Metrics and Guardrails
Every experiment measures a set of metrics. One of them, the primary target metric, decides whether the experiment succeeds. The rest act as guardrails that catch side effects. This page explains the metric types, the tracking behind them, and how they work together.
Metric kinds
Section titled “Metric kinds”Shuttlebase supports two kinds of metrics.
Conversion-rate metrics
Section titled “Conversion-rate metrics”Conversion-rate metrics measure the share of visitors who complete an action. Examples include clicks on a specific element, add-to-cart, and pageviews of a target page. The result is a rate: visitors who did the action divided by visitors in the variant.
Revenue metrics
Section titled “Revenue metrics”Revenue metrics measure money, and each one displays three views:
| View | What it shows |
|---|---|
| Purchase conversion rate | The share of visitors who made a purchase. |
| Average order value (AOV) | Revenue divided by number of orders. |
| Revenue per visitor (RPV) | Revenue divided by all visitors in the variant, buyers or not. |
The three views answer different questions. A variant can raise purchase conversion rate while lowering AOV, for example when it nudges visitors toward cheaper products. Revenue per visitor combines both effects into a single number, which makes it the most complete view of revenue impact.
Primary metric and guardrails
Section titled “Primary metric and guardrails”Each experiment has one primary target metric. Success is judged on that metric alone. Picking a single primary metric before launch keeps the decision honest, because a variant that misses its target cannot be declared a winner by pointing at some other number that happened to move.
Guardrail metrics protect the rest of your funnel. Shuttlebase watches them throughout the experiment. If a variant wins on the primary metric but hurts a guardrail, the platform flags it rather than treating it as a clean win.
For example, an experiment might target add-to-cart clicks with revenue as a guardrail. A variant that lifts add-to-cart but lowers revenue per visitor gets flagged, because more carts that produce less money is not the outcome you want.
Measurement data source
Section titled “Measurement data source”All experiment measurement runs through Shuttlebase’s own first-party snippet installed on your site. The snippet assigns visitors to variants and records the events described above, so exposure and outcome come from the same source. Details on what the snippet collects are at Tracking snippet.
If you connect GA4, Shuttlebase uses it only at the account level: for high-level performance analysis and for finding underperforming segments that suggest new experiment ideas. GA4 data never feeds experiment results. This separation avoids the mismatches that appear when a testing tool and an analytics tool define visitors and sessions differently.