Layer 3 — verify

Run tests you can actually trust

Every fix from a Polaris report can go straight into an experiment on the same data — with a calculator built in, so "it worked" has to survive a Wilson interval, not just a hopeful glance at the dashboard.

Insertion types

Three ways to run a variant

Standard

Redirect

Classic split URL redirect for full-page variants — simplest setup, works on any stack.

SPA

In-page JS insertion

DOM-level variant swaps for React/Vue/single-page apps, without a full reload breaking app state.

Event-based

Trigger on interaction

Variants that activate on a specific event — a click, a scroll depth, a cart state — not just page load.

Targeting & weighting

Point it at the right traffic

URL + audience targeting

Run a test on one URL pattern, one traffic source, one device class — or stack conditions to reach exactly the segment the report flagged.

Variant weights

Split traffic 50/50, or route more traffic to a promising variant early and adjust — the calculator recalculates sample size either way.

Free tool — no login

Sample size & significance calculator

The same math the Experiments module uses internally: Wilson confidence intervals, not raw percentage differences. Plan a test before you launch it, or check whether a result you're already seeing is real.

Result

visitors needed, per variant

days to reach that, at your traffic level

Result

Fill in both variants and check.

Wilson 95% CI — A: · B:

Uses a two-proportion z-test (α = 0.05, 80% power) for planning, and a Wilson score interval for the significance check. For guidance, not a replacement for your own statistician on high-stakes decisions.

Get started

Find what's worth testing first

A free audit tells you where to point your first experiment.

Get a free audit of your site