Segments
A segment is a subset of identities, defined by a set of rules that match identity traits. An identity always belongs to a single environment and can belong to any number of segments.
Once you have defined a segment, you can create segment overrides for features within an environment. A segment override allows you to control the state of a feature only for identities that belong to a specific segment. This is similar to how identity overrides let you control the state of features for an explicit set of identities that is known in advance.
Because segments are driven by identity traits, your application must identify the user when retrieving flags in order for segment overrides to be applied. If your user is not identified, no overrides will be applied and all flags will be returned exactly how they are defined in the current environment.
Segments and segment overrides can be used to implement many scenarios. For example:
- Test features in production before they are released by overriding them only for internal users or a QA team
- Deliver features only to "power users" who have logged in a certain number of times, have used specific functionality within your application, or any combination of factors
- Force a group of users into a specific A/B test variation by overriding weightings on multivariate flags
- Override behaviour based on the application version number, e.g. by using the SemVer rule operators
- Control features based on the time of day, date, weekday, etc. by passing it as a trait when evaluating flags for an identity
Security and privacy
The Flagsmith API to set user traits, e.g. the setTraits
method from the JavaScript SDK, does not require
authentication or credentials. This means that users can change their own traits, which could be a security problem if
you are using segments for authorisation or access control. If you must use segments for access control, make sure to
disable the
"Allow client SDKs to set user traits" option
on every environment that needs it, and use server-side SDKs to set traits instead. You can still use client-side SDKs
to read traits and flags derived from segments in this case.
Segment names and definitions might include sensitive or proprietary information that you do not want to expose to your users. Because of this, segments are transparent to applications and are not included in API responses when using remote evaluation mode.
Segment definitions are served to clients running in local evaluation mode, as this allows them to calculate segments without making requests to the Flagsmith API. This is only an implementation detail and no segment information is exposed when retrieving flags using any SDK method.
Creating project or feature-specific segments
Segments created from the Segments page of the Flagsmith dashboard can be used to override any feature within a single project.
To create a segment override, click on a feature in a specific environment and go to the Segment Overrides tab.
If you need to create a segment that will only ever be used to override a single feature, you can create a feature-specific segment by clicking on "Create Feature-Specific Segment" when creating a segment override. Feature-specific segments are otherwise functionally identical to project segments. By default, feature-specific segments are not shown in the Segments page, unless you enable the "Include Feature-Specific" option.
Once created, project segments cannot be changed into feature-specific segments or vice versa.
Order of rules within segments
Segment rules are evaluated in order, i.e. from top to bottom when viewed in the Flagsmith dashboard.
For example, consider the following segment:
- 10% percentage split
is_subscriber = true
This segment would first select 10% of all identities, and then choose subscribers from that cohort. Instead, if we used the opposite order:
is_subscriber = true
- 10% percentage split
This would first select all subscriber identities, and then randomly choose 10% of them.
Multiple segment overrides for one feature
If a feature has multiple segment overrides, they are evaluated in order, i.e. from top to bottom when viewed in the Flagsmith dashboard. The first matching override will be used to determine the state of a feature for a given identity.
Flag evaluation precedence
Identity overrides always take precedence over segment overrides. Simply put, the order of precedence when evaluating a flag is:
- Identity overrides
- Segment overrides
- Default value for the current environment
Trait data types
Each individual trait value is always stored as one of the following data types:
- String
- Boolean
- Integer
- Float
Values in segment rules, on the other hand, are always stored as strings. When segment rules are evaluated, rule values will be coerced to be the same type as the trait value. If the rule value cannot be coerced, that rule will evaluate as false. This provides some flexibility if you ever need to change the data type of a trait, e.g. from boolean to string, while maintaining backwards and forwards compatibility in your application.
For example, consider the following code using the JavaScript SDK:
flagsmith.identify('example_user_1234');
flagsmith.setTrait('accepted_cookies', true);
The value of the accepted_cookies
trait will be stored as a boolean for this identity. If you define a segment rule
like accepted_cookies = true
, the rule value true
is stored as a string. Because the accepted_cookies
was stored
as a boolean for this identity, the segment engine will coerce the rule's string value into a boolean, and things will
work as expected.
