# Targeting rules

> Attributes, operators, and how rule groups combine.

Targeting rules turn a flag on for **specific users** based on attributes you pass
at evaluation time — plan, country, email, internal team, anything. A user who
matches a rule gets the enabled value; everyone else gets the default.

## The user context

You describe the current user with a plain object of attributes:

```python
user = {
    "user_id": "42",
    "email": "alice@company.com",
    "plan": "enterprise",
    "country": "US",
}

client.enabled("advanced_analytics", user=user)
```

The keys are entirely up to you — they just have to match the attribute names you
target on in the dashboard. `user_id` (or `id`) is special: it's the input to
[percentage rollouts](/docs/concepts/rollouts).

## Conditions

A **condition** tests one attribute with one operator:

> `plan` **equals** `enterprise`

The operators are `equals`, `not_equals`, `contains`, `ends_with`, `in_list`,
`gt`, and `lt`. Each is defined with an example in
[Rule operators](/docs/reference/operators).

## How conditions combine: AND within a group, OR across groups

Targeting is built from **rule groups**. The logic is two levels:

- **Within a group**, all conditions must be true — they're **AND**ed.
- **Across groups**, any group matching is enough — they're **OR**ed.

So this set of groups:

```text
Group 1:  country equals "US"  AND  device equals "mobile"
Group 2:  plan equals "enterprise"
```

means *(country = US **and** device = mobile) **or** (plan = enterprise)*. A user
matches if they satisfy **either** group in full.

This "OR of ANDs" shape (formally, disjunctive normal form) can express any
combination of conditions. A group with no conditions never matches.

## Segments: define an audience once

If you find yourself repeating the same conditions across many flags — "internal
staff", "EU users" — define them once as a **segment** and attach the segment to a
rule group. Segments are project-scoped and reusable; editing a segment updates
every flag that references it. They're expanded inline when the config is
published, so the SDK only ever sees plain conditions. See
[Rules & rollouts in the UI](/docs/dashboard/rules).

## Targeting and rollouts together

A flag can have both rules and a rollout. Rules are checked first: a user who
matches any rule gets the enabled value regardless of the rollout percentage. Only
users who match no rule are then subject to the rollout. The full precedence is on
the [Evaluation order](/docs/reference/evaluation-order) page.

## Next

- [Rule operators](/docs/reference/operators) — every operator, with examples.
- [Percentage rollouts](/docs/concepts/rollouts) — release to a fraction of users.