> For the complete documentation index, see [llms.txt](https://docs.elevateab.com/elevate-helpcenter/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.elevateab.com/elevate-helpcenter/experiment-setup/ending-an-experiment.md). # Ending an Experiment *** Ending an experiment is two decisions, not one. First, whether the result is reliable enough to act on. Second, what to do with the winner. Marking an experiment complete stops traffic splitting and returns every visitor to your original experience. It does not apply anything to your store. Keeping a winning variation is always a separate, deliberate step. ### Is the result safe to act on? Check the significance status on the Results page before you end anything. | Status | What it means | What to do | | ------------------------------ | ---------------------------------------------------------- | ---------------------------------------------- | | **Significant** | The difference between variations is unlikely to be chance | Act on it | | **Near Significance** | Close, not confirmed | Let it run longer if traffic allows | | **Collecting Data / Trending** | Too early | Directional at best. Do not implement on this. | For what each status is calculated from, see [Statistical Significance](https://claude.ai/elevate-helpcenter/analyze/reports/statistical-significance). You can end an experiment at any time. Ending before significance is a legitimate choice when a variation is clearly harming revenue or is visibly broken. It is not a legitimate basis for rolling out a winner. #### Two things that are not a result **A variation that leads early.** Early leads reverse routinely. Significance exists precisely because a gap that looks large in the first few hundred visitors often is not there at all. **A losing variation given more time to recover.** If you would have stopped had the variation been ahead, stopping only when it is behind turns your stopping rule into part of the result. Decide the end condition before you launch. ### Step 1: Mark the experiment as complete 1. Open the experiment from your Elevate A/B Testing dashboard. 2. Select the **Settings** dropdown at the top right. 3. Select **Mark as Complete**. Traffic splitting stops immediately and all visitors see the original experience. The experiment moves to Done. Your results, reports, and raw data remain available. If you are only pausing for a promotion or an investigation, pause instead. See [Experiment Lifecycle](https://claude.ai/elevate-helpcenter/operate/experiment-setup/experiment-lifecycle). ### Step 2: Apply the winner If a variation won and you want to keep it, select **Apply Winner** from the experiment page. What that does depends on the experiment type, and for several types the change is yours to make in Shopify. | Experiment type | What happens when you keep the winner | | ------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | **Price Testing** | Elevate A/B Testing sets the winning price on the product | | **Product Image Testing** | Elevate A/B Testing applies the winning images to the product | | **Page Testing** | You publish the winning template as the default in your Shopify theme editor | | **Theme Testing** | You publish the winning theme from **Online Store → Themes** in Shopify | | **Split URL Testing** | You update your store to point at the winning URL permanently | | **Shipping Testing** | You update your Shopify shipping rates to match the winning configuration | | **Content Testing** | You either rebuild the changes in your theme, or create a [Personalization](https://claude.ai/elevate-helpcenter/personalizations/overview) with the same changes targeted to all visitors | | **Custom Code Testing** | You either add the winning JS or CSS to your theme permanently, or create a [Personalization](https://claude.ai/elevate-helpcenter/personalizations/overview) with the same code targeted to all visitors | | **Checkout Testing** | You create a [Personalization](https://claude.ai/elevate-helpcenter/personalizations/overview) with the same checkout configuration targeted to all visitors | #### Theme or personalization, for the types that offer both A personalization keeps the winning experience live immediately with no development work, which makes it the right choice when you want the result in production today or when the change may not be permanent. Putting the change in your theme is the right end state for anything you intend to keep. It survives if you ever remove Elevate A/B Testing, it is visible to whoever maintains the theme next, and it does not rely on a script running before the visitor sees the page. The usual sequence is a personalization now, a theme change at your next development cycle, and the personalization removed once the theme carries it. Leaving both in place means two systems making the same change, which is how a change becomes impossible to trace later. ### Step 3: Clean up #### Update everything downstream of the change If you changed what customers see, make sure the rest of your store agrees: * Email campaigns and automations * Social posts and ads * Google Shopping and other product feeds * Landing pages and content referencing the changed prices, pages, or shipping terms Price experiments are the sharp edge here. A price that changed on the product page while an old price sits in a live ad or a product feed produces support tickets and, in some markets, a compliance problem. #### Remove what the experiment added | What to check | Why | | ----------------------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ | | Custom event snippets in your theme | An event for a finished experiment keeps firing and clutters your theme. Leave it only if a future experiment will reuse it. See [Custom Events](https://claude.ai/elevate-helpcenter/operate/experiment-setup/custom-events). | | Variation templates that lost | Unpublished templates are harmless but accumulate. Remove the ones you will not reuse. | | Personalizations created from a previous winner | If you have since implemented the change in your theme, the personalization is now redundant. | ### What happens to my data? Nothing is deleted. Completing an experiment preserves your results, reports, and raw data indefinitely, and you can return to them at any time. Deleting an experiment is different, and permanent. If you want an experiment off your dashboard but want to keep the findings, mark it complete rather than deleting it. ### Can I resume an experiment I have ended? Yes. A completed experiment can be resumed back to Running, and existing variation assignments are preserved so returning visitors stay in the arm they were in. See [Experiment Lifecycle](https://claude.ai/elevate-helpcenter/operate/experiment-setup/experiment-lifecycle). Resume when you ended too early or when more data would settle an inconclusive result. Do not resume because you did not like the answer. ### What if the control won? Nothing to do. Your store continues unchanged. A control win is a real result, not a failed experiment. It tells you your current setup withstands the alternative you tried, and it saves you the cost of a change that would not have paid for itself. Log the hypothesis and what it scored, so nobody on your team proposes the same change in six months. ### Next steps * [Experiment Lifecycle](https://claude.ai/elevate-helpcenter/operate/experiment-setup/experiment-lifecycle) for pausing instead of ending, resuming a completed experiment, and what deleting removes * [Statistical Significance](https://claude.ai/elevate-helpcenter/analyze/reports/statistical-significance) for what each significance status is calculated from * [Custom Events](https://claude.ai/elevate-helpcenter/operate/experiment-setup/custom-events) for the snippets to remove from your theme during cleanup * [Personalizations](https://claude.ai/elevate-helpcenter/personalizations/overview) for keeping a winning experience live on the types with no automatic application *** \