The A/B Test Pre-Planning Calculator determines the required sample size and estimated test duration before running an experiment. It estimates the sample for detecting an effect of the planned size at the chosen power; a significant result is not guaranteed. The formula is an approximation based on the baseline rate, target rate, significance level and power. Common planning choices are 80% power and 95% confidence. PM Toolkit's free A/B test planning calculator helps product managers plan experiments with automatic duration estimation based on daily traffic and minimum detectable effect.
What is A/B Test Planning?
A/B test planning determines the sample size and duration needed before running an experiment. Planning shows the traffic needed for the effect size you want to detect.
Sample Size Formula
Approximation: n ≈ (Z_alpha + Z_beta)^2 x 2 x p(1-p) / MDE^2
Where: Z_alpha = z-score for two-sided significance level (1.96 for alpha = 0.05), Z_beta = z-score for statistical power (0.84 for 80%), p = baseline conversion rate, MDE = absolute minimum detectable effect expressed as a proportion; n is per group at equal allocation
Test Duration Formula
Duration (days) = total required sample / total eligible daily traffic (rounded up)
Illustrative planning settings
| Parameter | Standard | Conservative |
|---|---|---|
| Confidence Level | 95% (alpha=0.05) | 99% (alpha=0.01) |
| Statistical Power | 80% | 90% |
| Example duration to review | 1 week | 2 full business cycles |
Ask Claude, Cursor, or ChatGPT to run this calculator.
Install once. 24 web calculators · 17 in your AI · 12 PM workflows
Example
You plan test baseline 3% · relative MDE 10% · confidence 95% · power 80% · two-sided
pmtk → n = 51,487 per variant
Start with
A/B Test Sample Size Calculator
Plan sample size and duration for the improvement you want to detect. Compare the required traffic with what you can recruit.
Updated
Auto-computed
%
%
Looking for: 5% → 5.75%
users/day
Quick Scenarios
Choose a preset configuration based on your needs
Not sure? These presets illustrate planning trade-offs. Choose the effect size, confidence and power for your question before launching.
Sample per group at 50/50 allocation
13,534users
- Total
- 27,068
- Duration
- 28 days
- Target
- 5.75%
Test duration feasibility28 days
60d+42d21d1d
Estimated over 3 weeks
Why this matters
Choose the smallest improvement worth detecting, then check whether your traffic can support it. Confidence and power are planning assumptions, not guarantees. Allow for traffic changes and relevant usage cycles when setting the duration.
This test might take 28 days. Consider increasing MDE to shorten the timeline.
Required Sample Size
Review durationTotal Users Needed
27,068Control Group
13,534
Variant Group
13,534
Estimated Duration
28days
Per group at 50/50: 13,534 users
n = 2 × (Zα/2 + Zβ)² × p(1−p) ÷ MDE²
= n: 13,534 per variant · p: 5% · MDE: 15%
Equal-allocation approximation. The calculator uses the Evan Miller sizing formula and adjusts group totals for allocation.
What is A/B Test Planning?
A/B test planning estimates the sample and duration needed for an experiment. Set the minimum detectable effect, power, significance level, and stopping rule before launch so a promising early result does not change the plan.
Test Duration Formula
Duration (days) = Required Sample per Variant ÷ Daily Traffic per Variant
Planning defaults
95% confidence and 80% power are common planning defaults; document your choices before launch.
Rate this calculator:
Common questions
How do I calculate sample size for an A/B test?
Enter baseline conversion, minimum detectable effect, significance level, power, and allocation. Power describes how likely the test is to detect an improvement of the size you planned for. The planner estimates the required users per variant under its statistical assumptions.
What is a good minimum detectable effect (MDE) for an A/B test?
Start with an improvement large enough to affect your decision, then check its sample requirement. Smaller effects need much more traffic: under the usual approximation, reducing the effect by a factor of ten requires roughly a hundred times the sample. There is no universal MDE for a high- or low-traffic product.
How long should an A/B test run before reaching significance?
The planner estimates the time to collect your planned sample from daily traffic. It does not predict when significance will appear. Choose a duration that covers relevant usage patterns and follow the stopping rule set before launch.
Should I plan for 80% or 90% statistical power?
At 80% power, the test has an 80% chance of detecting the planned effect if that effect exists and the assumptions hold. Raising power to 90% requires more users. Choose based on the cost of missing the improvement and available traffic; inspect the planner’s sample estimate for each setting.