Telecom arguably went from an oligopoly in the 1980s and 90s to a highly competitive market today. We see the constant competition between both big players like Verizon and AT&T to much smaller, local companies to offer competitive prices without the brand prestige. Thanks to this, services have more affordable on a per-unit basis, as infrastructure continues to expand, on fairly often, telco companies share infrastructure. Customers seem to be willing to switch from one provider to another. But what are the reasons that customers are so readily willing to change despite the hassle? Churn rates for telco range from 1% to 5%, depending on a given timeframe or company. These numbers are not high, but when a single company has millions of customers, a percentage point change can represent a great sum of money and market share.
Use Case
What exactly are the reasons that customers are willing to switch from one telco provider to the other? Is price the most important factor? Do age or gender come into play? Maybe if you only have one service with a company versus multiple, you are more willing to move companies because the latter would be more complicated and time-consuming. We will highlight the most important factors in this decision-process based on a dataset from Kaggle.
Benefits
- Adapt the Telecom service offers & Product development department
- Modify Marketing 4 Ps' strategies: ( Place / Product / Promotion / Price )
- Get consumers insights that will help Sales team target the right audience with the right offer
- Help the customer service department offer the right solution
Hypotheses
Null: 1) Churn is not causally influenced by factors in our dataset 2) We cannot predict churn rates based on the factors - accuracy less than or equal to 50% (baseline figure)
Alternate: 1) Certain factors have statistically significant causal relationships with churn rates 2) Accuracy greatear than 50%
Type 1 and type 2 errors can exist for both hypotheses 1) and 2)
Data Exploration
See Pandas Profiling Report: https://github.com/McGill-MMA-EnterpriseAnalytics/FlamesChurn/blob/master/output_churn.html
Causal Analysis
Feature selection shows Tenure to be one of the most significant factor in dataset. Using the number of months the customer remains with the Telecom company as the treatment in our causal analysis, we determine its impact on the churn rate.
Conclusion
- Based on our model prediction we conclude:
1. Rejecting the null hypothesis, using Tenure, Payment Method, Paperless billing and the number of servives currently activated by the customer we can predict customer churn.
2. Focus on retaining customers on contract between 10 and 40 months
3. As customers with longer tenures are less likely to churn, telco companies should avoid providing month to month incentive packages.