Understanding Data For Churn Applications

A quick summary of Understanding Data For Churn Applications

Preparing and uploading data for Churn Model

Understanding Data Requirement for Churn Model

To build a highly accurate churn model for your business, you need to understand what kind of data will be used to train your models.
We have categorized data points into various indicator categories for you to understand what drives churn predictions.

  • Profile: These are data points which define the profile of the customer
  • Symptom: These data points represent customer health symptoms which may contribute to churn or non-churn
  • Causes: These data points represent causes that may lead the customer to leave or stay.

We have come up with a framework that is suitable for most of the business, You may improve it by adding more such data points based on your business context.

Type Field Name Indicator Description
String customer_id profile Unique id of the customer, it can be anonymized on your end as long as it identifies customer uniquely.
String age Profile Age of the customer
String sex Profile Is customer male or female
String city Profile City of the customer
Number days_since_joined Profile How old is the relationship with the customer?
Number days_since_last_purchase Symptom When did customer make his/her last purchase?
Number total_no_of_purchases Symptom How many times has customer made the purchase?
Number lowest_rating Symptom What is the lowest rating by the customer so far?
Number lowest_rating_l3m Symptom What is the lowest rating by a customer in the last 3 months ( l3m )?
Number total_engagements Cause How many times have you engaged with the customer so far? Engagement can be in the form of email, newsletters, calls, etc.
Number engagements_l3m Cause How many times have you engaged with a customer in the last 3 months?
Number total_sales Symptom What is the total sale of the customer so far?
Number total_sales_l3m Symptom What is the total sale of the customer in the last 3 months
Number average_session_time_secs Symptom What is the average time spent by the customer on your app?
Number average_session_time_secs_l3m Symptom What is the average time spent by the customer on your app in the last 3 months?
Number total_logins Symptom How many times did customer log into your app so far?
Number total_logins_l3m Symptom How many times did customer log into your app in the last 3 months?
Number complaint_count Cause How many complaints have you received from the customer so far?
Number complaint_count_l3m Cause How many complaints have you received from the customer in the last 3 months?
Number total_no_pricing_changes Cause How many times has price changed for your offering so far?
Number pricing_changes_l3m Cause How many times has price changed for your offering in the last 3 months?
Number total_no_customer_support_change Cause Has there been a change in your support team so far?
Number customer_support_change_l3m Cause Has there been a change in your support team in the last three months?
Number average_delivery_time_in_days Cause What was average delivery time for the customer so far?
Number average_delivery_time_in_days_l3m Cause What was average delivery time for customer in the last 3 months?
Number site_response_time_secs Cause What is the average response time of your app?
Number site_response_time_secs_avg_l3m Cause What is the average response time of your app in the last 3 months?
Number downtime_counts Cause How many downtimes did customer face so far?
Number downtime_counts_l3m Cause How many downtimes did customer face in the last 3 months?
Number competitor_promotions_l3m Cause How many competitors promotion campaigns happened in the last 3 months
Number promotions_l3m Cause How many promotion campaigns have you done in the last 3 months

Format : CSV

Once you prepare the data, make sure it is in CSV format with the headers similar to the above table.
You can add more fields to data set when you are ready to build your own ML model.

Download a sample data set: Datoin Sample Churn

For more details on Data Sets visit here