Survival Analysis Modeling for Customer Churn prediction on papAI

Survival analysis was developed to measure lifespans of individuals. It can be applied to any situation where we have births and deaths (duration). Medical professionals might be interested in the time between childbirths, where a birth in this case is the event of having a child, and a death is becoming pregnant again, Another example is users subscribing to a service: a birth is a user who joins the service, and a death is when the user leaves the service.

Survival analysis

What’s a customer churn ?

Customer churn is a problem that can be modeled using survival analysis. It is a business decision-making process to describe customer behavior during the time period before they stop doing business with us, and decide to move on to another supplier or not purchase anymore from any provider at all. 

Customer churn can be defined as every case where a customer decides that she/he wants to end the relationship with a company. There are several factors that can increase customer churn, such as poor service quality or pricing. Customer churn not only means that the customer is moving to another company, but it can also mean losing a customer in terms of reducing purchases or being late with payments.

How to use survival analysis for customer churn problems ?

  • identification of the survival time :
The first step is to identify the survival time or the survival event in our case. If we are interested in how long it takes for a customer to cancel their service (survival event), then we would consider “time-to” cancellation as survival data points and use them for studying this type of problem.
Survival Analysis Modeling for Customer Churn prediction on papAI
  • Identification of Data :
We must then identify the data we have or that we wish to collect and ensure that it is sufficient to answer our survival question. For example, if we want to study the survival time of clients between two time periods (e.g., January 2022 – September 2022), we must have survival data for the period January 2022 – September 2022.
  • Choose the model :
Next, we need to choose what type of survival analysis models we want to use in our use case. The following list includes the list of models presented on papAIa data scientist can help you in this phase:
 
– Cox’s proportional hazard model
– Weibull AFT model
– Log-Normal AFT model
– Log-Logistic AFT model
Survival Analysis Modeling for Customer Churn prediction on papAI

We can train multiple models with different combinations of different features at the same time and then compare their performances after being evaluated via papAI to select the best model that will be used later to do predictions on new individuals data.

Survival Analysis Modeling for Customer Churn prediction on papAI
  • Validate and interpret the survival model :
Some metrics and visualizations are exposed to well understand the results and performances of our trained models, mainly we can use AIC (Akaike information criterion) to compare between our models. And log-likelihood, concordance index to evaluate how good fitted our model is.

 

Survival Analysis Modeling for Customer Churn prediction on papAI

We can also take a look at the features coefficients to interpret the model and  see which features impact more the decision of churn to users and compare between different values of a specific feature or combination of more using the partial effects on outcome plot.

Survival Analysis Modeling for Customer Churn prediction on papAI

Customer churn is a major concern that affects not only the growth of your business, but also the profit.The churn analysis technique can help you to better understand the value of your customers and initiate a retention process based on the different metrics that the papAI platform offers.

 

Interested in discovering papAI ?

 Our commercial team is at your disposal for any question.

 
Survival Analysis Modeling for Customer Churn prediction on papAI
Scroll to top