The Amazon A/B Testing service is an effective tool to increase user engagement and monetization. It allows you to set-up two in-app experiences (each of which is presented to a different group of users) and evaluate the success of each, based on criteria you establish. Another, less obvious, use of the A/B testing service is feature timing. It’s easy to turn on a new feature at a specific time, or release a series of features on a schedule. You can do it in four simple steps, which I walk through here:

1. Creating a new A/B project and launch used for releasing a feature
2. Integrating the SDK and API into your Android app
3. Adding the Java code necessary to enable your feature based on the A/B launch
4. Releasing your feature via Amazon’s A/B Testing portal

Step 1: Creating a new A/B project and launch used for releasing a feature

Using the Amazon A/B Testing service requires an Amazon developer account and an Android app associated with the account. If you don’t already have a developer account you can sign up here. The first step for using Amazon’s A/B Testing to release a feature is to create an A/B launch used to determine whether the feature in your app is hidden from users or accessible by them. Once signed into your developer account the following steps can be used to accomplish this:

1. Navigate into your app. Click “My Apps” on the navigation bar of the Mobile App Distribution Portal to access your list of apps.
2. On the detail page of your app, click the “A/B Testing” link as shown below.
   
   



3. A screen displaying your list of A/B projects is displayed. Click “Add a New Project”.
   
   



4. Enter a title and description that indicates this test is used for the launch of a specific feature and click “Save Project”.
   
   



5. Click “Add a New Launch” on the detail page of the project you just created.
   
   



6. We are now looking at the page for creating an A/B launch. Enter a name and description that indicates this is used for a feature release. For the variable, we create a boolean that tracks whether the feature is enabled. The old variation is set to false and the new version is set to true. Make sure to keep the old variation set to 100% of the distribution. This means all users of your app will not have the feature enabled when you check this A/B launch’s variable. Click “Save” to add your A/B launch.

7. Click the “Start” button on the newly created A/B launch to make it available for use.

Step 2: Integrating the SDK and API into your Android app

We now have an A/B launch containing a corresponding variable for determining whether our feature is made available. The next step is to integrate the A/B Testing SDK into our app using the following steps so we can write code to obtain whether the feature is made available.

1. Download the Amazon A/B Testing SDK from here.
2. Set the required permissions INTERNET and ACCESS_NETWORK_STATE in your AndroidManifest.xml file.

<uses-permission android:name="android.permission.INTERNET" />
<uses-permission android:name="android.permission.ACCESS_NETWORK_STATE" />

3. Add the Amazon Insights Processing Service to your AndroidManifest.xml.

<application>
    ...
    <service android:name="com.amazon.insights.InsightsProcessingService" />
</application>

4. Add the AmazonInsights Android SDK jar and all 3 AWS Android jars to your Android project.  These jars can be found in the libs folder of the SDK download.
    

Step 3: Adding the Java code necessary to enable your feature based on the A/B launch

With the SDK setup its time to add code to either make our feature hidden or available based on the variable we configured for our A/B launch. The steps below detail the code needed for accomplishing this.

1. Import Amazon Insights in to your code.
   
   import com.amazon.insights.*;


2. Initialize the Amazon Insights SDK in the onCreate method of your Activity. Note: Your application key and private key can be found on the Amazon A/B Testing page for your application where we created the A/B project in step 1.

AmazonInsights

    .withApplicationKey(YOUR_APPLICATION_KEY)
    .withPrivateKey(YOUR_PRIVATE_KEY)
    .withContext(getApplicationContext())
    .initialize();

3. Add code to enable the feature based on the A/B launch variable.  We obtain our A/B launch variable and check whether we should enable access to the feature.

//Get our “Feature Launch - FeatureName” experiment and retrieve the variable
//for whether the feature is made available.
ABTest
   .getVariationsByProjectNames("Feature Launch - FeatureName")
   .withVariationListener("Feature Launch - FeatureName", new VariationListener() {
      public void onVariationAvailable(Variation variation) {
         boolean featureNameEnabled = variation. getVariableAsBoolean(
            "FeatureName_IsLive", false);
         if (featureNameEnabled) {
            enableFeatureName();
         }
    }
});

 Step 4: Releasing your feature via Amazon’s A/B Testing portal

With your app now supporting the release of the feature, the final step is to make your feature available at the appropriate time. This can be done completely via the Amazon A/B Testing portal and doesn’t require an APK update with the feature enabled. To launch the feature use the following steps:

1. Navigate back to the A/B launch experiment you created for your app’s feature release. This is accessible by clicking the “A/B Testing” link on your app’s detail page. Next click on the current experiment for our feature launch A/B project as displayed by Image 8 below.
   
   

2. Click the “Edit” tab to change the distrubtion of the variable we use to determine if the feature should be made available.
   
   

When the Launch was created the distribution was set to 100% to the “Old Version”. Change the distribution of the new variation to 100%. All users of your app are going to receive true for the variable used to enable your feature. Click “Save Launch” and your feature will now be live for all users. 

In some scenarios you might want to incrementally release a new feature over time. This can beneficial when releasing features for which scaling might be a concern. For example,  you can use Launch to safely increase the percentage of customers receiving the feature while you monitor the scalability of your service.   If a problem is detected you can stop incrementing the release or in an extreme case pull it back.

That is all it takes to support releasing a feature via Amazon A/B Testing in your Amazon app. You can find more information the Amazon A/B Testing service here.