Analytics Engine
Synthetic Account Experiments
Campaign and flow A/B testing performance from verified authenticated web-session contracts.
Distribution
Experiments by winner metric
| Experiment | Channel | Winner metric | Started | Test size | Win probability | Status |
|---|
Exploratory statistical layer
Within-experiment variation evidence
Revenue means attributed revenue per delivered recipient in neutral currency units because the source did not provide a verified currency code. Intervals use a Beta-Binomial conversion posterior multiplied by observed average order value. Order-value uncertainty is not included. These are not causal incrementality estimates.
Uncertainty
Revenue-per-recipient lift intervals
All buyer-reliable comparisons, sorted by probability that the variation beats the first listed baseline. Points show posterior medians; lines show 95% intervals. Values are neutral currency units per delivered recipient.
Exposure volume
Summed experiment message counts
Pooled message-exposure events for orientation. Profiles may recur across experiments, so these are not unique people or a customer funnel.
Selection alignment
Source selection versus RPR selection
Comparison is limited to experiments where both selections exist. Disagreement is not an error because Klaviyo may optimize a different winner metric.
Exploratory association
Click lift versus attributed RPR lift
Each point is a first-listed-baseline comparison. Amber points meet the 100-buyers-per-arm reliability threshold. Axes are clipped at the 95th percentile of absolute values for readability; clipped points remain in the plot. This shows association, not causation.
Program history
Experiment starts by quarter
The current quarter is provisional. Missing quarters mean no captured starts, not missing capture records.
| Experiment | Variation | RPR baseline | RPR variation | Observed lift | 95% posterior interval | Probability better | Decision category |
|---|
Lifecycle mapping
Flow names and IDs are retained, but no governed lifecycle-stage mapping exists in the source. The dashboard leaves lifecycle stage unmapped instead of inferring it from names.
Attribution sensitivity
Awaiting alternate-window capture
Required inputs: current and preview attribution settings plus campaign and flow revenue under each window. No comparison is drawn from a single attribution model.
Causal incrementality
Awaiting a true holdout
Required input: no-send, global holdout, geo experiment, or another causal control. Marketed-arm differences remain attributed performance only.
Global Experiments
No holdout experiments configured
Synthetic Account is eligible for Global Experiments. The verified holdout list and aggregate analytics contracts currently return an empty state.