> For the complete documentation index, see [llms.txt](https://docs.mediarithmics.io/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.mediarithmics.io/audience-and-activation/build-your-audiences/segment-reference/segment-types/cohort-lookalike-segment.md).
# Cohort lookalike segment
The lookalike type built by expanding a seed segment through cohorts, and the similarity index that controls how far it reaches.
## What it is
A **lookalike** segment holds users that share the characteristics of the users in another segment - the **seed segment**. It is how you increase the reach of an audience using only your own first-party data.
You can build a lookalike from any persisted segment that is not itself a lookalike.
Two flavours exist:
* **Cohort-based**, described on this page.
* **Score-based**, which lets you integrate your own algorithm - a score, a gender prediction, a socio-demographic model. Contact your Account manager to learn more about it.
{% hint style="warning" %}
Cohort-based Lookalike is not enabled by default. Get in touch with your Account manager for information on enabling it, and on configuring your cohort model.
{% endhint %}
The cohort model itself is configured once per organisation, on the datamart, before any lookalike segment can be built on it: see [Affinity & Intention](https://app.gitbook.com/s/XSPACE_SETUP/plan-and-integrate/set-up-advanced-features/affinity-and-intention-ml-algorithm#set-up-the-cohort-based-lookalike-model) in the `Setup & Monitor` section.
## What a cohort is
A **cohort** is a group of users sharing common traits. The traits can be defined by you, by mediarithmics, or both, from information on the user's profile and on events. Every user in your datamart is automatically assigned one, and only one, cohort.
A user's cohort is updated frequently - not only when a new activity is ingested live or a file is imported, but also for inactive users - so your datamart stays current.
{% hint style="info" %}
A cohort is an **unlabelled** group produced by an unsupervised clustering model. It is not a segment you named, and you should not expect it to carry a business meaning.
{% endhint %}
## How the segment is built
{% stepper %}
{% step %}
### Cohorts are assigned to users
Every user in the datamart is assigned a cohort. This happens once, during feature setup, and is updated automatically whenever a user changes.
{% endstep %}
{% step %}
### The seed / cohort overlap is calculated
For the seed segment you want to expand, the platform calculates how much of each cohort falls inside that segment, which identifies the cohorts most represented in it. This happens in real time, while you configure the lookalike.
{% endstep %}
{% step %}
### You configure the lookalike
You set the size of the lookalike segment through the minimum similarity index, and choose whether to include or exclude the seed segment's own users.
{% endstep %}
{% step %}
### The segment refreshes daily
Every day the lookalike is refreshed: new users are added, users no longer in the seed segment are removed, and users who switched cohorts are accounted for.
{% endstep %}
{% endstepper %}
## The similarity index
The **similarity index** is calculated for every cohort as:
```
overlap of users in the cohort within the seed segment
-----------------------------------------------------------
(users in the segment / users in the datamart)
```
In other words, it is the cohort's over-representation in the seed segment relative to the datamart. You set a **minimum** similarity index with the slider under the curve, and the platform shows in real time how many users the lookalike is expected to contain.
{% hint style="info" %}
**Worked through.** Set a minimum similarity index of **8.6** with the seed segment set to **include**: any user in a cohort scoring at least 8.6 against the seed segment is eligible, the expected size is **2,415,900 users**, and every user from the seed segment is in the result as well.
{% endhint %}
The final count can differ from the expectation, for two reasons: the seed segment itself varies, and cohorts are calculated live, so the number of users in a cohort moves between the segment update and the segment calculation.
## What happens at each update
Before each segment calculation, the platform rebuilds the list of cohorts used to populate the lookalike, from the settings you saved. It tries to **reach the expected number of users without going below the minimum similarity index** - adding cohorts largest-similarity-index first.
Following the example above, it adds cohorts until it reaches 2,415,900 users, considering only cohorts at 8.6 or above. Two outcomes are possible:
| Outcome | What you get |
| ------------------------------------------------------------ | ----------------------------------------------------------------------- |
| The target is reached with cohorts at or above the minimum | The lookalike is at least the expected size. |
| The target cannot be reached without going below the minimum | The lookalike is smaller than expected, and its precision is preserved. |
The minimum similarity index is therefore a floor on quality, not a target size.
## Creating one
{% hint style="warning" %}
**Before you start.** Wait for the seed segment's calculation to finish, and check that its volume is stable - minimal fluctuation over the last few days - before expanding it.
{% endhint %}
{% stepper %}
{% step %}
### Open the seed segment
Go to **Audience > Segments** and click the segment you want to expand.
{% endstep %}
{% step %}
### Start the lookalike
Click **Create Lookalike**, then select **Cohort based Lookalike**.
{% endstep %}
{% step %}
### Name it
Fill in the **Name**, and a **Description** if you want one.
{% endstep %}
{% step %}
### Calibrate it
In the **Lookalike calibration** panel, move the slider to set the minimum similarity index, and choose whether to **include** or **exclude** the seed segment's users. Click **Save those settings**.
{% endstep %}
{% endstepper %}
Like any other segment, it is populated at the next segment computation.
{% hint style="danger" %}
Lookalike segments are among the three types that cannot carry a processing activity. See [Enforce compliance on your audience](/audience-and-activation/build-your-audiences/enforce-compliance-on-your-audience.md).
{% endhint %}
## Related pages
Expand your audience with Lookalike
Deciding whether to expand, and judging the result.
---
# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.
## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation by asking a question.
Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters:
```
GET https://docs.mediarithmics.io/audience-and-activation/build-your-audiences/segment-reference/segment-types/cohort-lookalike-segment.md?ask=&goal=
```
`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.
Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.