Predicting Customer Churn for Nederlands Dagblad
With predictive analytics on Google Cloud, Nederlands Dagblad boosts subscriber retention.
.png?width=500&height=278&name=Customer%20Cases%20image%20(31).png)
The Challenge
It was challenging for the Nederlands Dagblad team to predict which customers might cancel their subscriptions within the next year. They had plenty of data from sources like Zeno, Pubble, and Google Analytics 4, but this data was scattered across different systems, making it hard to get a complete picture of their customers.
The existing churn model wasn’t providing the actionable results ND needed to make a difference. They wanted a reliable way to predict which subscribers might cancel within a year so their marketing team could intervene early. To achieve this, they needed to:
- Consolidate their scattered data into a single, unified view.
- Identify the most significant indicators of churn.
- Set up the right infrastructure on Google Cloud Platform to power their churn model effectively.
About our customer
Nederlands Dagblad, a well-known Dutch newspaper, wanted to keep its subscribers satisfied and engaged. With various subscription types—from paper to online—they needed a clear way to identify which customers would most likely cancel. By partnering with Crystalloids, Nederlands Dagblad used the power of Google Cloud Platform (GCP) to build a predictive churn model that provided clear, actionable insights and helped them take proactive steps to keep their subscribers onboard.
Start Your Success Story with Crystalloids
Partner with Crystalloids and Google Cloud to bring your project to life.
The Solution
The Crystalloids team started by diving into ND’s data, combining Zeno, Pubble, and GA4 information into a central GCP hub. From there, we identified 35-40 potential churn indicators, such as email unsubscribes and periods of inactivity.
Here’s how we brought it all together:
- Finding Key Indicators: After thorough testing, we narrowed the list to 12 key indicators that had the biggest impact on predicting churn.
- Building the Churn Model: Using BigQuery ML, we trained a machine learning model. We tested it by training on data from January to September and then checking its predictions against data from October to December. The model consistently delivered accurate results.
- Looker Studio: We visualized the churn predictions in Looker Studio, giving Nederlands Dagblad an easy-to-use tool to monitor and act on the insights.
- Keeping It Fresh: The model updates every month, providing ND's marketing team with a new list of high-risk customers to engage with. Regularly updating the model is a key part of their ML Ops process, ensuring it remains accurate and effective.
The Result
Nederlands Dagblad now has a predictive churn model ready to help identify customers at risk of leaving and enable the marketing team to take proactive steps with personalized campaigns.
We worked closely with ND’s team to ensure they understand how to use the model and apply its insights effectively in their marketing strategies.
With the churn model in place, ND is prepared to turn its data into a valuable tool for improving subscriber retention and understanding customer needs.
