> 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/price-testing/launching-a-price-experiment.md).

# Launching a Price Experiment

What happens when a price experiment goes live in Elevate A/B Testing, how long to run it, and what not to change while it runs.

***

### Before you launch

* [Preview and QA](https://claude.ai/chat/preview-and-qa.md) is complete.
* Your first price experiment has passed review. Later ones skip this.
* No sale or catalogue price change is planned during the run.
* No other running experiment includes the same products, unless one of them uses traffic isolation.

Select **Launch Experiment**. The experiment moves to **Running**.

### What happens at launch

Visitors are split according to your traffic allocation and assigned to a variation. Assignment is sticky: a visitor who returns sees the same price they saw before.

Visitors arriving from Google Ads are excluded and see your control price, so your ad price and your landing page price always match.

Results begin populating immediately, but do not read them yet.

### How long to run

Run until Elevate identifies a winner. There is no fixed duration, because the time needed depends on your traffic, your conversion rate, and how large the difference between variations turns out to be.

Three things extend a run:

* **Narrow audience targeting.** Every filter shrinks the pool entering the experiment.
* **Uneven traffic allocation.** Weighting the control at eighty percent means the variation collects data four times slower.
* **A small price difference.** A one percent change takes far longer to detect than a fifteen percent one.

Run for at least one full business cycle, usually a week, even if a winner appears sooner. Weekday and weekend buyers behave differently, and a result built on three weekdays does not describe your store.

### What not to do while it runs

**Do not stop early.** Results swing in the first days. A variation that looks like a clear winner after 48 hours frequently is not.

**Do not edit variations.** Changing a price mid experiment mixes two different tests into one dataset. The result becomes unreadable and cannot be recovered.

**Do not change catalogue prices.** If you run a sale by editing prices in Shopify, your control moves and your variations do not. Run sales as discounts instead.

**Do not disable the theme extension.** If it is switched off while an experiment is running, your storefront returns to catalogue prices and no new data is collected. Carts already in progress are not affected until the experiment is stopped, so a customer may still be charged a test price. If you need to stop an experiment, stop the experiment rather than disabling the extension.

**Do not add another experiment on the same products** unless one uses traffic isolation.

### Pausing

Pausing stops an experiment collecting data without ending it. Use it if something needs attention and you intend to resume.

A paused experiment is not a stopped one. See [Ending a Price Experiment](https://claude.ai/chat/ending.md) when you are ready to finish.

### Checking in

It is fine to watch the numbers. Just do not draw conclusions from them until Elevate indicates a winner.

Early in a run, look for problems rather than results: zero visitors on a variation, a conversion rate far outside normal, or orders at a price you did not set. Those mean something is wrong with the setup, not that a variation is winning.

See [Reading Price Experiment Results](https://claude.ai/chat/reading-results.md) for what the numbers mean.