Global Goals & Guardrails

Global Goals is where your testing program’s targets live: the north star it should move, the guardrails that protect other metrics while it does, the metric sources those numbers come from, and the program impact you’ve accumulated. You’ll find it at Testing Overview → Goals (workspace owners can edit; members see it read-only).

North Star

The north star is the one number the program optimizes — for stores, typically purchase conversion rate, revenue per session, or average order value; other site types can pick any conversion goal, optionally with a value per conversion so impact can be reported as estimated value.

The page shows the current value next to a baseline captured when your first metric source connects, plus the change vs. baseline. Changing sources starts a new baseline — series are never mixed.

Changing the north star doesn’t silently rewrite your plan. As the dialog puts it: “Changing your north star proposes a plan revision — nothing changes silently.” You’ll get a proposed plan update on the Testing Overview to review and approve, like any other plan change.

Guardrails

A guardrail is a metric a test is not allowed to hurt on its way to winning:

A test can’t be called a winner while it hurts a guardrail beyond its threshold. Insufficient data never blocks a winner — the guardrail simply shows as unmeasured.

How they work:

• Default: ecommerce workspaces start with a revenue-per-session guardrail at a −5% floor (relative). Tune the threshold or swap the metric any time.
• Up to 3 guardrails keep decisions crisp.
• Statuses: Passing, Watch, Breached, or Unmeasured — shown per guardrail here and rolled up into the Guardrails tile on the Testing Overview.
• Enforcement happens at decision time. A variant that clears the 95% winner bar but breaches a guardrail has its winner call withheld — you’ll see “Winner blocked by guardrail” on the experiment and dashboard instead of a celebration.
• Guardrails evaluate on in-experiment data (the same sessions the experiment measures), so they work even before any global source is connected.

Metric Sources

Your north star and program-impact series need a site-level data source. Three are supported, and each series is always labeled with the single source it came from — sources are never mixed within a series:

• Shopify analytics — connected automatically on Shopify workspaces: sessions, orders, revenue by day.
• Google Analytics 4 — works on both platforms and is the revenue answer for HTML sites. If you connected GA4 before Global Goals existed, you’ll see a one-time reconnect prompt: “Google added read access for reporting. Reconnect once to enable Global Goals — the AI assistant’s audiences keep working either way.” See GA4 integration.
• Site-wide tracking — Split Test Pro’s own opt-in, sampled measurement of pages you aren’t testing. Off by default; see Site-Wide Tracking.

Until a source is connected, goals use in-experiment data only — guardrails still work, but the north-star series and baseline wait for a source.

Program Impact

The impact section answers “is this program paying off?” two ways:

• Your north-star series from the connected source, with the baseline marked and shipped winners annotated on the timeline — this is the ground truth.
• Cumulative shipped-winner lift — an estimate summing the measured lifts of tests you shipped, with a plausible range, and an estimated monthly value when your metrics support it.

The estimate carries an honest caveat, straight from the page:

In-experiment estimates of tests you shipped. Because only strong-looking tests ship, realized site-wide impact is usually toward the low end; the Global Goals series above is the ground truth.

Next Steps

• The plan these goals steer: Testing Overview
• Fill in metrics for untested pages: Site-Wide Tracking
• What happens when a winner ships: Winner Rollout