When conducting an A/B test, not only do you need to distribute your users to different test groups, you also need to write a logic for each test group. After the integration stages, the distribution of users into different test groups can be done through the Hackle SDK.
The test groups are basically the different groups that are exposed to the versions (features, screens, algorithms, etc) of a page of an experiment. Test groups include both the control test group A and treatment test groups B, C, D, etc., that are exposed to the "improved" versions of the pages.
The test groups can be set on the dashboard and further information on the test groups can be found on the What is a Test Group? document.
Think of the variation as the term that is equivalent to different versions of the page being shown to users on your app or browser.
By passing the experiment key and user identifier to the
variation() method, you can distribute users and receive the results. After that, we implement the logic of the page version applicable for each test group.
In the example code below, we are passing an experiment key of 42, and there are two test groups, A and B.
# Determines the test group to assign the user "ae2182e0" # in an A/B tests with an experiment key of 42. # For undetermined cases, the user is returned to test group A. user = Hackle.user('ae2182e0') variation = hackle_client.variation(experiment_key=42, user=user) # Logic for the assigned test group. if variation == 'A': # Logic for test group A elif variation == 'B': # Logic for test group B
variation_detail() method works the same as the
variation() method but provides the reason for a user being distributed to a specific group. This method can be a useful tool to see if the distribution is working properly.
You must pass the experiment key as a parameter. For the example code below, we are passing experimental key 42.
# Traffic distribution details decision = hackle_client.variation_detail(experiment_key=42, user=user) # Test group determined from distribution variation = decision.variation # Reason for distribution to a test group reason = decision.reason
You will receive the reason for the distribution or the allocation of a specific user to a specific test group in the form of
SDK_NOT_READY. Please refer to the table below for the full list of different distribution reasons.
|SDK_NOT_READY||The SDK is not ready to use.|
(e.g. initialized with the wrong SDK key)
|(control) Test Group A|
|EXPERIMENT_NOT_FOUND||No A/B tests were found for the experimental key you passed.|
The experiment key may be incorrect or the experiment may be in the archive status.
|(control) Test Group A|
|EXPERIMENT_DRAFT||The A/B test has not yet been started (in draft mode).||(control) Test Group A|
|EXPERIMENT_PAUSED||The A/B test has been paused.||(control) Test Group A|
|EXPERIMENT_COMPLETED||The A/B test has ended.||Final winning test group from the experiment|
|OVERRIDDEN||Users are distributed to a specific test group by manual assignment.||Manually assigned test group|
|TRAFFIC_NOT_ALLOCATED||A/B test is running, but user has not been assigned to the experiment.||(control) Test Group A|
|TRAFFIC_ALLOCATED||User has been assigned to a test group in the A/B test.||Assigned test group|
|VARIATION_DROPPED||The test group was removed from the A/B test.||(control) Test Group A|
|INVALID_INPUT||The input value is not valid.|
(e.g. A number was entered in a parameter that requires a character)
|(control) Test Group A|
|EXCEPTION||An unknown error has occurred.||(control) Test Group A|
Deduplication of exposure event
If you use Backend-side SDK, Any successive Exposure events within one minute for same A/B test from single user will be counted as one.
- Parameter values of the distributed group can also be provided through the
- Through the config object and the
get()method, you can receive and use the parameter setting value set in the dashboard, and if you change the value in the A/B test parameter setting tab, the changed value is applied to the code.
Parameter setting is only available for SDK version 3.1.0 or higher.
from hackle.model import HackleUser user = HackleUser(id='ae2182e0') decision = hackleClient.variation_detail(experiment_key=42, user=user) # Get parameter value through get() method in traffic distribution details parameterValue = decision.get('parameterKey', 'defaultValue') # Example of string type parameter value strValue = decision.get('parmeterKey', 'defaultValue')
- The parameterKey of the
get()method is the key information set in A/B Test Parameter Settings and the defaultValue is the value returned when the distribution decision fails or when the wrong parameter type is entered.
- In order to properly receive the information you set in the dashboard, you must enter a value corresponding to the parameter type of defaultValue you set in the dashboard.
- JSON type is received in the form of a string, so in the case of JSON type, the defaultValue must be entered as a string type.
- The parameter types supported by SDK are string, number, and boolean. JSON types set in the dashboard can be received in the form of String. The default value of the JSON type must be entered as a string type.
Updated 12 months ago