The live calculator headline for this example is +50.00% relative lift.
NIRMION UTILITY SPACE
Calculateur de test
Créez un calculateur de test ab limité localement avec des hypothèses visibles, des exemples pratiques et un résultat révisible.
Le navigateur prêtRace localementRecommandation humaine
Révision du projet
Votre résultat est en attente.
Ajouter le contexte requis et une entrée par ligne, puis générer une sortie révisible.
- Définir le champ.
- Nom du propriétaire.
- Vérifiez la production avant de partager.
AB TEST EXPLAINER
Read the lift before you debate the winner.
This tool is built for a simple two-variant comparison. You paste traffic and conversion counts for the control and variant, and the page turns those four numbers into two observed rates, a percentage-point gap, and relative lift.
Worked example from the default input
With 100 control visitors and 100 variant visitors, the example records 10 control conversions and 15 variant conversions. The tool therefore reports 10.00% for control, 15.00% for the variant, a 5.00 percentage-point gap, and +50.00% relative lift.
The parser expects one labeled row per line. Keeping the labels exactly as shown is the safest way to match the calculator's current input contract.
Control visitors | 100
Control conversions | 10
Variant visitors | 100
Variant conversions | 15What to enter and how to enter it
This workspace does not use separate number boxes for each metric. The large input area is the real calculator definition, so the labels are part of the workflow.
Control visitors
Enter the total people, sessions, or units that saw the control experience during the same measurement window as the variant.
Control conversions
Enter only the conversions that belong to the same success event you will count for the variant.
Variant visitors
Enter the total people, sessions, or units that saw the variant experience under the same routing and time window.
Variant conversions
Enter the conversions for the variant using the same definition, attribution rule, and counting logic as the control.
How to interpret the result
Start with the two observed rates, then read the absolute gap, then the lift. That order keeps a relative percentage from hiding the size of the underlying rate change.
Control conversion rate
10.00%The control rate is 10 divided by 100, which is 10.00%.
Variant conversion rate
15.00%The variant rate is 15 divided by 100, which is 15.00%.
Absolute rate difference
5.00%The direct gap between the two observed rates is 5.00 percentage points.
Relative lift
+50.00%Relative lift compares the rate gap to the control baseline, so this example becomes +50.00%.
Method, assumptions, and limits
The page computes two proportions from the counts you supply. It then shows both the raw rate gap and the gap relative to the control rate so you can separate practical change from percentage framing.
Control rate = 10 / 100 = 10.00%Variant rate = 15 / 100 = 15.00%Absolute difference = 15.00% - 10.00% = 5.00 percentage pointsRelative lift = (15.00% - 10.00%) / 10.00% = +50.00%
Safe-use boundary
Use this readout only when both variants measure the same success event, use comparable traffic windows, and keep conversions less than or equal to visitors. The tool does not randomize traffic, deduplicate users, estimate uncertainty, or decide whether the experiment should stop.
Source behind the statistical boundary
For inference on differences between two proportions, use a formal method rather than relying on lift alone. The linked NIST reference covers the next step beyond this descriptive calculator.
NIST reference on difference of proportionsRelated tools that complete the workflow
These related pages help when you need traffic planning, a single-rate sanity check, or an actual significance decision after the counts are in.
Calculateur de taux de conversion
Check one observed rate in isolation before you compare it against another variant.
Calculateur de taille
Estimate how much traffic you need before the lift number is worth acting on.
Calculateur de durée A/B
Translate target traffic and cadence into a realistic A/B test run time.
A/B Test Signification Aideur
Test whether the observed gap is likely noise or evidence of a real difference.
Default example at a glance
The bars below show observed conversion rate, not raw conversion count. Equal traffic makes the five-point rate gap easier to read.
Because both variants have 100 visitors in the default example, the visual highlights the rate change directly: 10.00% versus 15.00%.
COMMON QUESTIONS
Questions worth answering before you trust the lift
The calculator is fast because it stops at descriptive comparison. These answers explain what that means in practice.
What is the difference between percentage-point change and relative lift?
Percentage-point change is the direct gap between the two rates. In this example, 15.00% minus 10.00% equals 5.00 points. Relative lift divides that same gap by the control rate, which is why the same result is +50.00%.
Does this calculator tell me whether an A/B test is statistically significant?
No. It reports observed rates, the absolute percentage-point gap, and relative lift only. Significance, confidence intervals, power, and stopping rules need a separate statistical method.
What if one variant has tiny traffic or zero conversions?
Treat the readout as fragile. Very small samples make rates unstable, and a zero control rate makes relative lift unhelpful in practice even though the page can still show a descriptive output. Use a sample-size or significance tool before making a decision.
Comment l’utiliser
Faites du résultat utile et responsable
- 011. Définir la question et le contexte.
- 022. Ajouter uniquement les informations nécessaires à ce résultat.
- 033. Vérifiez les hypothèses, les autorisations et les unités avant d’utiliser.
This workspace reads four labeled rows from the input box, converts them into control and variant conversion rates, then reports the absolute percentage-point gap and relative lift. It is descriptive only and does not estimate significance, confidence intervals, or sample size.