Experiment Types
EarlyCRO supports multiple experiment types to fit your testing needs:
A/B Test
Compare two or more variations against your original (control). Best for testing specific changes like headlines, CTAs, or layouts.
Split URL Test
Send traffic to entirely different URLs. Ideal for testing complete page redesigns or different landing pages.
Multivariate Test (MVT)
Test multiple elements simultaneously to find the best combination. Requires more traffic but provides deeper insights.
Setting Up Goals
Goals define what success looks like for your experiment. EarlyCRO tracks conversions against your goals automatically.
Goal Types
- Page Visit: Visitor reaches a specific URL (e.g., /thank-you)
- Click: Visitor clicks an element (e.g., "Add to Cart" button)
- Form Submit: Visitor submits a form
- Custom Event: Trigger via JavaScript API
- Revenue: Track purchase values for e-commerce
Primary vs Secondary Goals
Each experiment has one primary goal used to determine the winner. You can add unlimited secondary goals for additional insights, but they won't affect winner determination.
Traffic Allocation
Control how visitors are distributed across your variations:
- Equal Split: Each variation gets the same percentage of traffic (default)
- Custom Split: Manually set percentages for each variation
- Traffic Percentage: Only include a portion of total traffic in the test
Pro Tip: Start with 50/50 split for most tests. Use lower traffic percentages (e.g., 10%) for risky experiments to limit exposure.
Targeting Rules
Show experiments only to specific segments of your audience:
- URL Matching: Exact match, contains, regex patterns
- Device Type: Desktop, tablet, mobile
- Browser: Chrome, Safari, Firefox, etc.
- Geography: Country, region, city
- New vs Returning: First-time or returning visitors
- Custom Attributes: Target by user properties you set via API
When to Stop a Test
Knowing when to stop is crucial for valid results:
Stop when you have a winner
Probability to beat control is 95%+ with sufficient sample size
Stop for no clear winner
After sufficient time/traffic with no significant difference (iterate on hypothesis)
Don't stop early!
Early peeking leads to false positives. Wait for statistical significance.