Loading...
Loading...
Calculate standard deviation from raw data or conversion rates. Essential for A/B test design and sample size planning.
The Standard Deviation Calculator computes the dispersion of your dataset from raw values or a conversion rate. Standard deviation is essential for sample size calculations and understanding the variability of your metrics. If you are working with a continuous metric, paste your raw data. For binary metrics, simply enter the conversion rate.
Standard deviation quantifies how much individual data points deviate from the average. In the context of A/B testing, it represents the natural noise in your metric. A low standard deviation means your data points cluster tightly around the mean, making it easier to detect small changes. A high standard deviation means more spread, requiring larger sample sizes to achieve the same statistical power.
For conversion rate experiments, the standard deviation is determined entirely by the baseline rate using the Bernoulli formula: σ = √(p(1−p)). This reaches its maximum at p = 0.5 (50% conversion rate) and decreases as the rate moves toward 0% or 100%. This is why tests on metrics with extreme rates (like 1% error rates) need fewer samples than tests on 50/50 metrics.
Before running any A/B test, you need an estimate of your metric's standard deviation. This feeds directly into sample size calculations and determines how long your experiment needs to run. Underestimating σ leads to underpowered tests that miss real effects; overestimating it wastes time and traffic.
The coefficient of variation (CV = σ/μ) lets you compare variability across different metrics. Revenue per user typically has a CV of 100-300%, while conversion rates usually have a CV below 100%. Understanding your metric's CV helps you set realistic expectations for minimum detectable effects and experiment duration.
Explore more A/B testing and statistics tools
Calculate the minimum sample size for your A/B test.
Determine the minimum detectable effect for your experiment.
Detect sample ratio mismatch in your experiment.