Suppose later on you needed to store a third possible state for the trait accepted_cookies
, for example if users can
partially accept cookies. Your application can start storing this trait as a string without needing to modify your
existing segment:
flagsmith.setTrait('accepted_cookies', 'partial');
This would continue to work as expected for identities that already have this trait set as a string value. Always storing the trait as a string would also work, for example:
flagsmith.setTrait('accepted_cookies', 'true');
The following string trait values will evaluate to true
:
"True"
"true"
"1"
Rule operators reference
All rule operators are case-sensitive.
Name | Description |
---|---|
Exactly Matches (=) | Trait value is equal to the rule value |
Does not match (!=) | Trait value is not equal to the rule value |
% Split | Identity is in the percentage bucket. Learn more |
> | Trait value is greater than the rule value |
>= | Trait value is greater than or equal to the rule value |
< | Trait value is less than the rule value |
<= | Trait value is less than or equal to the rule value |
In | Trait value is equal to any element in a comma-separated list (case-sensitive). Learn more |
Contains | Rule value is a substring of the trait value |
Does not contain | Rule value is not a substring of the trait value |
Matches regex | Trait value matches the given regular expression |
Is set | Trait value is set for given identity and trait key |
Is not set | Trait value is not set for given identity and trait key |
SemVer | Trait value is compared against the rule value according to Semantic Versioning. Learn more |
Operator details
- In
- SemVer
- Percentage Split
- Modulo
The In
operator lets you match a trait value against a comma-separated list of values. For example, the segment rule
value might read 21,682,8345
. This would match against a trait value of 682
but not against a trait value of 683
or 834
.
The In
operator can be useful to build segments that represent a specific set of tenants in your application. For
example, you could create a segment with the following rule: tenant_id In tenant_1,tenant_2,tenant_3
SemVer operators compare semantic version values. Consider the following segment rule:
version
SemVer >=
4.2.52
This segment would include all users that have a version
trait set to 4.2.52
or greater. For example, any of the
following version
values would match:
4.2.53
4.10.0
5.0.0
Versions are compared as defined by the Semantic Versioning specification.
Percentage Split is the only operator that does not require a trait. You can use it to drive A/B tests and staged feature rollouts.
Percentage Split deterministically assigns a "bucket" to each identity solely based on its ID and not any traits, meaning that Segment overrides that use Percentage Split will always result in the same feature value for a given identity.
If you create a Segment with a single Percentage Split rule, Identities who are members of that split when the split value is set to, say, 10% will be guaranteed to also be in that split if it is changed to a value higher than 10%.
If the Percentage Split is reduced in value, some Identities will be removed from that Percentage Split to maintain the balance. The algorithm is fairly simple and good to understand - it is described here.
This operator performs a modulo operation, which returns the remainder
of dividing a numeric trait value by a given divisor. The operator accepts rule value in the format divisor|remainder
.
For example:
user_id
%
2|0
This segment will include all identities having an user_id
trait that is divisible by 2, i.e. even numbers. This is
equivalent to the following expression in many programming languages:
user_id % 2 == 0
Minimum SDK versions for local evaluation mode
When running in local evaluation mode, SDK clients evaluate segment rules locally, which means they must be updated to support the latest operators.
If an SDK client tries to evaluate a segment rule that has an unrecognised operator, that rule will silently evaluate to
false
. The table below lists the minimum required SDK version required by each operator:
Modulo | In | |
---|---|---|
Python | 2.3.0 | 3.3.0 |
Java | 5.1.0 | 7.1.0 |
.NET | 4.2.0 | 5.0.0 |
Node.js | 2.4.0 | 2.5.0 |
Ruby | 3.1.0 | 3.2.0 |
PHP | 3.1.0 | 4.1.0 |
Go | 2.2.0 | 3.1.0 |
Rust | 0.2.0 | 1.3.0 |
Elixir | 1.1.0 | 2.0.0 |
Limits
These are the default limits for segments and rules:
- 100 segments per project
- 100 segment overrides per environment
- 100 rules per segment override
- 1000 bytes per segment rule value
See the documentation on System Limits for more details.
Custom fields
Optional or required custom fields can be defined when creating or updating segments. Learn more