Experiments let you test different variants of your page to see which performs best. By splitting traffic between variants and measuring conversions, you can make data-driven decisions about your landing pages.
Creating an experiment
- Open your page in the editor
- Navigate to the Experiments panel
- Click Create experiment
- Configure your experiment settings:
- Name: A descriptive title for the experiment
- Hypothesis: What you’re testing and expect to happen
- Variants: Which variants to include in the test
- Goals: What conversions to measure
- Delivery: When to run and how to end
Selecting variants
Choose which variants to include in your experiment. You need at least two variants—typically your control (default) and one or more challengers.
Each variant gets a traffic weight that determines what percentage of visitors see it:
| Variant | Weight | Traffic |
|---|
| Control | 50 | 50% |
| New headline | 50 | 50% |
Setting conversion goals
Choose what you want to measure. You can set:
- Primary goal: The main metric for determining a winner
- Secondary goals: Additional metrics to track (up to 3)
Goal types include:
- Form submissions
- Link clicks
- Scroll depth
- Time on page
- External conversions (tracked from other pages)
Create goals in your brand settings before adding them to experiments.
Delivery settings
Configure when and how your experiment runs:
Start time
- Immediately: Start when you publish
- Scheduled: Choose a specific start date and time
End condition
- Run until significance: Automatically complete when a winner is detected
- Run for duration: End after a specific number of days
Auto-select winner
Enable this option to automatically set the winning variant as your default when the experiment reaches statistical significance.
Starting your experiment
After configuring your experiment:
- Review all settings
- Click Start experiment
- Publish your page to make the experiment live
Traffic splitting begins immediately after publishing.
Experiment status
Experiments progress through these stages:
| Status | Description |
|---|
| Draft | Configuration in progress, not running |
| Scheduled | Start time set for the future |
| Running | Actively collecting data |
| Paused | Temporarily stopped |
| Completed | Finished with or without a winner |
Monitoring results
While your experiment runs, the overview panel shows:
- Day counter: How long the experiment has been running
- Sessions: Total sessions per variant
- Conversions: Goal completions per variant
- Conversion rate: Percentage of visitors who converted
- Probability to be best: Statistical confidence each variant is the winner
Results update as new data comes in.
Statistical significance
Blox uses Bayesian statistics to calculate the probability each variant is the best performer. When one variant reaches 95% probability, it’s considered statistically significant.
For reliable results, experiments need:
- At least 100 visitors per variant
- A minimum of 14 days running (for auto-completion)
Pausing and resuming
You can pause a running experiment:
- Click Pause in the experiment controls
- Traffic stops splitting—all visitors see the control
- Click Resume to continue the experiment
Pausing creates a gap in your data, which is noted in integrity warnings. Try to avoid pausing unless necessary.
Completing an experiment
Experiments end when:
- Manual completion: You click “Complete” and optionally select a winner
- Auto-completion: Statistical significance is reached (if enabled)
- Duration ends: The scheduled end date arrives
When completed, you can choose to set the winning variant as your new default.
One experiment at a time
Each page can only run one experiment at a time. This ensures clean data and prevents interactions between tests. Complete or stop the current experiment before starting a new one. 
