A/B Tests are a great way to accurately measure the impact that Chameleon is having and compare different versions of a product tour.
Chameleon makes it very easy to run a Control Group (Chameleon vs. nothing) or A/B Test (two different Chameleon Tours). We recommend starting with a Control Group test and iterating until your Tour is effective, before testing Tour variations.
There are two ways to deliver Chameleon Tours and testing here is relevant for Automatic Tours only.
Once you have selected the Automatic option, you will be able to turn on Control Group testing and set the proportion of users that should be part of the control group -- as in the image below, where a random 20% of the target audience users will not see this Chameleon Tour.
Every user identified by Chameleon is automatically assigned a random number ("Testing ID" property) with a value between 0 and 100. This value is persistent for each user and can be leveraged for targeting and experimentation.
To select a control group, Chameleon randomly chooses a number between 0-100 as the start of the range for the group. The size of the group will determine the end range for the group (including going back to 0 after 100.) The corresponding users within this range will form the group.
e.g. if you choose a 20% control group for a Tour, the Testing IDs for the group could be 87.3 to 7.3. If you do this for another Tour, the group may be 12.9-32.9.
If you later increase the size of the control group, then the range is simply increased, from the same starting value.
An event will be logged whenever a user is first identified to be within the Target Audience, and thus is liable to see this Tour. This event* -- "Chameleon Experiment entered" -- will be available within all your connected analytics integrations.
Note: This event does not take into account URL Rules or other conditions for the Experience, only when a user first matches the Target Audience segment's conditions.
* Within Mixpanel, the event name is "Experiment Started" to better match Mixpanel's experimentation analysis framework.
For this event the following properties will also be logged:
The "Group" property values are either:
Control (Out) -- user is part of the control group and will not see the Tour
Test (In) -- user is part of the test group and can see the Tour
Other events (e.g. "Tour started") will also be logged as normal once a user starts interacting with the Chameleon experience. You will see all these within your analytics platform (e.g. Amplitude, Heap, Mixpanel, Google Analytics) and can use this to further analyze the conversion or relative impact of your Chameleon experiment*.
* It is within our plans to deliver reporting within the Chameleon Dashboard to measure experiment success and goal conversion, let us know what you'd like to see here.
Chameleon also lets you manually show an Experience to a random proportion of your Target Audience, so that you can show two different Tours to different groups or group multiple tests to the same random sample.
To create a Test Group, simply add an extra "sampling filter" filter to your Target Audience that selects a random sample of users of the desired size. This means you can still target users based on other conditions, such as user properties, events, data sources, etc.
To add the sampling filter:
Select Default properties as the type of filter
Select Testing ID in the next dropdown
Use more than or less than to define the range of users
Set the boundary number for this Testing ID value.
In the above example, users that have a Testing ID value between 50 and 100 would be targeted. This would constitute 50% of the users within the group defined by the other segment filters.
To target 10% of users, you could use either:
Testing ID more than 90
Testing ID less than 10
You could also use multiple filters to target another 10% group (e.g. Testing ID more than 10, and Testing ID less than 20).
You can use the same filter configuration in another Target Audience (by re-creating the filter) to target the same user group. This means you can run multiple A/B tests to the same user group.
To test two variations of a Tour:
Create the control variation of the Tour, including the Target Audience, using the sampling filter above
Duplicate the control Tour, update the Tour name (using the variant name/label), and then re-create the Target Audience. This time use the opposite sampling filter, so that you're targeting the alternative user group
Set both Tours live
You will be able to use the Tour name to segregate the data, as each Chameleon event contains additional attributes, such as Tour name, User ID, URL, etc.