Introduction
Customer churn is a major challenge for telecom companies, leading to significant financial losses. In this blog, we demonstrate how Decima2 can help reduce churn and generate savings of over £790,000 for our telecom case study.
The Importance of Marketing in Telecom
Marketing plays a crucial role in customer retention. Our case study involves 3,733 customers, of whom 1,869 are currently churning. With the value of retaining a customer at £2,000, a 50% churn rate represents a major financial loss.
Marketing campaigns can help prevent churn, but traditional approaches are costly and rely on trial and error. The telecom company's marketing department currently offers the following customer retention packages:
- Free Online Security Package: £500 per customer
- Free Online Security Package + Discounted Monthly Charges: £800 per customer
- Discounted Monthly Charges: £300 per customer
However, there is no way to determine in advance which package is most effective. Marketing teams must conduct A/B testing, where customers are randomly assigned to different groups receiving different packages. The results of these tests inform future marketing campaigns.
The Limitations of A/B Testing
A/B testing has several key drawbacks:
- High Cost & Risk: Assigning packages randomly without knowing their effectiveness can lead to financial losses, especially when some packages are significantly more expensive than others.
- Slow Results: It takes time to collect enough data for statistically significant results, delaying future marketing efforts.
- Operational Complexity: Randomly grouping customers requires substantial effort and access to demographic data, which may not always be available.
Decima2: A Smarter Approach to Reducing Churn
Instead of relying on costly and inefficient A/B testing, Decima2 uses a causal generative model to run synthetic experiments. This creates a virtual environment where different customer retention strategies can be tested instantly, without the cost of real-world implementation.
Understanding Churn with Decima2
When we load our telecom dataset into Decima2, we can see 3733 customers with a balanced number of churners and non-churners.
I’d like to understand my customers, have a look at which features are making them churn. The relevant features for the customer packages offerred by Marketing are OnlineSecurity_Yes, OnlineSecurity_No and MonthlyCharges. Use Decima2 to explore the causal effect of these features on churn:
- OnlineSecurity_No (0.406 causal effect): Customers without online security are more likely to churn.
- MonthlyCharges (0.218 causal effect): Higher monthly charges also contribute to churn.
Since OnlineSecurity_No is the fourth most influential factor while MonthlyCharges ranks 20th, we hypothesize that providing free online security will be more effective at reducing churn than offering discounted monthly charges. We test this using the Decima2 Optimize Outcomes Portal.
Using Decima2 to Find the Best Retention Strategy
After identifying what drives churn, we use the Optimal Outcomes Portal to run experiments. This portal allows us to simulate different interventions to determine the most effective strategy.
Defining Actionable Features
Actionable features are those that we can modify through external interventions. In this case, they include:
- OnlineSecurity_Yes: Setting this feature to 1 equates to offering a customer free online security.
- MonthlyCharges: Lowering this value equates to offering a customer a discount.
The Decima2 Algorithm
Decima2 runs thousands of synthetic experiments, testing different combinations of our actionable features. For example:
- Synthetic Experiment 1: OnlineSecurity_Yes = 0, MonthlyCharges = £55 -> Churn reduction = 0%
- Synthetic Experiment 2: OnlineSecurity_Yes = 1, MonthlyCharges = £30 -> Churn reduction = 43%
By running these experiments Decima2 identifies the optimal intervention: setting OnlineSecurity_Yes to 1 which reduces churn by 57%. This informs our marketing team that the optimal retention strategy is Package 1, offerring customers free onlline security.
The Power of Causal Analysis
Unlike correlation-based models, Decima2 understands causal relationships between variables. For example:
- Setting OnlineSecurity_Yes to 1 reduces churn.
- Setting OnlineSecurity_Yes to 1 has a causal impact on OnlineSecurity_No, setting it to 0. You can't have OnlineSecurity_Yes, OnlineSecurity_No simultaneously.
- Setting OnlineSecurity_Yes to 1 has a causal impact on OnlineBackup_No. Customers with online security are less likely to require online backup services.
These causal insights can inform future marketing strategies, helping companies optimize their service offerings based on real causal effects.
Business Impact: Decima2 vs. Random Allocation
Now we understand the Decima2 recommendation, we can evaluate its value compared to the telecom company's alternative strategies to reducing churn: A/B allocation strategy and doing nothing.
A/B Allocation Strategy
- Total cost: £1,999,200
- Total benefit: £1,400,000
- Net value: -£599,200
Doing Nothing
- Net value: -£3,738,000
Decima2-Optimized Strategy
- Total cost: £1,866,500
- Total benefit: £2,100,000
- Net value: +£191,350
By rolling out Package 1 (Free Online Security) to all customers, Decima2 helps maximize retention while minimizing costs, saving the telecom company £790,550 compared to A/B allocation, demonstrating the power of data-driven causal analysis.
Conclusion
Decima2 provides a smarter, more efficient approach to reducing telecom churn, eliminating the need for costly and time-consuming A/B testing. With Decima2, companies can:
- Understand why customers churn
- Identify the most effective retention strategy
- Optimize financial outcomes
With Decima2, telecom companies can move beyond guesswork and into precise, data-driven marketing strategies that deliver measurable success.