Marketing

A/BTestingCalculator

Determine the winning variation of your A/B tests with statistical significance calculator. Calculate lift, Z-score, and confidence levels.

A/B Significance Engine

Determine winning variations with 95% confidence

Statistical Z-Test

Variation A

Control Group
Conversion Rate
5.00%

Variation B

Test Group
Conversion Rate
7.14%
Measured Lift
+42.9%

Variation B outperformed Variation A by 42.9%.

2%Confidence
Trending Results

Results are not significant yet.

There is currently a 2.1% probability that Variation B is better. We recommend continuing the test until reaching 95% confidence.

Z-Score
2.027
P-Value
0.9787

What is Confidence?

Statistical significance (Confidence) measures the likelihood that the difference between variations is not due to random chance. In A/B testing, a 95% confidence level is the industry standard for making business decisions.

When to stop testing?

You should stop a test only after achieving a significant result and reaching your pre-determined sample size. Stopping early (peeking) can lead to false positives. Ensure your variations have run for at least one full business cycle.

What is A/B Testing?

A/B testing (split testing) is a method of comparing two versions of a webpage, app, or other product experience to determine which one performs better. You split your audience into two groups - one sees version A (control) and the other sees version B (variant).

Version A

Original/Control

Version B

Variant/Test

Winner

Statistically Significant

Key Metrics Explained

Statistical Significance

Confidence that results are not due to chance. Aim for 95%+

Conversion Rate

Percentage of users who complete desired action

Lift

Percentage improvement of variant over control

Sample Size

Number of visitors needed for reliable results

P-Value

Probability results are due to chance (lower is better)

Confidence Level

1 - p-value, typically 95% or higher

Best Practices

Test One Thing

Change only one element at a time

Run Long Enough

Collect enough data before concluding

Use Real Traffic

Test with actual target audience

Avoid Peeking

Don't stop test early if results look good