Experiment measurement

Compare control and variation metrics, adjust traffic, review AI suggestions, and decide whether to apply experiment changes.

Experiment measurement helps you compare the control with the active variation before making a live-store decision.

Entry path

Open a store project → More → Experiments → select an experiment → Performance.

Confirm the experiment goal and current traffic allocation.
Wait for a useful sample before interpreting small percentage changes.
Review variation changes in the Experiments tab before considering Apply Changes.
Ask a workspace owner before an action that deploys changes or stops live testing.

Control or section	What happens or appears
Experiments	Opens cycle messages, plans, preview, changes, and pending requests.
Performance	Opens runtime, sample size, traffic split, chart, and metric table.
Metric selector	Changes the chart to a supported measure such as conversion, add to cart, checkout, purchase, order value, revenue per visitor, banner clicks, cart abandonment, or error rate.
Control / Variation chart	Compares the recorded series for the original and experiment version.
Traffic control → Update	Saves a new experiment allocation within the allowed range.
AI Insight → Allow	Approves the displayed traffic suggestion.
AI Insight → Cancel	Dismisses that suggestion without approving it.
Ask Runner AI	Sends a message inside the experiment workspace.
View Plan	Opens the current cycle plan and to-do list.

Open Performance.
Confirm runtime, sample size, and traffic split before reading the trend.
Select the metric that matches the experiment goal.
Compare control and variation over the same period.
Check supporting metrics for error, abandonment, or revenue tradeoffs.
Return to Experiments and review the exact changes that produced the variation.
Read any AI Insight request before allowing a traffic change.
Choose to keep measuring, adjust traffic, stop, or prepare to apply only after reviewing both behavior and implementation.

Action	What the first click does	What confirmation means
Apply Changes	Opens Apply Experiment Changes.	Confirming applies the experiment changes to the main store path and starts production deployment. Review the named experiment first.
Stop	Opens Stop Experiment with warnings.	Confirming ends the current experiment. If experiment authorization remains enabled, Runner may start another; use project settings when you intend to disable experiments completely.
Traffic Update	Submits the new allocation.	Future eligible traffic uses the updated split after success.

State	Meaning and next step
No cycle data	Measurement has not produced a cycle result yet. Keep the experiment running or inspect its status.
Unable to load metrics	Reload Performance and retry later; do not apply changes from missing data.
Flat or mixed result	The available evidence does not show a clear improvement. Continue measuring or stop without applying.
Traffic update failed	The current value was not stored. Reload before trying a valid allocation again.
Apply failed	No successful production application was confirmed. Preserve the error and verify the live store before retrying.