Churn Prediction and Transforming Data Using Bold Data Hub

In this article, we will demonstrate how to import tables from a CSV file, to predict churn using transformations, and move the cleaned data into the destination database using Bold Data Hub. Follow the step-by-step process below.

Sample Data Source:
Sample CSC Data

Step-by-Step Process in Bold Data Hub

Step 1: Open Bold Data Hub

• Click on the Bold Data Hub.

Step 2: Create a New Pipeline

• Click Add Pipeline in the left-side panel.
• Enter the pipeline name and click the tick icon.

Step 3: Choose the Connector

• Select the newly created pipeline and opt for the CSV connector. You can either double-click or click on the Add Template option to include a template.

Step 4: Upload Your CSV File

• Click the “Upload File” button to select and upload your CSV file.

Step 5: Set the Properties

• Copy the file path and paste it into the filePath property field.

Step 6: Save and Choose the Destination

• Click Save, choose the destination, and confirm by clicking the Yes button.

Note: On-Demand Refresh will be triggered when the pipeline is saved. If needed, the pipeline can be scheduled in the Schedules tab.

Step 7: View Logs and Outputs

• Click the pipeline name in the left-side panel and switch to the Logs tab to view logs.

Step 8: Apply Transformations

• Go to the Transform tab and click Add Table.

• Enter the table name to create a transform table for customer satisfaction summary.

Note: The data will initially be transferred to the DuckDB database within the designated {pipeline_name} schema before undergoing transformation for integration into the target databases. As an illustration, in the case of a pipeline named “customer_service_data”, the data will be relocated to the customer_service_data table schema.

Learn more about transformation here

Churn Prediction Feature

Overview

Churn prediction models are used to forecast the likelihood of a customer discontinuing their relationship with a company. By creating features such as the time since the last contact, frequency of tickets, or changes in support issues, we can enhance the model’s ability to predict churn. These features provide valuable insights into customer behavior patterns and engagement.

Approach

We can derive the following features from support ticket data:

• Time Since Last Contact: The time difference between the most recent support ticket and the current date.
• Frequency of Tickets: The number of tickets raised by the customer within a given timeframe (e.g., monthly, quarterly).
• Changes in Support Issues: Tracking the nature or type of support issues over time to detect shifts that might indicate dissatisfaction.

SQL Query for Creating Churn Prediction Features

WITH Customer_Activity AS (
    SELECT 
        Customer_ID, 
        Customer_Name, 
        MAX(Ticket_Creation_Date) AS Last_Interaction, 
        COUNT(Ticket_ID) AS Total_Tickets, 
        SUM(CASE WHEN Ticket_Status = 'Resolved' THEN 1 ELSE 0 END) AS Resolved_Tickets, 
        AVG(Customer_Satisfaction) AS Avg_Satisfaction 
    FROM {pipeline_name}.sample_csc_data 
    GROUP BY Customer_ID, Customer_Name
)
SELECT 
    c.*, 
    (CURRENT_DATE - Last_Interaction) AS Days_Since_Last_Contact, 
    (Resolved_Tickets * 1.0 / NULLIF(Total_Tickets, 0)) AS Resolution_Rate 
FROM Customer_Activity c 
ORDER BY Days_Since_Last_Contact DESC;