# Start an A/B test

> Start a site-wide A/B test under Analytics › A/B Tests by choosing the control group fraction and the hypothesis, so you can measure what Releva adds to your shop.

An **A/B test** holds back a share of your visitors from Releva so you can compare them with everyone else. While it runs, the **control group** gets no product blocks, banners, stories or personalised search and is left out of campaigns; the **treatment group** gets Releva as usual. This guide starts a test with a 20 % control group. For what the two settings mean and how the results are computed, read [How A/B tests work](https://releva.ai/documentation/learn/ab-tests/how-ab-tests-work/).

## Before you start

* No test is running. A shop can run one A/B test at a time, and **Start A/ B Test** is disabled while one runs. Stop the running test first: see [Check and stop an A/B test](https://releva.ai/documentation/guides/ab-tests/check-and-stop-an-ab-test/).
* Your plan includes A/B tests. If **A/B Tests** is locked in the menu, or **Start** shows a lock, ask your account owner to unlock the feature.
* You accept the trade-off. Fewer visitors see product blocks and banners, and fewer enter your campaigns and receive your newsletters, for as long as the test runs.
* Plan for at least three days. The results leave out the first and the last day of the test, so a test started today shows nothing before its third day.

## Steps

1. Go to **Analytics › A/B Tests**

   Under **Running A/B tests** you see **No active A/B test**. Earlier tests are listed below, with the date they ended.

   ![Step 1](https://releva.ai/documentation/_astro/step-1.CXco7kHY_Z1H2Wxh.webp)

2. Under **Running A/B tests**, select **Start A/ B Test**

   ![Step 2](https://releva.ai/documentation/_astro/step-2.Dm6ofdCX_vawRC.webp)

3. In **Control Group Fraction**, move the slider to the share of visitors who should not get Releva, for example **20%** (the slider shows **2000**)

   The slider runs from **0** to **10000** (the right-hand label can look cut off) in steps of **1000**, where **10000** is 100 %. The default is **5000**, half of your visitors. The arrow keys move it one step at a time. The text below the slider shows how 1000 interactions split for a given value.

   ![Step 3](https://releva.ai/documentation/_astro/step-3.CtINn_Jf_2hhrHp.webp)

4. Under **Hypothesis**, keep **One sided** to test whether Releva does better than the control group

   Choose **Two sided** if you want to know whether Releva does better *or worse*. A two-sided test needs a bigger difference to become significant.

   ![Step 4](https://releva.ai/documentation/_astro/step-4.BxS8Hyf9_Z29IEKj.webp)

5. Select **Start**

   ![Step 5](https://releva.ai/documentation/_astro/step-5.D0BgNnEu_ZFXazg.webp)

6. Read what the test changes and select **Confirm**

   The dialog reminds you that **fewer** users will see product blocks and banners on your website, and **fewer** users will enter your campaigns and receive your newsletters. **Close** goes back without starting anything.

   ![Step 6](https://releva.ai/documentation/_astro/step-6.DZP4_Egu_NxUs9.webp)

7. You are back on **A/B Tests** and the new test is listed under **Running A/B tests** with **Running: YES**

   **Start A/ B Test** is now greyed out and shows a lock, because only one test can run at a time.

   ![Step 7](https://releva.ai/documentation/_astro/step-7.ApPmSjAc_YT1yb.webp)

## Check it worked

* The card under **Running A/B tests** shows **Started at:** today’s date, **Control group fraction:** **20%**, **Hypothesis:** **ONE SIDED** and **Running:** **YES**, with a **Stop** button on the right.
* **Start A/ B Test** is now greyed out. Only one test can run at a time.
* The start date is always the moment you confirm. You cannot backdate a test or schedule it.

## Troubleshooting

**Start A/ B Test is greyed out and shows a lock.** A test is already running. If you can manage the subscription, the page also says “Upgrade your subscription to access this feature.” This message appears whenever the button is disabled, even when your plan includes A/B tests. Stop the running test and the button comes back.

**Start shows a lock with the tooltip “Unlock feature.”, or is disabled.** Your plan does not include starting A/B tests. The lock is shown to users who can manage the subscription; select it to request the feature. Everyone else sees a disabled **Start** and should ask their account owner.

**A red error line appears under the form after Confirm.** The server refused the test. The usual cause is a test that started in another browser tab in the meantime; go back to **A/B Tests** and check **Running A/B tests**. The message shown is a generic “An internal error occurred.”

**I want a different control group fraction.** A running test cannot be edited. Stop it and start a new one with the new value.

**The search box on the list does nothing.** A/B tests have no name to search by. Running and previous tests are always listed in full, newest first.

## Next steps

* [Check and stop an A/B test](https://releva.ai/documentation/guides/ab-tests/check-and-stop-an-ab-test/): Read the status, influence and significance of each indicator, and stop the test when you have your answer.
* [How A/B tests work](https://releva.ai/documentation/learn/ab-tests/how-ab-tests-work/): Learn how visitors land in the control or treatment group and which days and events the results count.
* [How product blocks work](https://releva.ai/documentation/learn/product-blocks/how-product-blocks-work/): Learn what the product blocks your control group does not see do, and what the numbers on their cards count.
* [How campaigns work](https://releva.ai/documentation/learn/campaigns/how-campaigns-work/): Learn how campaigns reach your segments, the other part of what your test measures.