Free Split Test Calculator

Analysis Settings

Analysis Methods Explained

Frequentist (Z-Test)

The traditional statistical method that answers: "If there were no real difference between A and B, what's the chance we'd see these results by luck?"

  • Uses p-values and confidence levels (95% confidence is standard)
  • Tells you if results are "statistically significant"
  • More reliable with larger sample sizes (1000+ visitors)
  • Familiar to most analysts and stakeholders
Bayesian Analysis

A more intuitive approach that directly answers: "What's the probability that B is actually better than A?"

  • Shows direct probability of B being better
  • Works well even with smaller sample sizes
  • Takes into account uncertainty in the data
  • Provides continuous updates as data comes in
Both Methods

Using both methods gives you the most complete picture:

  • See if results are significant (Frequentist)
  • Understand the probability of improvement (Bayesian)
  • Make more confident decisions
  • Better understand the risks involved
Which Should You Choose?
  • Frequentist: When you need a clear yes/no decision and have large sample sizes
  • Bayesian: When you want probability-based decisions or have smaller samples
  • Both: When you want the most comprehensive analysis (recommended)

Revenue Data Analysis

Why Include Revenue?

Revenue data provides deeper insights into the true business impact of your test variants beyond just conversion rates.

  • A variant with lower conversions might actually generate more total revenue if its customers make larger purchases
  • Helps identify which variant is more profitable, not just which converts better
  • Shows revenue per visitor for each variant, a key business metric
How Revenue is Analyzed

The calculator performs two separate statistical tests:

  • Conversion Rate Test: Uses a Z-test to determine if the difference in conversion rates is statistically significant
  • Revenue Test: Uses a t-test to determine if the difference in revenue per visitor is statistically significant

Each test provides its own confidence level, helping you understand:

  • How confident you can be about the conversion rate difference
  • How confident you can be about the revenue difference
  • Whether the results align or conflict between metrics
Example Scenario

Consider two variants:

  • Variant A: 10% conversion rate, $50 per conversion = $5.00 per visitor
  • Variant B: 8% conversion rate, $75 per conversion = $6.00 per visitor
  • While Variant A has better conversions, Variant B generates 20% more revenue

The calculator will show:

  • Statistical confidence for the conversion rate difference
  • Statistical confidence for the revenue per visitor difference
  • Whether these differences are significant enough to make a decision
When to Use
  • When testing pricing strategies or purchase flows
  • If different variants might attract different customer segments
  • When optimizing for revenue is as important as conversion rate
  • To understand the full economic impact of your test
Variant A
Variant B
Statistical Factors
Sample Size Ratio (A:B)
-

Sample Size Ratio

This shows how evenly your visitors are split between variants A and B.

  • A ratio of 1.00 means you have equal visitors in both groups (ideal)
  • Numbers far from 1.00 mean uneven split, which can affect reliability
  • Try to keep this number between 0.8 and 1.2 for best results
Minimum Detectable Effect (MDE)
-

Minimum Detectable Effect

The smallest improvement you can reliably detect with your current sample size.

  • Smaller numbers are better - they mean you can detect subtle differences
  • Large numbers mean you can only detect big changes
  • To decrease MDE, increase your sample size
Power Analysis
-

Power Analysis

Shows how likely you are to detect a real difference if one exists.

  • Higher is better - aim for at least 80%
  • Low power means you might miss real improvements
  • To increase power, get more visitors or look for bigger improvements
Frequentist Analysis Results
Conversion Rate A: -
Conversion Rate B: -
Relative Improvement: -
Statistical Confidence: -
Bayesian Analysis Results
Probability B is Better: -
Expected Loss: -
Suggestions or Questions? Contact Us
ยฉ 2024 Miles Baker. All rights reserved.