Statistical Inference

Metric Type

Binary (conversion rate, success/failure) Continuous (revenue, time, score)

Optional Settings

Test Direction Confidence Level (%) Minimum Meaningful Lift (optional)

Data Relationship ⓘ Independent
Different users or entities are used in control and test.

Example:
One group of users sees the old feature, a different group sees the new feature.

Paired
The same users or entities are measured twice — before and after a change.

Example:
Measure the same users before a feature launch and again after the launch.

Independent Paired

Paired Conversion Changes

Why these inputs changed

You selected Binary and Paired. This means we are measuring the same users before and after a change.

In this situation, comparing overall conversion rates is not statistically valid. Instead, the analysis focuses only on users whose conversion status changed.

This method is called McNemar’s test. It is the standard and statistically correct approach for paired conversion data.

Converted before, but not after

Did not convert before, but converted after

Users whose conversion status did not change do not need to be entered.

Results

No analysis run yet.

Test used
Test statistic
P-value
Alpha (α)
Statistically significant?
Confidence interval
Number of comparisons

Statistical Inference

Metric Type

Optional Settings

Paired Conversion Changes

Control Group

Test Groups

Results

📊 Business Impact

🔍 Statistical Confidence

👉 Recommended Action

📐 Statistical Details