5 min read
Rollouts
A rollout is a set of store changes that goes live at a time you choose, with nobody clicking at midnight. Find it under Markets › Rollouts in the sidebar.
There are two kinds:
- Launch: the changes go live at the start and stay. Use it for a new look you want live on Monday morning.
- Event: the changes run from the start to the end. At the end they roll back (the store goes back to what was live before) or you keep the changes, as you chose. Use it for a Diwali sale, a weekend offer or a festive theme.
What a rollout can change
| Change | What happens at the start | What rolling back does |
|---|---|---|
| Online store theme | The chosen theme is published. | The theme that was live before is published again. |
| Checkout branding | The checkout’s logo, colours, font and corners are replaced. | The previous checkout look comes back. |
| Discounts (up to 250) | Each chosen discount is turned on or off. | Each discount goes back to how it was. |
One theme change and one checkout change per rollout. A discount can be in only one rollout that has not ended.
Edit theme for this rollout makes a copy of your live theme in the theme library and picks it. Open the copy in Customize to make the sale look; your live theme is not touched until the rollout starts.
Create a rollout
- Go to Rollouts and click Create rollout.
- Give it a name (shoppers never see it) and pick Event or Launch.
- Click Add changes and choose a theme, the checkout branding or discounts.
- Under Schedule, set when it starts and, for an event, when it ends and what happens then. Times are India time (IST).
- Click Save. With a start date and at least one change, it is Scheduled; without them it stays a Draft.
Statuses
| Status | Meaning |
|---|---|
| Draft | No start date, or no changes yet. Nothing will happen. |
| Scheduled | It will start by itself at its start time. |
| Active | Its changes are live (an A/B test: its Treatment visitors see them). An event shows when it will end. |
| Paused | An event rolled back early. Resume puts its changes back until the end. |
| Ended | Done. A launch ends as soon as it starts. |
| Archived | Kept for the record only. |
While it runs
- Start now applies a scheduled rollout straight away.
- Pause (events) puts the store back early; Resume applies the changes again.
- End now (events) asks whether to roll back or keep the changes. This cannot be undone.
- A running rollout’s changes cannot be edited; its name, end time and end choice can.
Rolling back puts back what was replaced, as it is now. If you edited your everyday theme while the sale ran, those edits stay. If you changed something the rollout set by hand (turned a sale discount off yourself, for example), rolling back leaves your choice alone.
A/B test a theme
Not sure the new look sells better? Run the rollout as an A/B test first. Under Audience, turn on Run as an A/B test:
- Control visitors see the store as it is. Treatment visitors see the rollout’s theme. Nothing changes for everyone until you decide.
- Traffic split: how the tested visitors divide between Control and Treatment, 50/50 unless you move it.
- Traffic allocation: the share of all visitors in the test. The rest see the store as it is and are not counted.
- Eligible markets: test in chosen markets only (leave all unticked for every market).
- Target visitors: how many visitors the test should reach before you decide. The Reached ring on the results fills up towards it.
- Notes: write down what you are testing and what would make you ship it. Only your team sees them.
A visitor stays in the same arm on every visit. Only visitors who accepted analytics cookies are tested; everyone else, and every search engine bot, sees the Control.
What can be tested. An A/B test splits theme changes only. Checkout branding and discounts apply to every shopper at once (a discount decides the price in the cart, the checkout and the order), so a rollout with those cannot be an A/B test. Test the theme first, then run a normal rollout for the sale.
Results
Once the test starts, Results shows, for Control and Treatment side by side: visitors, sessions, add-to-cart rate, conversion rate, orders and revenue per visitor. Above them a hint reads Not enough data yet until each arm has 100 visitors and there are 10 orders in all, then Treatment ahead with ~95% confidence (or Control, at ~90, 95 or 99%), or No clear winner yet. The hint is a two-proportion z-test on conversion rate.
Ending the test
- Ship Treatment publishes the rollout’s theme for every visitor and ends the test.
- Keep Control ends the test and leaves the store as it is.
- Pause shows everyone the Control until you resume; the numbers so far are kept.
An Event that is also an A/B test ends by itself at its end time: Keep the changes ships the Treatment, Roll back keeps the Control. While a test runs, its split, traffic and markets stay fixed so the numbers stay fair; its name, notes, target and end can still change.
Conflicts
When you save, the rollout is checked against your other rollouts:
- Resolve rollout conflicts: another rollout changes the theme or the checkout at overlapping times, or already holds one of the discounts. Change the times or the changes, then save.
- Review rollout conflicts: the theme to publish was deleted. If it is still gone at the start, the rollout does not start, goes back to Draft, and says why at the top of its page.
The theme library, the discounts list and Settings → Checkout → Branding show an In rollout chip on anything a rollout will change.
Who can use it
Rollouts have their own permission in Settings → Staff → Roles: Rollouts: read to see them, Rollouts: manage to create, start, pause, end and delete them. Administrators have both. A rollout publishes themes and switches discounts when it runs, so give manage only to people trusted with those.
Over MCP, an AI assistant can read and plan rollouts; ending or deleting one always waits for a person’s approval.