Recognize it? As an organization with subscription services, reducing churn is probably on the agenda. Not the most popular topic, because we prefer to look at what is going well. So what is the best way to analyze churn? Where do you start? In this blog, we explain how you can use data to analyze churn with the hope of churn reduction.
What is churn? And why is reducing churn so important?
Churn is the number of customers who stop using your company’s services or products. It indicates how well a company manages to retain its customers. Churn has a direct negative impact on your organization’s revenue and profitability. Losing existing customers means an immediate drop in recurring revenue, and acquiring new customers is often more expensive than retaining your customers.
And what about your competitive advantage? Retaining your customers gives you predictability in your revenue, which provides stability. As a business, it allows you to focus on growth rather than replacing your lost customers. This allows you to get ahead of your competitors. Plus, what could be finer than a strong product with satisfied customers?
1. Collecting and organizing subscriber data over a period of time.
In our analysis, we focused on the online usage of subscribers to two news websites and their apps.
Our goal sounds simple: improve subscriber retention. We want to understand why some subscribers cancel their subscriptions and see what factors contribute to this. By understanding these patterns, targeted strategies can be developed to improve subscriber retention.
For our churn analysis, we started by organizing a large amount of previously collected subscriber data. This data is spread across several tables, so the first task is to go through these tables carefully and merge the relevant information. In doing so, we selected the most important factors based on domain knowledge and supported by literature review (Belchior, L. M., António, N., & Fernandes, E. (2024).
To merge the data, we used dbt (Data Build Tool) in combination with Snowflake. Snowflake acts as the central repository for all our tables, while dbt allows us to transform this data in an efficient and structured way. With dbt, we can build models that select the most important columns, perform calculations, and ultimately create a composite dataset. This allows us to process our data consistently and repeatably, giving us a robust basis for further analysis.
We chose to use a one-year analysis period, after consulting with experts within the domain. This period provides a good overview of subscriber behavior. A disadvantage of this choice is that seasonal factors were not included in the analysis, which could affect the interpretation of the results, especially if churn patterns fluctuate greatly seasonally.
Due to the sheer size of our dataset, it is essential to break up the transformations into short periods. In this case, we chose to add the data to our dataset on a monthly basis, although this is highly dependent on the number of clients you work with. For this task, we deployed the tool Conveyor.
2. Defining churn
An important step in analyzing churn is to formulate a tight churn definition. This is because the definition you use can have quite an effect on the analysis, so it is wise to make it dependent on the purpose of the analysis. Since we are interested in the behavior behind churn, we chose to make the rules quite rigid. For example, we did not mark subscribers who return within two months as churners. This way, we can avoid creating noise in the dataset, as many subscribers temporarily cancel their subscriptions to get extra discounts. If you are more interested in revenue or outflow per month, it may be wise to include these types of subscribers in the analysis though.
3. Data analysis and interpretation.
After aggregating and organizing the data, we identified key variables that may influence subscriber attrition (churn). We then analyzed these variables to discover patterns that indicate an increased risk of churn. During this analysis, it became clear that segmenting subscribers into groups is essential to refine insights.
We chose to base the segmentation on three main criteria: subscription type, length of relationship with the organization and subscriber activity. These segmentations allowed us to see which factors within each segment played a significant role in churn risk.
Segmenting the data in this way allowed us to better identify specific patterns and behaviors by subscriber group.
Conclusion
Analyzing and addressing churn provides the following valuable insights:
- A churn analysis provides insights on why customers leave, which helps with product improvement.
- Identifying risk factors for churn enables proactive retention strategies.
- Lowering churn forces you to continually add value for your customers.
- By actively combating churn, you can retain your customer base, grow and improve your services as a subscription company. This leads to a stronger market position and sustainable long-term success.
