Experiment Lifecycle
How experiments progress from creation to winner promotion
Every experiment follows a four-stage lifecycle. Understanding each stage helps you get the most out of your tests.
Stages
Draft
Your experiment is created and configured but not yet live. Set your two variants, choose your termination rule, and select your primary goal KPI. No traffic is split yet — you can freely edit all settings.
Running
The experiment is live. Carts are being assigned to variants and data is accumulating. You can monitor results in real time, but avoid drawing conclusions or ending the experiment too early.
Completed
The experiment has ended — either manually, by reaching its time limit, or by hitting its volume target. Results are finalized and ready to analyze.
Winner Promoted
You've reviewed the results and promoted the winning variant. Its full configuration is now applied as your store's default prevention settings.
Termination Rules
When setting up an experiment, choose one of three ways to end it:
| Rule | How It Works |
|---|---|
| Manual | The experiment runs indefinitely until you or your account manager stops it |
| Time-based | Set a duration between 1 and 90 days |
| Volume-based | Set a total cart sample size (e.g., stop after 1,000 carts across both variants) |
Promoting a Winner
Once your experiment is complete, review the results for both variants and promote the winner. When you promote a variant, its entire configuration is applied as your store's new default prevention settings:
- Incentive type and amount
- Minimum purchase threshold
- Discount stacking rules
- Product and collection exclusions
Promoting a winner permanently changes your store's prevention settings. If you want to revert, you can update your settings manually from the Settings page.
The winner can be selected manually based on your own analysis, or your account manager can assist with interpreting the results.
What Happens After an Experiment
After an experiment ends, your store continues to run with whatever settings were active before — the experiment does not automatically change anything. You take action by:
- Reviewing the results in the experiments dashboard
- Promoting the winner to apply its settings as your new store defaults, or
- Keeping your existing settings if the experiment did not produce a meaningful difference
Previous experiment results are preserved in your experiments history for future reference.