Reading results

How to interpret the Results tab and decide when to end an experiment.

Results tab

The Results tab on each experiment shows a table with one row per variant. Results are computed by the hourly rollup job, or on demand when you click ↻ Refresh results.

Funnel overview

At the top of every experiment page, the funnel shows aggregate stats across all variants:

Metric	Source
Visitors	Unique visitor IDs allocated to the experiment
Sessions	PAGE_VIEW events
Add to Cart	ADD_TO_CART events
Checkout	INITIATE_CHECKOUT events
Orders	Orders attributed via cart attributes
Revenue	Order revenue attributed to the experiment

Conclusion banner

The banner above the results table tells you where the experiment stands:

State	Meaning
Collecting data	Fewer than 100 sessions per variant — too early to draw conclusions
Not yet significant	Enough data but p ≥ 0.05 — keep running
Ready to conclude	p < 0.05 — results are statistically significant

When to stop

Don’t stop an experiment just because you see a positive trend early. Early results are volatile and will often revert. The minimum recommended runtime is 7 days.

Stop the experiment when all three conditions are met:

1. At least 7 days have elapsed (one full day-of-week cycle)
2. You have at least 100 sessions per variant
3. p < 0.05

If an experiment is running for 4+ weeks with no significance, the effect size is likely too small to be meaningful — consider ending it.

Guardrails

The Auto-stop guardrails on the Overview tab can automatically pause your experiment if:

• Sample Ratio Mismatch — visitor allocation drifts significantly from target weights (broken assignment pipeline)
• Control CVR drop — your control variant’s conversion rate drops >20% (broken variant hurting the store baseline)