---
metadata:
- name: generator
content: Diplodoc Platform v5.52.0
alternate:
- https://yandex.com/support/partner/en/web/units/experiments.md
- https://yandex.com/support/partner/ru/web/units/experiments.md
- href: en/web/units/experiments.md
type: text/markdown
title: Markdown version
- href: ../../llms.txt
type: text/markdown
title: llms.txt
---
> **Documentation Index:** Fetch the complete configuration index at https://yandex.com/support/partner/en/llms.txt
# A/B experiments
The Yandex Advertising Network interface has a dedicated split testing technology called [Varioqub](https://yandex.com/support2/varioqub/en/).
You can use it to run A/B experiments with your ad units to:
1. Test out new [overlay units](*overlay-block). Add [Fullscreen](https://yandex.com/support/partner/en/web/units/types/fullscreen.md), [Floor Ad](https://yandex.com/support/partner/en/web/units/types/floor-ad.md), and [Top Ad](https://yandex.com/support/partner/en/web/units/types/top-ad.md) units and see how they work on your site.
1. Compare ad units with different parameters to determine which are the most effective. For example, you can compare banners with different CPM floors and blocked topics.
{% note info %}
To test out different ad unit designs, use the [design auction](https://yandex.com/support/partner/en/web/units/profit.md).
{% endnote %}
## Creating an experiment {#create-experiment}
1. [Add](https://yandex.com/support/metrica/quick-start.html) a Yandex Metrica tag to your site. It collects the data necessary for running experiments.
1. [Connect](https://yandex.com/support2/varioqub/en/connect) Varioqub.
1. In the Yandex Advertising Network interface, go to **Website Ads** → **Sites** and open the settings of the desired site.
1. Go to **A/B experiments** → **To create**.
1. Add the experiment conditions:
- **Title**: Experiment name.
- **Counter**: Yandex Metrica tag ID.
- **Experiment Application Pages**: Site domain and URL without the protocol (`HTTP`/`HTTPS`) and `hash` (the `#` character and the value that follows it in the URL).
For example, if you want your experiment to include only the site's home page, write it as `example.com`. For information on adding more flexible conditions, see the [Varioqub Help](https://yandex.com/support2/varioqub/en/create/conditions#pages).
- **Audience share**: Percentage of the site's audience (from 1% to 100%) that will participate in the experiment. The share is evenly divided between the samples. By default, experiments cover 100% of the site's audience.
- **Period**: Experiment duration, from 14 to 90 days.
1. Set up ad unit versions:
1. If you have a reference ad unit you want to compare other units to, select it from the drop-down list in the **The control option** field. During the experiment, the control unit is replaced with ad units from your experimental versions.
If you don't need a control version, leave the **Without a block** value to measure effectiveness with and without different ad unit versions.
1. To add an experimental ad unit, enter the unit ID in the **Option 1** field or select it from the list.
To add multiple experimental versions, click **Add an option** and select an ad unit. You can choose up to five versions: they will alternate over the course of the experiment.
To open an ad unit's settings, click **View settings** to the right of the ad unit selection field. You can only change the settings when the experiment isn't running.
1. Click **Save and Run**.
1. If you used a control ad unit, replace the ad tag. Copy the new ad tag that will appear after you launch the experiment and add it to your site.
{% note info %}
If you used the `clientFeatures` argument or modified its parameters when connecting Varioqub, do the same for the ad tag of the control unit.
{% endnote %}
You can view the ad tag by clicking **The code for the site** on the experiment page.
1. Click **Complete**.
## Viewing the experiment results {#view-experiment}
To view the experiment report, go to the **A/B experiments** tab, select the experiment, and click **Extended Report**. For more information on the report objectives, structure, and settings, see the [Varioqub Help](https://yandex.com/support2/varioqub/en/stat).
Brief statistics are also available under **Statistics** on the experiment page:
- **Option**: Ad unit version.
- **Block**: Ad unit name and ID.
- **Time on the site**: How much time the user spent on the site.
- **Bounce rate**: Percentage of visitors who [bounced](https://yandex.ru/support/metrica/general/glossary.html#glossary__failure) from the site.
- **Remuneration on the website**: Revenue from ad impressions.
## Stopping an experiment {#stop-experiment}
The experiment stops automatically when it reaches the end of the specified period. You can also stop it manually at any time:
1. Select the experiment in the **A/B experiments** tab.
1. At the top right, click **Stop** → **Complete**.
{% note warning %}
Once an experiment is stopped, you can't resume it.
{% endnote %}
## Extending an experiment {#extension-experiment}
You can extend an experiment after it reaches the end of the specified period.
1. Select the experiment in the **A/B experiments** tab.
1. Change the end date and click **Save** → **Extend**.
## Applying the experiment results
1. In the **A/B experiments** tab, click **How to apply the result** next to the desired experiment.
1. Copy the code and add it to your site between the `` and `` tags at the location where the ad unit should be displayed. The ad unit will be displayed to all site visitors.
If you don't want to apply the result, you don't need to do anything.
## Contact support
Write to WhatsApp
Go to WhatsApp
Write to Telegram
Go to Telegram
Write to chat
Send an email
You can also go to
[](https://partner.yandex.com)
[*overlay-block]: Ad units that overlap page content.