Targeting Rules

Targeting rules let you control which users, organizations, or contexts receive a specific flag variation. Flagmint evaluates rules in priority order — the first rule that matches determines the result. If no rules match, the flag’s fallback value is returned.

How Evaluation Works

When a flag is evaluated, Flagmint follows this sequence:

Flatten the context — The nested evaluation context (user, organization, custom) is flattened into a simple key-value map for attribute matching.
Evaluate targeting rules in order — Rules are processed by order_index, lowest first. The first matching rule wins.
Serve the matched rule’s result — A matched rule can return either a specific variation or a rollout result.
Fall back — If no rules match, the flag’s fallback value is returned, coerced to the expected type.

For flags without targeting rules, the evaluation is simpler: if a rollout is configured, it is applied directly. Otherwise, the flag’s base value is returned.

Context arrives → Flatten context
  → Has targeting rules?
    → Yes → Evaluate rules in order_index order
      → Rule matches?
        → Has variation_id? → Return that variation's value
        → Has rollout_id?   → Apply rollout strategy → Return result
        → Neither?          → Return fallback (misconfigured rule)
      → No match → Next rule
    → All rules checked, none matched → Return fallback value
  → No targeting rules?
    → Has rollout? → Apply rollout → Return result
    → No rollout  → Return base value

Context Flattening

Before rules are evaluated, the nested evaluation context is flattened into a simple key-value map. This is how attribute paths in rules map to context values. Nested context:

{
  "kind": "multi",
  "user": {
    "kind": "user",
    "key": "user123",
    "email": "alice@example.com",
    "plan": "premium"
  },
  "organization": {
    "kind": "organization",
    "key": "org456",
    "country": "DE",
    "employeeCount": 50
  },
  "custom": {
    "source": "mobile_app"
  }
}

Flattened result:

Attribute	Value
`user.key`	`"user123"`
`user.email`	`"alice@example.com"`
`user.plan`	`"premium"`
`organization.key`	`"org456"`
`organization.country`	`"DE"`
`organization.employeeCount`	`50`
`custom.source`	`"mobile_app"`

Rule attributes like user.plan or organization.country reference keys in this flattened map.

Rule Types

Flagmint supports two kinds of targeting rules: segment rules and custom rules. Each rule can optionally reference a variation (return a specific value) or a rollout (apply a rollout strategy).

Segment Rules

Segment rules reference a predefined, reusable segment by ID. A segment contains its own set of conditions and a logical operator that determines how those conditions are combined.

{
  "kind": "segment",
  "segment_id": "beta_users",
  "order_index": 0,
  "variation_id": "var_enabled"
}

Segment definition:

{
  "id": "beta_users",
  "name": "Beta Users",
  "logical_op": "AND",
  "rules": [
    {
      "attribute": "user.plan",
      "operator": "in",
      "value": ["premium", "growth"]
    },
    {
      "attribute": "organization.country",
      "operator": "in",
      "value": ["DE", "NL", "FR", "IE"]
    }
  ]
}

Segment Logical Operators

Segments support three logical operators that control how their conditions are combined:

Operator	Behavior	Description
`AND` (default)	All conditions must match	User must match every condition
`OR`	Any condition can match	User matches if at least one condition is true
`NOT`	No conditions should match	User matches only if all conditions are false

Example — OR segment (any EU country):

{
  "id": "eu_users",
  "name": "EU Users",
  "logical_op": "OR",
  "rules": [
    { "attribute": "organization.country", "operator": "eq", "value": "DE" },
    { "attribute": "organization.country", "operator": "eq", "value": "FR" },
    { "attribute": "organization.country", "operator": "eq", "value": "NL" }
  ]
}

A segment with force: true always matches regardless of its conditions. This is useful for testing or emergency overrides.

Custom Rules

Custom rules define conditions inline on the targeting rule itself, without referencing a reusable segment. Like segments, custom rules support logical operators.

{
  "kind": "custom",
  "order_index": 1,
  "logical_op": "AND",
  "conditions": [
    {
      "attribute": "user.plan",
      "operator": "eq",
      "value": "enterprise"
    },
    {
      "attribute": "organization.employeeCount",
      "operator": "gt",
      "value": 100
    }
  ],
  "variation_id": "var_enterprise_feature"
}

Custom rules support the same AND, OR, and NOT logical operators as segments.

Operators

Conditions within segments and custom rules use operators to compare attribute values. All string comparisons are case-sensitive.

