/Docs

Metrics

Metrics are how A vs B knows whether your experiment succeeded or failed. You define what a "conversion" means — a button click, a page visit, a form submission, a purchase — and the platform measures how often each variation achieves it. This section covers the foundational metric types as well as the advanced metric flavors used by stats-conscious experimenters: ratio metrics with the delta method, quantile metrics with bootstrap CIs, composite metrics with weighted statistics, and winsorization for outlier-resistant analysis.

In this section

Tracking EventsThe new prerequisite — define click, pageview, and custom events once, then reuse them across every metric.
What Are Metrics?Understand primary vs secondary metrics and the metric types available.
Click MetricsTrack clicks on any element using a CSS selector.
Pageview MetricsTrack visits to a URL using simple, exact, substring, or regex matching.
Custom EventsFire metric conversions from your own JavaScript code using avsb.track.event().
Revenue TrackingPass a revenue value with custom events to see Revenue Impact in your results.
Ratio MetricsTrack per-visitor ratios like revenue-per-visitor with delta-method confidence intervals.
Quantile MetricsMeasure p90 / p95 / p99 of continuous values such as page-load time, with bootstrap CIs.
Composite MetricsCombine multiple metrics into a single weighted decision signal for multi-objective experiments.
WinsorizationCap extreme values at a chosen percentile so outliers don't swing the test outcome.
Managing MetricsCreate, edit, delete, and reuse metrics across experiments.

Key concepts

Every experiment must have exactly one primary metric. This is the single measurement that determines which variation wins. You can add as many secondary metrics as you like — they provide additional context without affecting the winner calculation.

Metrics are defined at the project level, not the experiment level. That means you create a metric once and can attach it to as many experiments as you need. Any changes you make to a metric will apply to all experiments using it.