Field guide

Shopify Rollouts Turned Your Theme Into an Experiment. Run It Like One

November 3, 2026

Shopify's Rollouts, announced in the June 5, 2026 changelog, let you schedule a theme or checkout configuration, release it gradually to a percentage of visitors, revert it automatically, and A/B test two full versions. The capability is new. The discipline it demands, a baseline, a holdout, and a pre-agreed verdict metric, is not.

What Rollouts actually changed

Before Rollouts, a theme launch was a cliff: edit in a duplicate, preview it yourself, publish to everyone at the busiest moment you dared. The June 2026 changelog changed the shape of the risk. A rollout can be scheduled for a set time, swapped in temporarily with an automatic revert, released to a slice of traffic, and run as an A/B test between two complete versions. Shopify copies your published theme automatically, and one rollout can bundle theme, checkout, and customer-account changes together. A September 2026 update redesigned the setup flow with explicit traffic splits and conflict warnings.

Gradual release is the feature with the money in it. A design that quietly costs 4 percent of conversion is a bad quarter when it ships to everyone on day one. Shipped to 10 percent with a comparison running, it is a two-week lesson with a small, known price.

Reset the yardstick before you test

An experiment is only as good as its baseline, and Shopify moved the baseline in the same season. The September 2026 session-counting change, sessions running through activity, ending after 30 minutes idle, bot sessions excluded, means conversion rates from before the change are not comparable with rates after it. Test a rollout against a simultaneous control, never against last month's rate. The whole point of the A/B structure is that both arms live under the same counting rules, the same traffic mix, and the same promotions.

Pick the verdict metric before launch and write it down: conversion rate on the tested templates, with orders and revenue per session as guards. A rollout judged afterward on whichever metric happened to move is a story, not a test.

What to test first, and what to leave bundled out

Start with the change you most fear being wrong about, on the template that carries the revenue: usually the product page. Keep the first rollout to one bundle with one idea in it. Rollouts allow theme, checkout, and account changes in a single release, and that power is a trap for a first test, because a bundled win or loss cannot be attributed and the losing element rides along with the winning one.

Run long enough to cover your full weekly cycle twice. A test read on a weekend-heavy slice will happily crown a design your weekday traffic hates.

The revert is a feature, rehearse it

Automatic revert is the part teams skip configuring because it feels pessimistic. Configure it anyway, with a floor on the verdict metric that triggers it without a meeting. The stores that get hurt by gradual releases are not the ones whose test lost; they are the ones whose test lost on a Friday and was discovered on a Tuesday. A rehearsed, automatic way back is what makes a 10 percent release genuinely cheap to be wrong with.

Frequently asked questions

What are Shopify Rollouts?

A Shopify feature announced June 5, 2026 that schedules theme or checkout configurations, releases them gradually to a percentage of visitors, reverts them automatically, and A/B tests two full versions against each other. One rollout can bundle theme, checkout, and customer-account changes.

How long should a theme A/B test run?

Long enough to cover your full weekly traffic cycle at least twice, with enough orders in each arm to judge your chosen metric. Reading a test on a weekend-heavy slice is the classic way to crown a design your weekday buyers dislike.

Can I compare a rollout against last month's conversion rate?

No. Shopify changed session counting in September 2026, so earlier conversion rates use a different denominator. Judge a rollout against its simultaneous control arm, which shares the same counting rules, traffic, and promotions.

What should my first Rollouts test be?

One idea, on the template that carries the revenue, usually the product page. Resist bundling theme, checkout, and account changes into the first test: a bundled result cannot be attributed, and the losing part hides inside the winning result.