Stop customer churn with data science - high res - rockfeather

Churn prevention

- Keeping existing customers is more cost-efficient than trying to gain new customers.
- Data science can help to prevent churn.

What is churn?

Lots of companies are focusing on getting new customers rather than keeping their existing customer base. For example, why are gyms and telecom providers giving a discount to their new customers, and not to their current customers? As a result of this strategy, customers hop from provider to provider, which leads to a lack of a loyal customer base.

Why prevent churn? 

Getting a list of your future churners is incredibly valuable to your business. You could give them a little bit extra TLC (tender, love & care) or a discount, depending on which marketing strategy is the best fit for each churner, to make sure that these churners will not leave.

You will need to dig through some data if you want to understand which customer characteristics are typical for churners.

To define which characteristics of customers are good predictors for customer churn, you must look at their historical data. This dataset must contain data from customers that have left your company in the past, and from customers that have remained loyal. Next, in this dataset, you can try to uncover any hidden patterns in the customers’ behavior that separate the churners from the non-churners. For example, you might find out that most customers that churned filed a complaint in the last six months before switching. This might be a hint that you don’t follow-up on customer complains thoroughly enough. Thereby, improving the follow-up could lead to fewer churners.


In this section, we will give an example of a classification case. Imagine Football Club A. Each season every football club sells season tickets to its fans. At the end of a season, Football Club A asks their fans if they wish to renew their season ticket. Most fans will, but a small amount of these fans will churn, which means that they will not renew their membership. As a result of this churn, there may be a drop in revenue for Football Club A, which is undesirable but often preventable. If this football club had insight on which fans are likely to churn, they could act on this by giving them extra attention and maybe even a discount. But how on earth could Football Club A know which fans will churn?

Features and target variable

Machine learning can help Football Club A with identifying these fans. Based on the seasons before, the FC knows which fans churned in the past. The FC has information about these fans, like how often they visited the matches and what the prices are for their season tickets. The football club can use this type of information as features in a machine learning algorithm. The target variable we want to predict is, in this case, ‘churn’.

The prediction

The algorithm looks at the data and tries to understand the behavior and characteristics of fans that churned in the past. Based on these patterns, the model can make a prediction for the fans that are currently having a season ticket. The outcome of the model provides a list of fans and their probability of churning. This information can be passed on to the marketing and sales department, who can then play their part to prevent these customers from churning by giving them better reasons to renew their season ticket with the Football Club.


Preventing churn can be valuable for your organization. Knowing which customers are likely to churn not only brings in extra money, but it can also change your business model. If you reward your loyal customers regularly, you do not need to spend a fortune on finding new customers.

Want to know more?

If you have any questions or want to know more about advanced forecasting, feel free to contact Alexander by sending him an email.