Loyalty Management

Problem Statement

One of the leading Retailer in the US having in excess of 300 million customer was looking for means to

null

Reduce User Churn.

null

Predicting User Engagement based on Profile and Activity Data.

null

Predicting Total User activity in a day.

null

Personalizing campaigns and Loyalty Programs.

Solution

Sentienz proposed a customer engagement Platform that contains custom framework for customer micro segmentation and Engagement Prediction. Customer profile and activity data was ingested into specific modules for each tasks like Engagement and Segmentation. The Key solution components applied were :

  • Streamsets for Ingestion and Data pipeline creation
  • DataRobot and H20 for creating AI and ML Models
  • Unsupervised Models for Customer Segmentation
  • Classification Models for User Engagement Prediction
  • Deep Learning Time Series Models for Total Events Prediction

Actions

null

Implemented Streamsets

For ingesting profile and activity data and to create data pipeline for micro segmentation of the customers.
null

Implemented Campaign Management System

Segmented data of the customer was fed to the Campaign system and targeted campaign.
null

Implemented DataRobot and H20 for generating models ( AI and ML)

  • Customer demographic and 30 days cumulative activity data was used to predict year end customer engagement.
  • Customer activity data was used to predict total events per day

null

Implemented Streamsets

For ingesting profile and activity data and to create data pipeline for micro segmentation of the customers.
null

Implemented Campaign Management System

Segmented data of the customer was fed to the Campaign system and targeted campaign.
null

Implemented DataRobot and H20 for generating models ( AI and ML)

  • Customer demographic and 30 days cumulative activity data was used to predict year end customer engagement.
  • Customer activity data was used to predict total events per day

Business Benefits

  • 150%
    Improved the User engagement
    metrics
  • 90%
    Predicted User Engagement with
    over Accuracy
  • 10%
    Predicted total events per day with Mean absolute Percentage Error (MAPE)

Get a copy of use case document