Metrics & Results
What NavonaAI tracks during experiments and how winners are determined
Metrics Tracked Per Variant
NavonaAI tracks the following for each variant throughout the experiment:
| Metric | Description |
|---|---|
| Sample size | Number of carts assigned to this variant |
| Checkouts started | Carts that proceeded to checkout |
| Orders completed | Carts that resulted in a purchase |
| Revenue generated | Total order value attributed to this variant |
| Cart-to-order rate | Orders ÷ total carts |
| Cart abandonment rate | Carts that never became orders |
| Checkout abandonment rate | Shoppers who reached checkout but didn't complete the purchase |
All metrics are calculated per variant, so you can directly compare how each configuration performed with its share of traffic.
Choosing a Goal KPI
When creating an experiment, you select the primary KPI used to determine the winner:
| KPI | Best For |
|---|---|
| Conversion rate | Maximizing the share of carts that become orders |
| Abandonment rate | Minimizing lost carts (lower is better) |
| Orders created | Maximizing order volume |
| Revenue total | Maximizing recovered revenue (accounts for average order value differences) |
Which KPI should I choose? For most stores, revenue total is the best primary KPI — it accounts for both conversion rate and order value. If you're optimizing for volume over revenue, use conversion rate instead. Your account manager can help you decide.
How Winners Are Determined
When an experiment completes:
- NavonaAI aggregates all events from carts assigned to each variant during the experiment window
- The primary goal KPI is compared between the two variants
- The variant that performed better on the chosen KPI is identified as the winner
The winner can then be promoted to apply its configuration as your store's default settings.
If both variants show identical performance on the primary KPI, the experiment is flagged for manual review. Your account manager can help interpret borderline results.