Variation Distribution

When running an A/B Test, you need to distribute users across Variations and implement the logic for each Variation. You can perform this user distribution through the Hackle SDK.

variation

When using the Python SDK, duplicate exposure events for the same A/B Test Variation Distribution result generated by the same user within 1 minute are deduplicated.

Events within 1 minute are counted as one.

Server SDKs do not support cohort targeting. If targeting includes a cohort, run the experiment using a Client SDK.

By passing an Experiment Key and User Identifier to the variation() method, you can distribute the user and receive the result.

Parameter	Type	Required	Constraints
Experiment Key (key)	int	Required	-

Example

In the example code below, Experiment Key 42 is passed and two Variations, A and B, exist.

from hackle.model import HackleUser

# 실험 키가 42인 A/B 테스트에서
# "ae2182e0"라는 식별자를 가진 사용자에게 노출할 테스트 그룹을 결정합니다.
# 결정하지 못하는 상황인 경우 테스트 그룹 A를 반환합니다.

user = HackleUser.builder() \
                 .device_id('ae2182e0') \
                 .build()

variation = hackle_client.variation(experiment_key=42, user=user)

# 할당받은 그룹에 대한 로직
if variation == 'A':
    # 그룹 A 로직
elif variation == 'B':
    # 그룹 B 로직

variation_detail

The variation_detail() method works the same as the variation() method and additionally provides the reason for distribution. This method is useful for checking whether distribution is working correctly.

Parameter	Type	Required	Constraints
Experiment Key (key)	int	Required	-

Example

You must pass the Experiment Key as a parameter. In the example code below, Experiment Key 42 is passed.

# 분배 결정 상세
decision = hackle_client.variation_detail(experiment_key=42, user=user)

# 분배 그룹
variation = decision.variation

# 분배 결정 사유
reason = decision.reason

Distribution Reason

The distribution decision reason is returned in the form of SDK_NOT_READY. Refer to the table below for details.

Distribution Reason	Description	Distribution Result
SDK_NOT_READY	The SDK is not ready to use. (e.g., initialization attempted with an invalid SDK Key)	A (Control Group)
EXPERIMENT_NOT_FOUND	No A/B Test was found for the provided Experiment Key. The Experiment Key may be incorrect, or the experiment may be archived.	A (Control Group)
NOT_IN_MUTUAL_EXCLUSION_EXPERIMENT	The experiment is included in Mutually Exclusive Settings, but the user was not assigned to that Mutually Exclusive Group.	A (Control Group)
EXPERIMENT_DRAFT	The A/B Test is in draft state.	A (Control Group)
EXPERIMENT_PAUSED	The A/B Test is paused.	A (Control Group)
EXPERIMENT_COMPLETED	The A/B Test has ended.	Winning Variation selected at completion
OVERRIDDEN	The user was assigned to a specific group by Override.	Override-assigned group
NOT_IN_EXPERIMENT_TARGET	The user is not an A/B Test target.	A (Control Group)
TRAFFIC_NOT_ALLOCATED	The A/B Test is running, but the user was not allocated to the test.	A (Control Group)
TRAFFIC_ALLOCATED	The user has been allocated to the A/B Test.	Allocated group
VARIATION_DROPPED	The originally allocated group has been removed from the test.	A (Control Group)
INVALID_INPUT	The input value is invalid.	A (Control Group)
EXCEPTION	An unknown error occurred.	A (Control Group)

Parameter Settings

• You can also receive the parameter values of the distributed Variation through the variation_detail() method.

• You can use the config object and get() method to retrieve and use the Parameter Settings values configured on the A/B Test screen. If you change a value on the A/B Test parameter settings screen, the updated value is applied to your code.

• The parameterKey of the get() method is the key configured in the A/B Test parameter settings, and defaultValue is the value returned when the distribution decision fails or an invalid parameter type value is provided.

• To receive the configured information correctly, you must enter a value of the type that matches the parameter type you configured in defaultValue.

Method	Value Type	Description
get	string, number, boolean	* Returns the configured parameter value. * JSON type can be received as a string. * The default value for JSON type must be entered as a string type.

Example

from hackle.model import HackleUser

user = HackleUser.builder() \
                 .device_id('ae2182e0') \
                 .build()

decision = hackleClient.variation_detail(experiment_key=42, user=user)

# 분배 결정 상세에서 get() 메소드를 통해 parameter 값 가져오기
parameterValue = decision.get('parameterKey', 'defaultValue')

# string 유형의 parameter값 예제
strValue = decision.get('parmeterKey', 'defaultValue')