In order to create a new experiment for the game select A/B TESTING in the game dashboard menu and click on Create Experiment.
The settings are divided into 3 groups:
- Common data (1)
- Experiment audience (2)
- Experiment groups (3)
Common data of the experiment contains:
- Experiment name (1) - the title of the experiment.
- Description (2) - additional information about the experiment.
- Experiment key (3) - a special short key. It is sent to the game side. The value should include only alphanumeric characters, numbers and underscores.
Experiment audience section allows defining how many players will take part in the experiment and set some rules of including players to it.
- Experiment audience (1) - the max value of players for the experiment.
- New users only (2) - if the option is checked only players joined the game after the start of the experiment will be included in it. In the other case, any player can be included.
By default, the audience size is set as an average daily historical number of new users, if New users only option is checked. In the other case, it is set as an average daily historical number of daily active users.
Advanced settings allow defining filters by country, platform and the game version.
Every experiment should contain at least 2 groups. The maximum number of groups is 10.
Pay attention that the results for each group are compared with the results for group A in the experiment report. So we offer to use Group A as a control group and not to change the behavior for the players from this group.
In order to add a group, click on + ADD GROUP.
Groups can be removed only in turn starting from the last group. So Remove group option is shown only for the last group.
The name and the color of every group are set automatically and can't be changed.
By default, the part of players for group A is 50%, but it can be changed. The minimum value is 25%, the maximum value is 75%. The rest of the experiment audience is distributed evenly between the other groups.
For example, the experiment audience is divided into 4 groups and Group A includes 40% of players. In this case, each group of Group B, Group C and Group D will include 20% of the audience.
Every group has the approximated number of players (1) and the following attributes to set:
- Description (2) - the additional information about the group. It will be shown in the tooltips with the KPI values in the experiment report.
- Payload (3) - some instructions to transform the game if the player was included in the group.
When the experiment is set, you can save it as a draft by clicking on SAVE AS DRAFT or start the experiment by using START EXPERIMENT.