A/B Testing Memory Requirement Formula
Learn how an A/B testing memory requirement calculator estimates sample size, expected conversions, and test duration.
In A/B testing, “memory requirement” means the visitor data needed to detect a planned conversion-rate improvement, not computer storage. The estimate helps teams check whether their available traffic can support a meaningful experiment before launching it.
- 100% Free
- No Sign-Up Required
- Private & Secure
- Mobile Friendly
Visitors Needed per Version
Where:
The formula estimates the number of visitors each version needs by comparing the baseline conversion rate with the smallest conversion-rate uplift you want to detect. Higher confidence, higher power, rarer conversions, and smaller target changes all increase the required sample.
Variables Explained
| Variable | What It Means | Unit |
|---|---|---|
| n - Sample size per version | Estimated number of eligible visitors required in each equally allocated test version. | visitors |
| baselineRate - Baseline conversion rate | The control conversion rate expressed as a decimal rather than a percentage. | percent |
| minimumDetectableEffect - Minimum detectable uplift | The smallest relative improvement the test is designed to detect. | percent |
| effectSize - Absolute detectable effect | The absolute conversion-rate difference implied by the selected relative uplift. | percent |
| variantRate - Expected variant conversion rate | The conversion rate expected if the chosen uplift is present. | percent |
| averageRate - Average conversion rate | The midpoint of the baseline and expected variant conversion rates. | percent |
| confidenceLevel - Confidence critical value | The selected critical value used to control false-positive risk, such as 1.96 for 95% confidence. | number |
| statisticalPower - Power critical value | The selected value representing the desired probability of detecting the planned effect when it exists. | number |
| numberOfVariants - Number of versions | The total number of equally allocated versions, including the control. | number |
| dailyEligibleVisitors - Daily eligible visitors | Average daily visitors who can be included in the experiment. | visitors per day |
Step-by-Step Calculation
Convert the baseline percentage to a decimal
A baseline such as 5% becomes 0.05 for use in the statistical calculation.
baselineRate = baselineConversionRate / 100
Calculate the absolute effect size
A relative uplift is translated into a percentage-point difference. For example, a 10% uplift on a 5% baseline equals a 0.5 percentage-point change.
effectSize = baselineRate * (minimumDetectableEffect / 100)
Estimate the expected variant rate
This is the conversion rate that would result if the selected uplift occurred.
variantRate = baselineRate + effectSize
Calculate the pooled average rate
The two-proportion sample-size approximation uses the average of the expected control and variant rates.
averageRate = (baselineRate + variantRate) / 2
Estimate visitors needed per version
The formula combines the selected confidence level, power, expected rates, and effect size, then rounds up to a whole visitor.
sampleSizePerVersion = ceil(pow((confidenceLevel * sqrt(2 * averageRate * (1 - averageRate))) + (statisticalPower * sqrt((baselineRate * (1 - baselineRate)) + (variantRate * (1 - variantRate)))), 2) / pow(effectSize, 2))
Calculate total traffic and duration
Each version receives its planned sample. Total traffic is then divided by daily eligible traffic and rounded up to full days.
totalSampleSize = sampleSizePerVersion * numberOfVariants; estimatedTestDays = ceil(totalSampleSize / dailyEligibleVisitors)
Example: 5% baseline rate and 10% relative uplift
Convert the baseline rate
5 / 100
0.05
Find the absolute detectable effect
0.05 × (10 / 100)
0.005
Calculate the expected variant rate
0.05 + 0.005
0.055 or 5.5%
Calculate visitors per version
ceil([(1.96 × sqrt(2 × 0.0525 × 0.9475)) + (0.842 × sqrt((0.05 × 0.95) + (0.055 × 0.945)))]^2 / 0.005^2)
31,243 visitors
Calculate total visitors
31,243 × 2
62,486 visitors
Estimate test duration
ceil(62,486 / 1,000)
63 days
Final Result
Plan for about 31,243 visitors per version, 62,486 visitors overall, roughly 3,281 expected conversions, and an estimated duration of 63 days.
Assumptions
- ✓Traffic is allocated evenly across the control and every variation.
- ✓The baseline conversion rate remains broadly stable while the experiment is running.
- ✓The planned uplift is a relative change from the control conversion rate.
- ✓The calculation uses a standard two-proportion sample-size approximation.
- ✓Visitors and conversion opportunities are treated as sufficiently independent for planning purposes.
- ✓The test has a predefined primary conversion metric.
Limitations
- !This is a planning estimate rather than a guarantee that a result will be valid or useful.
- !The estimate does not automatically adjust for multiple comparisons, many variants, or many outcome metrics.
- !Seasonality, traffic-source changes, product releases, and tracking changes can affect the observed conversion rate.
- !Unequal traffic allocation can increase the total sample requirement compared with equal allocation.
- !The formula does not determine whether the detected change has enough business value to act on.
Common Mistakes to Avoid
Entering a percentage-point change as though it were a relative uplift. A move from 5% to 5.5% is a 10% relative uplift, not a 0.5% uplift.
Using all site visitors rather than only visitors who are actually eligible to enter the test.
Forgetting to count the control in the total number of versions.
Choosing a very small detectable uplift without checking whether available traffic can support the resulting duration.
Stopping when an early result looks favorable instead of using a predefined sample and stopping approach.
Changing the primary metric or audience definition after the test has started.
Related Formulas
Frequently Asked Questions
How is A/B testing sample size calculated?
This calculator converts the target relative uplift into an absolute conversion-rate difference and applies a two-proportion sample-size approximation using the selected confidence level and statistical power.
Why does a smaller minimum detectable effect need more traffic?
Smaller differences are harder to separate from normal random variation, so more visitors are required to detect them reliably.
What does 95% confidence mean in this calculator?
It uses a critical value of 1.96 in the planning formula to reduce false-positive risk. It does not mean there is a 95% chance that one version is better.
What is the difference between confidence level and statistical power?
Confidence level relates to controlling false positives, while power is the planned chance of detecting the selected uplift when it is genuinely present.
Does the number of variants change the sample size per version?
Not in this equal-allocation calculation. Each version keeps the same estimated per-version requirement, while total traffic increases with every additional version.
Why are expected conversions shown?
Expected conversions provide a practical sense of how much outcome data the planned sample may contain. They are estimated from total sample size and the average expected conversion rate.
Ready to calculate your result?
Use the calculator to get instant results with your own inputs.