Calculated Metrics

Split Test Pro ships a handful of built-in ratio metrics — Revenue per Session, Average Order Value — but they’re fixed to the Shopify checkout. Calculated metrics let you build your own, GA4-style: pick a numerator, pick a denominator, name the metric, choose a display format, and every experiment’s results tab renders it per variant with the same Bayesian lift, credible interval, and probability-to-be-best treatment the built-ins get.

Calculated metrics are defined once per workspace (like conversion goals and funnels) and work on both platforms — HTML and Shopify.

What You Can Build

Metric	Numerator	Denominator	Format
Signup value per session	Value of Signup	Sessions	Currency
Revenue per checkout	Checkout revenue (built-in)	Orders	Currency
Quiz completions per PDP view	Count of Quiz Completed	Count of PDP Views	Percent
Upsells per purchase	Count of Upsell Accepted	Count of Purchases	Number

Numerators aggregate a quantity per variant:

• Value of a goal — the summed value field on the goal’s events (see Continuous Metrics for how events carry values).
• Count of a goal — the goal’s conversion count, the same count the binomial goal columns report.
• Checkout revenue (Shopify only) — the built-in checkout revenue aggregate behind Revenue per Session.

Denominators count the trials the numerator is spread over:

• Sessions — per-variant session count.
• Count of a goal — any countable built-in event or custom goal.
• Orders (Shopify only) — the checkout count behind Average Order Value.

Defining a Metric

1. Open Settings → Calculated metrics (workspace owners only — members see the definitions read-only, like goals and funnels).
2. Click New metric, name it, and pick a numerator and denominator from the platform-aware dropdowns.
3. Choose a format — Currency, Number, or Percent — which controls how the value renders on results.
4. Save the definition, then add it to any experiment that should report it (below).

Adding a Calculated Metric to an Experiment

Calculated metrics are tracked per experiment exactly like goals. On the experiment’s General tab, open the Metrics panel’s + Add Metric dropdown — workspace calculated metrics appear there alongside your goals (marked Calculated). Once added, the metric behaves like any tracked metric: it renders on the Results tab (and in the AI analysis), it can be removed with the same ×, and it’s eligible as the primary metric.

• An experiment reports only the calculated metrics you’ve added — nothing is tracked by default.
• A tracked metric that’s later deleted in Settings is simply skipped — a stale selection never breaks results.
• Like goals and funnels, the tracked set is editable while the experiment is a draft or paused.

Reading the Results

Each calculated metric gets its own card on the experiment Results tab, one row per variant:

• Trials — the denominator count (sessions, orders, or the goal’s conversions).
• Value — numerator ÷ denominator, rendered in the metric’s format.
• vs Control — the percent change of the variant’s value against control.
• 95% CI (lift) — the credible interval of the lift, as a percent range.
• Prob. Best — the probability this variant has the best true value of the metric.

Under the hood every calculated metric is modeled as a continuous measurement: the engine derives a per-trial standard error with the same zero-inflated delta-method formula the built-in Revenue per Session uses, then feeds it through the Normal-model Bayesian pipeline. In fact, a calculated metric defined as Checkout revenue ÷ Sessions reproduces the built-in Revenue per Session numbers exactly.

Using a Calculated Metric as the Primary Metric

A calculated metric can be the experiment’s primary (stopping) metric — the measurement the winner verdict is judged on. Add the metric to the experiment’s tracked metrics, then mark it with the same Primary radio as any goal. The results tab badges its card as Primary, and the winner is declared from its probability-to-be-best, exactly like a conversion goal.

Two things to know:

• The headline conversion count and the daily chart still come from a countable event (your first goal, or purchases on Shopify) — a ratio has no per-day event series of its own.
• If the metric is later deleted in Settings or removed from the experiment’s tracked metrics, the primary becomes stale: starting the experiment is blocked with a clear message until you pick another primary.

Currency Handling

Monetary numerators aggregate in your shop/site currency — the same policy as the built-in revenue metrics — and the metric is labeled with that currency code. There’s no exchange-rate conversion: if your events report mixed currencies, the amounts are aggregated as recorded.

Caps and Validation

• Up to 10 calculated metrics per workspace.
• Names must be unique; numerator and denominator must differ.
• Sources are validated against your platform — HTML workspaces can’t reference checkout revenue or orders — and against the workspace’s goals, with the same server-side checks the Settings UI mirrors.

Common Mistakes

• Dividing by a rare denominator. Value of Signup ÷ Count of Purchases on a store with 3 purchases produces a huge, unstable ratio. Prefer sessions or a high-volume event as the denominator.
• Expecting a value where none is tracked. A “Value of …” numerator only counts events that carried a value payload. If the goal fires without values, the metric reads zero — see Continuous Metrics for wiring values.
• Reading too much into early lift. Ratio metrics inherit the variance of their numerator. Give them the same patience you’d give any continuous metric — 100+ trials per variant before drawing conclusions.

Next Steps

• Wire up the valued events your numerators aggregate: Continuous Metrics.
• Define the goals your sources draw from: Conversion Goals.
• Understand the model behind the lift and probability numbers: Bayesian Stats Explained.