CHURN PREDICTION
FOR MOBILE OPERATORS
Digitata's Churn Prediction allows you to instantaneously predict, with high accuracy, which of your subscribers are going to churn.
Quickly and easily create your own version from the AWS Marketplace or contact us for more details.
Enable now!82.8%
Of Churners Identified
94.3%
Correct Churn Predictions
9.7%
Actual Churners
500+
Model Attributes
It costs between 5 times and 25 times as much to find a new customer than to retain an existing one.*
*According to Harvard Business Review
Quick Predict
Voice Active Days 4weeks
Recharge Count 4weeks
Voice OOB Spend 4weeks
Rechage Spend 4weeks
Voice Active Days 4weeks 2
Change in Recharge Count
Change in Recharge Amount
Change in Voice Active Days
Total Recharges 4weeks
Out of Bundle Call Count
Select calculate to start
Don't be startled by churn.
Predict it.
Churn prediction for marketing.
Plan with it.
Churn prediction to increase.
Customer loyality.
Churn has too often been approached with scattershot, manual methods. By using machine learning algorithms this enables a scientific, accurate and automated method of managing churn.
Here are some interesting churn related resources: AWS Machine Learning Blog - Preventing customer churn by optimizing incentive programs using stochastic programming
Enable prediction now!
Follow the steps below:
1. Start
Click on this link to open the AWS marketplace.
2. Region and Setup
Follow the steps on the AWS marketplace to deploy and setup the churn model in your AWS account. You will choose your AWS region and click Continue to subscribe.
3. Subscribe
You will be presented with a summary of what will be deployed. Review the listed product and license agreement :wand select Accept Offer.
4. Configure
After subscribing, select Continue to configuration. You can choose to launch either using the AWS CLI (Command Line Interface) or using the Sagemaker Console. If you are not familair with the AWS Command Line Tools, it is suggested that you select Sagemaker Console. Keep the selection at Create real-time inference endpoint under Sagemaker Options.
5. Install
Select View in Amazon Sagemaker or enter the commands if you are launching through the CLI. If launching through the CLI, skip to step 8. Otherwise continue onto the next step.
6. Create Model
On the Create Endpoint/Model page, enter a name for the model that will be deployed. Something like churn-prediction for example. Select Next at the bottom, unless you have specific networking requirements we don't require Container or VPC setup at this point in time.
7. Create Endpoint
On the Create Endpoint/Endpoint page, enter a name for the endpoint that will be deployed. This name is what we will use when we are doing real-time inferences against the model. Select Create endpoint configuration. Wait for the Console to update your Endpoint Configuration, then select Submit.
8. Run
Recall the Endpoint name you selected in the previous step. Using your favorite language of choice, you can now use the AWS SDK to perform real-time inferences against the model. Such as the following: Use the following commands to run an inference against your endpoint:
const AWS = require('aws-sdk');
const sagemaker = new AWS.SageMakerRuntime();
const Params = {
EndpointName: 'model_name_from_setup',
Body:'data in csv format',
ContentType: 'csv',
}
sagemaker.invokeEndpoint(Params, function (err, data) {
if (err) { console.log('error', err); return }
console.log('probability to churn', data.Body.toString());
});
9. Data Preparation