Source: https://docs.amadalis.ai/insights/measures-and-retention/

# Measures and returning visitor groups

Use **Insights** > **Analytics** in the site admin to define what a report counts and compare groups of visitors who returned later.

Checked against the product’s code on 7 November 2026.

Open **Insights** > **Analytics** and choose the **[Measures](https://docs.amadalis.ai/reference/glossary/#measure)** tab. It lists the measures, dimensions and [goals](https://docs.amadalis.ai/reference/glossary/#goal) available to reporting. A display name alone does not tell you the definition or its version.

## Build a compound measure

1.  Choose **New measure** to open the studio.
2.  Choose **Journey**, **Ratio** or **Capped count**.
3.  Enter **Display name**.
4.  Review **Stable key**.
5.  Fill in **What this measures**.

The studio builds from supported catalog objects with saved versions. For a journey, choose events the site actually records; a ratio needs existing measures to use as its two sides.

An ordered journey counts a defined sequence. A ratio divides one eligible measure, the numerator, by another, the denominator. A capped count limits how much one visitor can contribute. These count different groups and should not be compared as if they were the same simple event total.

For **Journey**:

1.  Choose the event for each journey step.
2.  Choose **Order semantics**.
3.  Set **Convert within**.
4.  Set **Exclude visitors who show** only if the journey needs that exclusion.

For **Ratio**:

1.  Choose **Numerator**, the measure to divide.
2.  Choose **Denominator**, the measure to divide by.

For **Capped count**:

1.  Choose **Event to count per visitor**.
2.  Choose whether to **Cap extreme visitor values**.
3.  If the cap is on, review **Upper percentile**.

Then save the definition:

1.  Choose **Review definition**.
2.  Read the proposed definition.
3.  Choose **Save metric** when it matches your question.
4.  Read its receipt.
5.  Check its entry in **Compound registry**.

A saved metric does not by itself lock it into an active experiment. Experiment planning has its own version and approval rules.

## Read returning visitor groups

This retention means returning activity.

1.  Open **Reports** in Analytics.
2.  Open **Do visitors come back?**.
3.  Choose **Day**, **Week** or **Month**.
4.  Read each group’s size and interval together.

The report groups people by the period they first arrived and shows how many came back in each later interval. Each group is called a cohort. Recent groups have not had time to fill every future cell.

It is separate from the site’s data deletion policy. Erased visitor records leave the cohort calculations. The available retained events determine what can be counted.

For deletion policy and archive coverage, read [Privacy and consent](https://docs.amadalis.ai/trust/privacy-and-consent/#retention). For experiment use, read [Experiments](https://docs.amadalis.ai/optimization/experiments/).