The customer churn prediction template on the AI & Analytics Engine supports two different types of churn use cases
On the AI & Analytics Engine, upon choosing the customer churn prediction template, users are asked to choose between two types of churn. The user is expected to choose based on how their customers pay for the business' products and service:
Predicting whether transactional activity will decline in volume or frequency for each customer who is currently active. We refer to this as the “transactional option” in our documentation on this site.
Predicting for each customer the likelihood that they will terminate their ongoing paid subscription service. We refer to this as the “subscription option” in our documentation on this site.
This option is suitable for industries like retail, e-commerce, wholesalers, personal banking etc, where:
Customers decide when to spend and how much to spend.
There is no contract or subscription requiring regular payments, which implies customers do not actively inform a business if they will stop/have stopped using their product by a certain date.
Under this option, the business user defines “churn” based on the customers' activity level (either total number of transactions or total transaction amount) going below a certain threshold, for a certain period of time. For example, a business could decide that a 30-day period with total transaction value less than $100 indicates that the customer has churned.
This graphic illustrates the concept of transactional churn
- In the first scenario, the customer made a small number of low-value transactions over a 30-day period.
- In the second scenario, the customer made no transactions over the same period.
- In the third scenario, the customer has not churned and is still very active: many high-value transactions over the same period
This option is suitable for products and services purchased periodically (weekly, monthly, or annual etc.) This includes subscription fee, such as telecom/internet services, on-demand video/audio streaming, software-as-a-service (SaaS) solutions, etc.
Under this option, customers actively terminate their service by initiating a “cancel” request through their online account or through a customer-service call, and hence the business holds records of whether and when customers who singed up in the past cancelled their services. To use the template, the business users also needs to provide additional data that helps the models learn about possible signs of upcoming customer churn, in the form event logs of various kinds of interaction their customers had with the business. For instance, for telecom churn problem, event logs datasets can be about:
Text and MMS messages