> For the complete documentation index, see [llms.txt](https://docs.roivenue.com/roivenue-resources/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.roivenue.com/roivenue-resources/roivenue-training/roivenue-app/umm/experiments.md).

# Experiments

Add results of real-world tests, such as lift or geo tests, and use them to calibrate your MTA or UMM models.

Experiments let you add the measured result of a real-world test to Roivenue. A model combined with an experiment is closer to the real impact of your marketing, so its results are more reliable.

Use Experiments when you want to:

* record the result of a test for a specific channel, platform or campaign
* calibrate a UMM model with a test result
* create a version of an MTA model that includes a test result

The page has 2 parts:

1. **Experiments list:** your experiments with **Name**, **Site**, **Period** and **Attribution Model**, and **Add Experiment**.
2. **Experiment form:** **Experiment Settings** and **Assign to MTA & UMM Models**.

To open the page, go to **UMM → Experiments** in the left navigation.

***

### How to add an experiment

1. Click **Add Experiment**.
2. Fill in **Experiment Settings**:

| Setting                          | What to fill in                                                                                                                                         |
| -------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Name of the experiment**       | A unique name.                                                                                                                                          |
| **Site**                         | The site where the experiment ran.                                                                                                                      |
| **Period**                       | **From** and **To** of the experiment. **To** must be at least one day after **From**.                                                                  |
| **Assign to Specific Dimension** | The rule that selects what the experiment tested, for example **Platform Name** equals "Facebook Business". Combine conditions with **And** and **Or**. |
| **Result of experiment**         | The measured **Value** (a positive number) and its **Metric**: **Conversions** or **Net Revenue**.                                                      |

3. Under **Assign to MTA & UMM Models**, click **Add Another Model**, select the model and set the weight between the experiment and the model.
4. Click **Save**.

***

### How to choose between an MTA and a UMM model

| You assign the experiment to | Result                                                                                                        |
| ---------------------------- | ------------------------------------------------------------------------------------------------------------- |
| An MTA model                 | Roivenue creates a version of the MTA model that includes the experiment. This version cannot be used in UMM. |
| A UMM model                  | The experiment is used in the UMM calculation.                                                                |

To use an experiment in UMM, connect it directly to a UMM model. You can do it here or in the site configuration of the model in [UMM Models](https://docs.roivenue.com/roivenue-resources/roivenue-training/roivenue-app/umm/umm-models).

The model must use the same site as the experiment. If the model takes its totals from MMM, its mixing metric must match the metric of the experiment.

{% hint style="info" %}
**Tip:** The weight decides how much the experiment influences the final result compared with the model. Use a higher weight for experiments you trust more, for example tests with a clean setup and enough volume.
{% endhint %}

***

#### FAQ

<details>

<summary>1. What kind of experiment can I add?</summary>

Any test that measured the effect of a marketing activity on conversions or net revenue for a site and period, for example a lift test or a geo test.

</details>

<details>

<summary>2. Why can't I select a model?</summary>

The model uses a different site, or its mixing metric does not match the metric of the experiment. Change the model or the experiment settings.

</details>

<details>

<summary>3. What does "Experiment Configuration Error" mean?</summary>

The model the experiment is assigned to has no configuration for the experiment's site or metric. Add the configuration in [UMM Models](https://docs.roivenue.com/roivenue-resources/roivenue-training/roivenue-app/umm/umm-models), or assign the experiment to another model.

</details>

<details>

<summary>4. Can I use an experiment assigned to an MTA model in UMM?</summary>

No. The MTA version with the experiment cannot be used in UMM. Connect the experiment directly to a UMM model.

</details>

<details>

<summary>5. When will the experiment affect the results?</summary>

After you save, the model recalculates. This usually takes a few hours.

</details>

<details>

<summary>6. Do I need MMM data to use experiments?</summary>

Not for MTA models. For UMM models, you need the full UMM setup. See [UMM](https://docs.roivenue.com/roivenue-resources/roivenue-training/roivenue-app/umm).

</details>
