Mitigate Churn with Machine Learning

August 30, 2022

Customer churn is a major pain point for companies and should be carefully monitored. Churn rate is the number of existing customers who cancel or don’t renew their subscriptions, either because they no longer desire your products/services or because they left for a competitor.

Leichtman Research Group, Inc. (LRG) found that the largest pay-TV providers in the U.S. – representing about 95% of the market – lost about 5,120,000 net video subscribers in 2020, compared to a pro forma net loss of about 4,795,000 in 2019.[1]

Knowing the number of customers leaving the business will help you know how it’s affecting your revenue. This information will allow you to create a long-term strategic plan to improve your company and reduce the number of churned customers while acquiring new customers.

How do you calculate the churn rate?

Take the number of lost customers and divide that by the number of total customers at the beginning of a specific time frame (e.g., one year, one quarter, etc.) and multiply that by 100.

For example, if your company had 1,000 subscribers at the beginning of the year and lost 150 by the end of the year, you would divide 150 by 1,000. That equals 0.15. Multiply that by 100 and you have a 15% annual churn rate.

150 ÷ 1,000 = 0.15 x 100 = 15

That percentage is very large. After determining your churn rate, your next step is to find a way to reduce that number. You want to increase your number to over 1,000 customers, not lose them.

The projected decline in subscribers will mean a drop of about $25 billion in cable subscription revenue plus associated advertising losses for the largest U.S. media companies.[2]

Many churn rate use cases can be solved with ElectrifAi’s Churn Mitigation machine learning model. The following are just a few examples:

  • Mature wireless market: saturation and slow growth rates
  • Substantial churn across both pre-paid and postpaid accounts with no clearly identifiable cause for churn
  • Determine specific marketing actions to successfully mitigate churn

By using applied machine learning, ElectrifAi can identify customer behavior that leads to churn. And that information is not theoretical. We use this information to help businesses solve real problems. In fact, we created a personalized churn reduction treatment for a Tier-1 global wireless telecommunications provider that saw impressive results.

What was the business impact for the client?

  • Reduced churn: Advanced 20+ bps annual churn lift
  • Revenue lift: Achieved $145MM/year in revenue lift or approximately $2/subscriber across the targeted customer base
  • Results A/B tested against existing efforts: Churn and lift are relative to existing campaigns

How does the churn mitigation machine learning model work?

  • Multiple models are configured to identify who, when, and how

         o  Who: Identify high-risk subscribers

         o  When: Identify likely time frame subscribers will churn

         o  How: Select retention strategy to best reduce churn

  • The model output drives highly targeted marketing campaigns

The following data sources and features are used to power the machine learning model:

  • Subscriber Data

         o  Usage and Consumption

         o  Billing and Payments

  • Product Information

         o  Handset Details

         o  Additional Features

  • Interaction Patterns

         o  Customer Service

         o  Marketing and Promotions

         o  Customer Experience

  • External Data

         o  Competitor Data

         o  Demographics Data

  • 4,000+ Signals give a 360o view across the entire lifecycle of every customer
  • Injected intelligence in customer behavior patterns
  • Automated campaign targeting through highly targeted and granular offers to address specific churn drivers

How is the model output leveraged?

  • Model scores individual churn drivers—cost, technical issues, old phone, etc.
  • Understanding individual churn drivers help customize treatment options for more effective mitigation
  • Reinforcement learning algorithms create a feedback loop—promotions adjusted as world dynamics change

Machine learning is currently playing a big role in many companies, some of whom may even be your competitors. Staying ahead of that competition means you need to act now.

According to Gartner, the number of CSPs investing in artificial intelligence (AI) technologies for improving their infrastructure planning, operation, and products will rise from 30% in 2020 to 70% in 2025.[3]

ElectrifAi is ready to begin your digital transformation. Our high-ROI, domain-driven machine learning solutions will help you increase your revenue, decrease your costs, and reduce your risk. We have 17 years of data science and analytics experience with over 2,000 customer engagements around the world solving real business problems.

With minimal operations and maintenance, you can achieve faster, better, and cheaper time-to-value through our large library of machine learning models. Our solutions work with third-party platforms or your own tech stack.

Let us know how we can help you achieve more with your data. Contact us today for a custom demo!

[1] MajorPay-TV Providers Lost About 5,120,000 Subscribers in 2020. Leichtman Research Group (LRG), March 4, 2021.

[2] Media executives are finally accepting the decline of cable TV as they plot a new path forward. Alex Sherman, CNBC, October 24, 2020.

[3] 2021:Emerging AI trends in the telecom industry. Liad Churchill, Customer Think, February 10, 2021.