ElectrifAi
July 8, 2021

How to Mitigate Churn Rate with Machine Learning

Most businesses experience churn rate: particularly, subscription-based businesses in which the most crucial factor is getting customers to remain subscribers. Acquiring new customers and getting them to subscribe for a set timeframe is the initial goal. Losing that customer, however, ends the subscription and the rest of that customer’s on-going value is lost.

What is churn rate?

Churn rate is the rate at which customers stop doing business with a company. To be successful, the company’s growth rate (number of new customers) must exceed the churn rate.

How is churn rate calculated?

To calculate churn, take the number of customers who subscribed to your service and divide that by the number of customers who cancelled their subscription. For example, if you acquired 60 new customers in June and you had 40 customers already, you had 100 customers. If 10 customers canceled the subscription at the end of the month, you had a 10% churn rate for June.

Acquiring new customers is very costly compared to retaining customers.

It depends on the industry, but it typically costs five times or more to acquire a new customer compared to the incentives you give to retain existing customers.[1]

“If you’re not convinced that retaining customers is so valuable, consider research done by Frederick Reichheld of Bain & Company[2] (the inventor of the net promoter score[3]) that shows increasing customer retention rates by 5% increases profits by 25% to 95%.” - Amy Gallo, Harvard Business Review

Reducing the amount of churn with better customer retention rates can bring in substantial profits for your business.

How do you mitigate churn rates with machine learning?

Companies have been putting effort into churn mitigation even prior to the advance of machine learning. Traditionally, a one box fits all type of marketing approach was used where incentives to stay with the company was given to a large group of people.

With machine learning models, however, you’re able to really look at each customer as an individual and understand their probability of churn. The more you know about a customer, the more nuanced you can get in your approach. You can now rank order customers individually and target the different churn possibilities with higher confidence of success.

How does ElectrifAi help to mitigate churn?

At ElectrifAi, we excel not only in understanding who is likely to churn, but why people churn. We can determine the various factors driving those customers to churn. It’s not only about using machine learning models to score outcomes. You must go one step further and try to pinpoint the customer’s pains or risk factors.

If businesses where churn is an issue continue to merely give people incentives to stay, those businesses never address the fundamental problem driving churn. Over time, money is just spent. No returns are gained because customers continuously churn.

ElectrifAi addresses the driving factors of churn.

Our clients use our machine learning models to great effect. Out of our vast library of prebuilt models, we can use several models together to create a customized plan for your specific needs. Churn mitigation may be the first step.

Our churn mitigation model is a nonlinear model with many interaction factors. Oftentimes it’s difficult to decipher the driving factors of churn. We use various techniques to identify at each customer level: what are the driving factors leading to a high probability of churn?

Let’s break down a few examples.

Example 1

The churn mitigation model predicts a high probability of churn for a particular customer, but it also gives you an opportunity to retain the customer. If the customer had a phone upgrade three years ago and has not upgraded, a competitor could potentially send out a phone upgrade promotion and you would lose that customer.

Say the same customer adds a new person to their account. That’s a lifetime event with a high probability of churn because now the subscriber is going to shop around to see if there are better deals. To get in front of the competitor, it’s important to take action first.

You can incentivize the customer to stay by offering a family plan or to get an additional phone for free with the new line. But these incentives are only useful for the company if they work to solve the problem of churn mitigation. The company only sends out incentives if they know the customer is very likely to shop around, thus not sending out needless incentives to satisfied customers.

Example 2

Another approach to churn mitigation is to consider other factors such as customer sentiment.

Applying our machine learning model Opinion Mining to call center logs, we look at not only customer sentiment but we are able to really understand: what are the aspects of the business subscribers are concerned and talking about? The model brings that understanding to a higher level so we can understand the crucial pain points.

In this example, customers might be calling the call center to complain about dropped calls. Just looking at one or two customers, you’re not able to really identify that this is a large-scale problem. But when you apply our opinion mining to thousands of customers, then you aggregate these problems to find the ones most in need of immediate attention.

A particular neighborhood may be experiencing elevated numbers of dropped calls. Now you can connect that problem to the network data and verify that you know a certain portion of the network has a problem. You can tie that back to the churn probability and know those subscribers are likely to churn.

Providing incentives can attempt to mitigate the issue. But you should also look at ways to upgrade the networks in that neighborhood or to improve that customer’s experience so that this factor doesn’t become an issue for future churn.

Additionally, there could be other customers in that neighborhood who might have lower churn possibilities. Now, since you’re addressing these customer experience factors, those customers churn probability will stay low and should not come up as a problem in the future.

Conclusion

ElectrifAi machine learning models give you an advantage over the competition.

The key here is that our machine learning models allow you to identify the truly risky customers who are likely to churn. We not only identify those customers; we are able to go one or two steps further and understand the drivers of that churn. Understanding how it ties to other customer experiences and trying to address those pain points is the first step to mitigate churn today so that future churn never appears.

Interested in learning more about how churn mitigation? Contact ElectrifAi today and let us create a plan that meets your specific business needs.

[1]Harvard Business Review, https://hbr.org/2014/10/the-value-of-keeping-the-right-customers

[2]Frederick Reichheld of Bain & Company, https://media.bain.com/Images/BB_Prescription_cutting_costs.pdf

[3]Net promoter score, https://www.netpromoter.com/know/

You have data, we have solutions!

Find out what ElectrifAi has to offer by filling out the information below.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.