Survival analysis is a form of retention rate analysis that enables businesses to predict a customer's churn time by applying statistical techniques. Survival analysis can determine customers' activity before they move on to another seller.
Businesses can evaluate a customer's lifetime value (CLV), customer's loyalty and future expected revenue with survival analysis.
Benefits of survival analytics
Survival analytics forms the basis of determining your customer's response to your business. Product managers and marketers can use the resulting information to improve their efforts:
- Study the retention rates to determine the effectiveness of customer acquisition channels like PPC and affiliates
- Retain high-value customers who have lower survivability rates using churn prevention methods
- Consider the timing of marketing campaigns that allow customer acquisition, depending on the weekday and date that show the greatest-value customer cohorts
- Carry out LTV calculations as the amount of time a customer stays active with your business is determined by the predicted monetary value in the future
Survival analytics models
Two of the most common survival analytics models are:
Kaplan-Meijer estimate of the survival function
As a non-parametric statistic, this model uses lifetime data for an approximate survival function. For example, you can use the calculation to show:
- In a given time “T”, the survival function shows the probability of a customer not churning.
- On the contrary, the cumulative hazard function depicts the total risk of the customer churning during period “T”.
A customer's survival function and the probability of them still being a customer at day “T” (when their subscription ends) are also important findings with this calcuation. The estimator considers the churned customers and those who are "at-risk" of churning in the future.
This model's workings give a real insight into how long a customer can potentially stay with you and ultimately decide its lifetime value.
The Cox Model: Survival Regression
Risk factors and exposures are simultaneously taken into account with survival time, which is precisely what the Cox proportional hazards model calculates. In this model, the hazard rate measures the effect (the failure risk), provided the customer has survived up until a specific time.
This model assumes that every impact is constant over time, which is not always true, and certain factors could go against this assumption.
For instance, a customer's like refund over time could be different and have a varying effect on survival over a period, making it unsuitable to put into the Cox model. It is essential to carefully examine all aspects before customer modeling and their lifetime value.
Are you looking to improve retention rates at your organization? Build a free retention query, in seconds, with Scuba Analytics self-service explorer demo.