Operator	Description	Value Type	Example
`eq`	Equals	Any	`user.plan` eq `"premium"`
`neq`	Not equals	Any	`user.plan` neq `"free"`
`in`	Is in list	Array	`organization.country` in `["DE", "FR", "NL"]`
`nin`	Not in list	Array	`organization.country` nin `["US", "CN"]`
`gt`	Greater than	Number	`organization.employeeCount` gt `10`
`lt`	Less than	Number	`organization.employeeCount` lt `100`
`contains`	String contains substring	String	`user.email` contains `"@acme.com"`
`not_contains`	String does not contain substring	String	`user.email` not_contains `"@test.com"`
`startsWith`	String starts with prefix	String	`user.email` startsWith `"admin"`
`endsWith`	String ends with suffix	String	`user.email` endsWith `"@flagmint.com"`
`exists`	Attribute is present	—	`user.email` exists
`not_exists`	Attribute is absent	—	`user.phone` not_exists

The gt and lt operators coerce both sides to numbers. If either value cannot be parsed as a number, the condition does not match. The eq and neq operators include special handling for boolean and numeric string comparisons — for example, the string "true" will match a boolean true attribute.

Rule Ordering and First-Match Wins

Targeting rules are evaluated in order_index order (lowest first). The first rule that matches determines the result — subsequent rules are not evaluated. This means rule order matters. Place more specific rules (like individual user overrides) before broader rules (like segment-based rollouts). Example — override for a specific user before a general rollout:

{
  "targeting_rules": [
    {
      "kind": "custom",
      "order_index": 0,
      "conditions": [
        { "attribute": "user.key", "operator": "eq", "value": "user_alice" }
      ],
      "variation_id": "var_enabled"
    },
    {
      "kind": "segment",
      "order_index": 1,
      "segment_id": "eu_premium",
      "rollout_id": "rollout_50_percent"
    }
  ]
}

Alice always gets the enabled variation. Everyone else matching the eu_premium segment goes through the 50% rollout.

Rule Outcomes

When a rule matches, it must specify what to return. There are two options:

Variation

A variation_id returns a specific, predefined value. Variations are configured per flag and can be of any type (boolean, string, number, JSON).

{
  "kind": "custom",
  "order_index": 0,
  "conditions": [{ "attribute": "user.plan", "operator": "eq", "value": "enterprise" }],
  "variation_id": "var_premium_feature"
}

Rollout

A rollout_id references a rollout configuration, which dynamically determines the value based on the user’s hash bucket. See Rollout Strategies below.

{
  "kind": "segment",
  "order_index": 1,
  "segment_id": "all_users",
  "rollout_id": "rollout_ab_test"
}

Rollout Strategies

Flagmint supports four rollout strategies: off, percentage, variant, and gradual.

Off

The off strategy explicitly disables the rollout and returns the flag’s fallback value. Useful for pausing a rollout without deleting the configuration.

{
  "strategy": "off"
}

Percentage

Percentage rollouts are boolean-only. They determine whether a user is “in” (feature enabled) or “out” (fallback value). This is the simplest way to progressively enable a feature.

{
  "strategy": "percentage",
  "percentage": 25,
  "salt": "experiment-2026-q2"
}

Users whose hash bucket falls below the percentage get true
Users above the percentage get the flag’s fallback value
A percentage of 100 short-circuits and returns true for all targeted users

Percentage rollouts only work with boolean flags. If applied to a string, number, or JSON flag, the evaluator logs a warning and returns the fallback value. Use variant rollouts for non-boolean experiments.

Variant

Variant rollouts split users into buckets, each receiving a different variation value. This is how you run A/B tests and multi-variant experiments. Variant rollouts work with any flag type (boolean, string, number, JSON).

{
  "strategy": "variant",
  "salt": "checkout-experiment",
  "variants": [
    { "variation_id": "var_control", "weight": 50 },
    { "variation_id": "var_redesign", "weight": 30 },
    { "variation_id": "var_minimal", "weight": 20 }
  ]
}

Weights represent percentages and must sum to exactly 100. This is validated at configuration time — if weights don’t sum to 100, the configuration is rejected. How variant assignment works:

The user’s stable key is combined with the salt
A hash produces a deterministic bucket (0–99)
The bucket maps to a variant based on cumulative weights

Using the example above: users hashing to 0–49 get var_control, 50–79 get var_redesign, and 80–99 get var_minimal.

Variant rollouts reference variation_id values, not raw values. The variation’s configured value and type are looked up and coerced to the flag’s expected type. If a variant references a missing variation, the fallback value is returned.

Gradual

Gradual rollouts progressively increase the percentage of users who see a feature over time. Like percentage rollouts, they are boolean-only.

{
  "strategy": "gradual",
  "salt": "new-dashboard-rollout",
  "target_percentage": 100,
  "increment": 10,
  "interval_hours": 24,
  "start_at": "2026-04-01T00:00:00Z"
}

This configuration starts at 0% and adds 10% every 24 hours until reaching 100%:

Day	Percentage
Day 0 (Apr 1)	0%
Day 1 (Apr 2)	10%
Day 2 (Apr 3)	20%
…	…
Day 10 (Apr 11)	100%
Day 11+	100% (capped at target)

Gradual rollouts are computed based on elapsed time from start_at. If start_at is in the future, the rollout returns 0% until that time. If start_at is omitted, the rollout starts immediately.

