A/B Testing Screen Resolution Monthly Formula
Learn how eligible traffic, sample size, and estimated test duration are calculated for a screen-resolution A/B test.
This calculation estimates whether visitors using one screen resolution can provide enough traffic for a two-variation A/B test. It first filters monthly website traffic to the selected resolution, then estimates the sample required to detect a chosen absolute conversion-rate improvement.
- 100% Free
- No Sign-Up Required
- Private & Secure
- Mobile Friendly
Estimated Test Duration in Months
Where:
Calculate the visitors needed in each test group, double that amount for control and variant, and divide it by the monthly visitors using the selected screen resolution.
Variables Explained
| Variable | What It Means | Unit |
|---|---|---|
| monthlyVisitors - Total monthly website visitors | Average number of all website visitors in one month before screen-resolution filtering. | visitors |
| resolutionShare - Screen resolution traffic share | Percentage of monthly visitors who use the screen resolution being tested. | percent |
| eligibleMonthlyVisitors - Eligible monthly visitors | Estimated monthly traffic available to the experiment after filtering to the selected resolution. | visitors |
| baselineProportion - Baseline conversion proportion | Current conversion rate for the selected resolution expressed as a decimal. | number |
| variantProportion - Variant conversion proportion | Expected conversion rate after adding the minimum detectable effect to the baseline rate. | number |
| conversionDifference - Conversion difference | Absolute conversion-rate change to detect, expressed as a decimal. | number |
| confidenceZ - Confidence z-score | Z-score selected for the confidence level, such as 1.96 for 95% confidence. | number |
| powerZ - Power z-score | Z-score selected for statistical power, such as 0.84 for 80% power. | number |
| pooledProportion - Pooled conversion proportion | Average of the expected control and variant conversion proportions. | number |
Step-by-Step Calculation
Estimate resolution-specific traffic
Filter total monthly traffic to visitors using the screen resolution included in the experiment.
eligibleMonthlyVisitors = monthlyVisitors * (resolutionShare / 100)
Convert the baseline rate to a decimal
A conversion rate entered as a percentage must be expressed as a proportion for the sample-size calculation.
baselineProportion = baselineConversionRate / 100
Set the expected variant rate
The variant rate equals the baseline rate plus the smallest absolute improvement the test should detect.
variantProportion = (baselineConversionRate + minimumDetectableEffect) / 100
Calculate the pooled proportion
The pooled proportion is the average expected conversion rate across the two groups.
pooledProportion = (baselineProportion + variantProportion) / 2
Calculate visitors required per variation
This two-proportion sample-size formula estimates the eligible visitors needed in both the control and variant groups.
sampleSizePerVariation = ceil(pow((confidenceZ * sqrt(2 * pooledProportion * (1 - pooledProportion))) + (powerZ * sqrt((baselineProportion * (1 - baselineProportion)) + (variantProportion * (1 - variantProportion)))), 2) / pow(conversionDifference, 2))
Estimate total traffic and duration
Double the per-variation sample for an even A/B split, then divide by eligible monthly traffic.
estimatedTestDurationMonths = (sampleSizePerVariation * 2) / eligibleMonthlyVisitors
Example: Testing a layout on a screen resolution with 25% traffic share
Eligible monthly visitors
100000 * (25 / 100)
25,000 visitors
Baseline proportion
5 / 100
0.05
Variant proportion
(5 + 1) / 100
0.06
Visitors per variation
ceil(pow((1.96 * sqrt(2 * 0.055 * 0.945)) + (0.84 * sqrt((0.05 * 0.95) + (0.06 * 0.94))), 2) / pow(0.01, 2))
8,149 visitors
Estimated duration
(8149 * 2) / 25000
0.65 months
Final Result
About 8,149 eligible visitors are needed in each variation, or 16,298 in total. At 25,000 eligible visitors per month, the estimated duration is about 0.7 months.
Assumptions
- ✓Visitors are assigned evenly between the control and variant, with 50% in each group.
- ✓Monthly traffic and the selected resolution's traffic share remain broadly stable during the test.
- ✓The calculation compares two conversion rates using a two-sided test.
- ✓The minimum detectable effect is an absolute change in percentage points, not a relative percentage lift.
- ✓Each eligible visitor contributes an independent observation.
Limitations
- !Actual traffic can vary by month, campaign, device mix, geography, and season.
- !The calculation does not adjust for multiple comparisons, sequential monitoring, or several metrics.
- !Resolution reporting may group users differently across analytics tools or change as device behavior changes.
- !The estimate addresses sample quantity, not implementation quality, tracking accuracy, or experiment validity.
Common Mistakes to Avoid
Entering a relative lift as the minimum detectable effect; a move from 5% to 6% is 1 percentage point, not 20 percentage points.
Using total site conversion rate when the tested resolution has a different baseline conversion rate.
Forgetting that only resolution-specific visitors are eligible for the calculation.
Treating monthly traffic as equally available every day despite uneven traffic patterns.
Stopping when an early result looks favorable instead of collecting the planned sample.
Related Formulas
Frequently Asked Questions
How is A/B test sample size calculated for a screen resolution?
The calculation uses the baseline rate, absolute effect to detect, confidence z-score, power z-score, and a two-proportion comparison formula. It then applies the result only to visitors using the chosen resolution.
Why does a smaller minimum detectable effect require more traffic?
Smaller conversion differences are harder to distinguish from normal random variation, so the formula requires a larger sample.
Is the minimum detectable effect relative or absolute?
It is absolute percentage points. For example, a change from 5% to 6% is a 1 percentage-point effect.
What does 95% confidence mean in this calculator?
It uses a z-score of 1.96 in the planning formula to set a stricter evidence threshold than 90% confidence. It does not guarantee a result.
Why is the sample size rounded up?
A partial visitor cannot be assigned to a group, and rounding up avoids falling below the calculated planning requirement.
Ready to calculate your result?
Use the calculator to get instant results with your own inputs.