> ## Documentation Index
> Fetch the complete documentation index at: https://docs.getjurni.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Create and Run an A/B Test in Jurni

> Choose a new or existing campaign link, configure page variants and traffic weights, and check the test before sending visitors.

This guide covers the **Jurni setup steps** for an A/B test. If you are still deciding what to test or how to structure the hypothesis, start with [Run your first A/B test](/academy/playbooks/first-ab-test).

## Before you create the test

Have these ready:

* **Control** — the current page or destination.
* **Variant** — the alternative you want to compare.
* **Hypothesis** — one sentence explaining what changed and why it may matter.
* **Primary metric** — the result that will guide your decision.
* **Campaign link plan** — whether you need a new Smart Link or will intentionally use an existing one.

Both pages should already be reviewed on mobile/desktop and through the real product or conversion path.

## Create the experiment

<Steps>
  <Step title="Start a new experiment">
    Open **Experiments / Tests** and create a new experiment. Give it a name your team will understand later, such as **Meta — Hero proof test — September** instead of “Test 4.”
  </Step>

  <Step title="Record the hypothesis">
    Use the description to write what you are testing. A useful format is: **We believe \[change] will improve \[metric/behavior] because \[reason].**
  </Step>

  <Step title="Choose the Smart Link">
    Under **Link**, choose **Create new link** when this campaign needs a new entry URL, or **Use existing link** when you deliberately want an available Smart Link's current traffic to enter the experiment.
  </Step>

  <Step title="Add the control and variant destinations">
    Give each variant a clear **Variant Name** and choose its **Destination**. Names such as **Current hero** and **Benefit-led hero** are more useful later than A and B.
  </Step>

  <Step title="Review Weight %">
    With two variants, the initial setup starts evenly. Change the weights only when your test plan calls for it. Adding or removing variants redistributes weights, so recheck the final numbers.
  </Step>

  <Step title="Add Routing Rules only when intentional">
    Routing Rules send matching URL-parameter traffic to a specific variant. Leave them out for a straightforward randomized comparison unless you have a campaign-routing reason to use them.
  </Step>

  <Step title="Start the experiment">
    Review the Smart Link, destinations, weights, and any routing rules. Save/start the experiment and wait for confirmation.
  </Step>
</Steps>

## Create new link or use existing link?

### Create new link

Choose this when:

* the campaign is new;
* no existing Smart Link should be repurposed;
* you want a clean campaign entry URL for the test.

The slug becomes part of the campaign URL you will share or put in ads.

### Use existing link

Choose this when:

* the campaign already uses a Smart Link;
* you intentionally want that link's traffic to start entering this test.

<Warning>
  An existing Smart Link can already be live in Meta ads, emails, creator campaigns, QR codes, or other traffic. Connecting it to a new experiment can change where those visitors go. Treat this as a live routing change.
</Warning>

The existing-link selector excludes Smart Links already connected to another experiment.

## Set the variants clearly

For each variant, check:

* **Variant Name** — describes the idea being tested;
* **Destination** — the exact page or supported custom URL;
* **Weight** — intended share of normal experiment traffic.

For a custom or external destination, the page still needs the appropriate Jurni tracking/attribution path. A redirect working is not proof that experiment results will be measured correctly.

[Test against your existing website page →](/experiments/testing-against-website)

## What should I put in the ad?

Use the **experiment's Smart Link** as the campaign destination.

Do not put the direct variant URL in the ad when you expect Jurni to split experiment traffic.

After setup:

1. copy the Smart Link;
2. open that exact URL yourself;
3. add the campaign parameters you expect to use;
4. verify the page and purchase path;
5. check tracking and consent.

[Test the campaign link →](/publishing/check-campaign-link)

## Why do I keep seeing the same variant?

Jurni can keep a visitor's assignment consistent. Repeatedly refreshing the same browser is **not** a useful 50/50 traffic test.

Use a fresh browser/context when appropriate for assignment QA, and read [Traffic assignment](/experiments/traffic-assignment) before interpreting a few manual visits as traffic imbalance.

## Routing Rules: when should I use them?

Use Routing Rules when a parameter should **deterministically** control the variant.

Examples:

* `utm_term=persona1` → Persona 1 page;
* a specific campaign parameter → matching page.

This is different from a normal weighted test because matching traffic is intentionally routed before the standard split.

[Configure Routing Rules →](/experiments/routing-rules)

## Before sending traffic

QA both variants through the Smart Link and confirm:

* correct page/product;
* real offer/price;
* product options/subscription where relevant;
* Add to Cart/checkout;
* mobile/desktop;
* forms/review widgets;
* tracking/consent;
* campaign parameters;
* intended routing rules.

For a copy-ready AI QA prompt, use [Experiment prompts](/academy/prompt-library).

## While the test is running

Avoid casual changes to:

* the pages;
* the product/offer;
* tracking;
* weights;
* routing rules.

If a change is necessary, record it. Combining results from materially different versions can make the test harder to interpret.

The destination URL for an existing variant is locked in the current variant-editing flow; do not plan on silently swapping the underlying destination in place.

## Reading and ending the test

Use [Read experiment results](/experiments/reading-results) with the same test period and primary metric you chose before launch.

Before changing live routing:

* [Pause, end, and reuse](/experiments/pause-end-reuse)
* [Optimization Loop](/academy/playbooks/optimization-loop)

Selecting a winner when ending the experiment is a **traffic-routing decision**. It is not automatically a statistical claim.


## Related topics

- [Run your first Jurni A/B test](/academy/playbooks/first-ab-test.md)
- [What Are Experiments? A/B Testing in Jurni](/experiments/what-are-experiments.md)
- [How Experiment Traffic Is Assigned](/experiments/traffic-assignment.md)
- [Experience Analytics](/analytics-reporting/experience-analytics.md)
- [Test a Jurni Page Against an Existing Website Page](/experiments/testing-against-website.md)