Validation rules for gradual rollouts:

target_percentage must be between 0 and 100
increment must be between 0 and 100
interval_hours must be greater than 0
start_at (if provided) must be a valid ISO 8601 date-time string

Consistent Hashing

All rollout strategies that involve user bucketing (percentage, variant, gradual) use consistent hashing to ensure stable assignment. The same user always receives the same rollout result as long as the salt remains unchanged. The stable key used for hashing is resolved in this order:

user.key (from kind: "user" or kind: "multi" with a user block)
user_id (legacy flat context)

If no stable key is found, the rollout returns the fallback value with a warning. Changing the salt redistributes all users. This is useful when you want to re-randomize an experiment without changing the targeting rules.

Type Coercion

All flag values are coerced to the flag’s expected type before being returned. This ensures consistent types regardless of how the value is stored.

Expected Type	Coercion Behavior
`boolean`	`"true"` → `true`, `"false"` → `false`, otherwise `Boolean(value)`
`number`	`Number(value)`, returns `0` if not a valid number
`string`	`String(value)`, returns `""` for null/undefined
`json`	Must be a non-null, non-array object, otherwise returns `{}`

Examples

Feature gate for enterprise users

{
  "targeting_rules": [
    {
      "kind": "custom",
      "order_index": 0,
      "conditions": [
        { "attribute": "user.plan", "operator": "eq", "value": "enterprise" }
      ],
      "variation_id": "var_enabled"
    }
  ]
}

Enterprise users get the enabled variation. Everyone else gets the fallback.

Gradual rollout to all users

{
  "targeting_rules": [],
  "rollout": {
    "strategy": "gradual",
    "salt": "new-dashboard",
    "target_percentage": 100,
    "increment": 10,
    "interval_hours": 24,
    "start_at": "2026-04-15T00:00:00Z"
  }
}

No targeting rules means all users are eligible. The gradual rollout increases coverage by 10% per day.

A/B test for a segment with user override

{
  "targeting_rules": [
    {
      "kind": "custom",
      "order_index": 0,
      "conditions": [
        { "attribute": "user.key", "operator": "eq", "value": "user_qa_lead" }
      ],
      "variation_id": "var_redesign"
    },
    {
      "kind": "segment",
      "order_index": 1,
      "segment_id": "eu_enterprise",
      "rollout_id": "rollout_pricing_ab"
    }
  ]
}

The QA lead always sees the redesign. All other EU enterprise users are split by the rollout.

Kill switch

A flag with no targeting rules, no rollout, and is_active = true acts as a simple on/off toggle. The base value is returned for every evaluation. Set is_active = false or use a rollout with strategy: "off" to disable it.

Email domain targeting

{
  "kind": "custom",
  "order_index": 0,
  "conditions": [
    { "attribute": "user.email", "operator": "endsWith", "value": "@acme.com" }
  ],
  "variation_id": "var_internal_feature"
}

Only users with @acme.com email addresses see the feature.

Getting Started

Concepts & Features

Billing & Plans

Security & Compliance

Resources

Targeting Rules

Targeting Rules

How Evaluation Works

Context Flattening

Rule Types

Segment Rules

Segment Logical Operators

Custom Rules

Operators

Rule Ordering and First-Match Wins

Rule Outcomes

Variation

Rollout

Rollout Strategies

Off

Percentage

Variant

Gradual

Consistent Hashing

Type Coercion

Examples

Feature gate for enterprise users

Gradual rollout to all users

A/B test for a segment with user override

Kill switch

Email domain targeting

Getting Started

Concepts & Features

Billing & Plans

Security & Compliance

Resources

​Targeting Rules

​How Evaluation Works

​Context Flattening

​Rule Types

​Segment Rules

​Segment Logical Operators

​Custom Rules

​Operators

​Rule Ordering and First-Match Wins

​Rule Outcomes

​Variation

​Rollout

​Rollout Strategies

​Off

​Percentage

​Variant

​Gradual

​Consistent Hashing

​Type Coercion

​Examples

​Feature gate for enterprise users

​Gradual rollout to all users

​A/B test for a segment with user override

​Kill switch

​Email domain targeting

Targeting Rules

How Evaluation Works

Context Flattening

Rule Types

Segment Rules

Segment Logical Operators

Custom Rules

Operators

Rule Ordering and First-Match Wins

Rule Outcomes

Variation

Rollout

Rollout Strategies

Off

Percentage

Variant

Gradual

Consistent Hashing

Type Coercion

Examples

Feature gate for enterprise users

Gradual rollout to all users

A/B test for a segment with user override

Kill switch

Email domain targeting