there are diminishing returns. Running much longer than needed wastes time and can introduce novelty effects or seasonal bias. If your test has not reached significance after 2-3x the recommended duration, the effect is likely too small to matter or the test needs to be redesigned.
MDE is the smallest percentage improvement you want your test to be able to detect. A 10% MDE means: if the real effect is at least 10% relative to your baseline CR, the test is designed to catch it. Set a realistic MDE — chasing tiny lifts requires enormous traffic.
Mida is 10X faster than anything you have ever considered. Try it yourself.
Mida.so is a super lightweight A/B testing tool to help you experiment, analyze and implement conversion strategies in minutes.
