Stopping a theme experiment has two parts: concluding it in Cooee, and tidying up the challenger theme in Shopify. The second part is easy to overlook and matters more than it looks.
Concluding the experiment
Open the experiment in Cooee and click Stop. New visitors are no longer assigned.
Review your primary metric alongside the guardrails.
Take the matching action in Shopify Admin under Online Store → Themes.
Decision | Action in Shopify Admin |
Variant wins | Publish the challenger theme as your new live theme |
Control wins | Delete the challenger theme |
Inconclusive | Extend the runtime — or, if you are stopping for good, delete the challenger theme |
Your experiment data stays available in Cooee after stopping, whichever action you take.
Why you must publish or delete the challenger theme
When Shopify serves a visitor the challenger theme, it remembers that choice in that visitor's browser. That record belongs to Shopify, not to Cooee, and stopping the experiment does not clear it.
So if you stop the experiment and leave the challenger sitting in your theme library, visitors who were in the variant arm keep seeing it when they come back — potentially for a long time. They are looking at a storefront you are no longer testing and may no longer be maintaining.
Both closing actions resolve it properly:
Publishing the challenger makes it the live theme, so every visitor sees it. That is the intended outcome when the variant wins.
Deleting the challenger removes it from your store, so Shopify falls back to your live theme for everyone — including visitors still holding the old preference.
Duplicate before you delete. In Shopify Admin, use Themes → ⋯ → Duplicate first. The copy keeps all your design work for later reference and is never served to visitors. Deleting a theme cannot be undone.
Cooee does not publish or delete themes for you. Stopping an experiment never changes your theme library. That action is always yours.
If a theme changes while the test is running
Cooee watches your theme library and responds automatically when a theme in the experiment changes:
What happened | What Cooee does |
You published the challenger theme as your live theme | Stops the experiment — once the challenger is live, both arms are the same and there is nothing left to compare |
You deleted the challenger theme | Stops the experiment, so no visitor is sent to a theme that no longer exists |
You deleted the live (control) theme | Pauses the experiment and notifies your team |
If you publish a different theme mid-test — one that is neither arm — your control experience has changed underneath the experiment. Cooee flags this, and the results from before and after that point are not comparable. Restarting is usually the honest choice.