Skip to main content
Version: v1.31.0

How to target specific users

Evaluation Context

An evaluation context in a feature flagging system is crucial for determining the output of a feature flag evaluation. It's a collection of pertinent data about the conditions under which the evaluation is being made. his data can be supplied through a mix of static information (server name, IP, etc ...) and dynamic inputs (information about the user performing the action, etc ...), along with state information that is implicitly carried through the execution of the program.

When using GO Feature Flag, it's often necessary to personalize the experience for different users. This is where the concept of a targeting key comes into play. A targeting key is a unique identifier that represents the context of the evaluation (email, session id, a fingerprint or anything that is consistent), ensuring that they are consistently exposed to the same variation of a feature, even across multiple visits or sessions.

For instance, GO Feature Flag ensures that in cases where a feature is being rolled out to a percentage of users, based on the targeting key, they will see the same variation each time they encounter the feature flag.

The targeting key is a fundamental part of the evaluation context because it directly affects the determination of which feature variant is served to a particular user, and it maintains that continuity over time.

Rule format

A rule is a configuration that allows to serve a variation based on some conditions.

Format details

Name of your rule.
This is needed when your are updating a rule using a scheduled rollout.

Query represents an antlr query in the nikunjy/rules format.
This field is mandatory in every rule used in the targeting field.

See query format to have the syntax.

Note: if you use the field query in a defaultRule it will be ignored.

Name of the variation to return.

Represents the percentage we should give to each variation.

variationA: 10.59
variationB: 9.41
variationC: 80

The format is the name of the variation and the percentage for this one.

Note: If your total is not equal to 100%, this rule will be considered invalid.


Allows you to ramp up the percentage of your flag over time.

You can decide at which percentage you start and end with in your release ramp. Before the start date we will serve the initial percentage and afterwards, we will serve the end percentage.

See progressive rollout to have more info on how to use it.


Set to true if you want to disable the rule.

Default: true.


variation, percentage and progressiveRollout are optional but you must have at least one of the three.

If you have more than one field we will use the first one in the order progressiveRollout > percentage > variation.

Query format

The rule format is based on the nikunjy/rules library.

All the operations can be written in capitalized or lowercase characters (ex: eq or EQ can be used). Logical Operations supported are AND & OR.

Compare Expression and their definitions (a|b means you can use one of either a or b):

eq, ==equals to
ne, !=not equals to
lt, <less than
gt, >greater than
le, <=less than equal to
ge, >=greater than equal to
swstarts with
ewends with
inin a list
notnot of a logical expression


  • Select a specific user: key eq ""

  • Select all identified users: anonymous ne true

  • Select a user with a custom property: userId eq "12345"

  • Select on multiple criteria: All users with ids finishing by that have the role backend engineer in the pro environment for the company go-feature-flag

    (key ew "") and (role eq "backend engineer") and (env eq "pro") and (company eq "go-feature-flag")


When you initialise go-feature-flag you can set an environment for the instance of this SDK.

// ...
Environment: "prod",
// ...

When an environment is set, it adds a new field in your user called env that you can use in your queries. It means that you can decide to activate a flag only for some environment.

Example of flag configuration based on the environment:

A: "A"
B: "B"
C: "C"
- name: Target pre environment
query: env eq "pre"
variation: A
- name: Target pro environment
query: env eq "pro"
variation: B
variation: C

Get the rule name in the metadata

When you use a rule in your targeting, you can get the name of the rule in the metadata of the variation.
The information on what rule has been used to serve the variation is available in the metadata of the variation in the field called evaluatedRuleName.

If you are interested about this information, you have to name your rules by adding the field name in your rule. This name will be extract and added in the evaluatedRuleName field of the metadata.

Get the latest GO Feature Flag updates