Predict & Reduce Customer Churn using Machine Learning

As promised, here’s the second article in the series of customer case studies and use cases that I’d like to share for learning purposes.

This post covers the application of embedded machine learning to prevent customer churn. I’ll be using a sample customer case study to define the need, problem, solution, approach and key benefits. We will also go over customer’s improved business outcome realized by leveraging the Tellius platform.

Written by: Chris Walker,

Head of Product Marketing at Tellius

What is customer churn?

Customer churn is the term used when an existing customer stops using a company’s services and/or stops buying their products. In other words, the customer chooses to cut his ties with the company. A few types of churn can’t be avoided – e.g. churn due to death. Such churn is categorized as non-addressable churn. In this post, we will focus on the other class of customer churn – addressable churn.

Churn is a huge problem for companies as it contributes to a reduction in the revenue. At the same time, it puts additional pressure on teams to make up for the lost revenue. One of the ways of doing that is by acquiring new customers. But as we all know, acquiring customers is hard and expensive. Also, customer acquisition costs will further put a dent to the overall revenue of the company.

There are some other indirect effects with customer churn as well. Churned customers start a new relationship with the company’s competitors. They might at times even try to take away other loyal customers with them.

One of the recent research note from PWC concluded that:

“Financial institution will lose 24% of revenue in the next 3-5 years, mainly due to customer churn to new fintech companies.¹”

For the reasons mentioned above and many more, reducing customer churn is a key business strategy for most businesses. Even if reducing customer churn is not a company’s strategic objective, it is definitely in their best interest to retain any and every customer possible.

So now that we understand what’s customer churn and why it needs to be addressed with urgency, let’s look at various approaches companies take today towards solving this problem.

Typically, tribal knowledge and/or biased judgment is used to identify customers that are likely to churn. In such scenarios, you are most likely to end up targeting customers that aren’t going to churn. Giving such customers retention incentives adds unnecessary operational costs. Also, you risk not reaching out to customers that might actually churn. Hence, you end up losing their business.

So, that brings us to the meat of our discussion – what’s the right way to prevent customer churn? Let’s talk about a specific problem one of our customers recently put in front of us.

Can AI-driven Analytics help predict/reduce customer churn?

So we will dive deeper into a case study that lays out the need, problems, solution, approach and, benefits for addressing. Data, names, numbers and a few other items have been anonymized/randomized to protect confidential information and respect their privacy.

Business: Multinational Bank.

Industry: Financial Services.

Need: Large banks are losing customers to fintech startups. Startups have taken a data-driven approach towards acquiring, serving and retaining customers. Large banks need to innovate to retain customers. They need to proactively reach out to customers who are at a risk of leaving.

Problem: Our leading multinational bank focuses on private banking. The bank is facing increased customer churn over the past period due to increased competition in the market. The bank possesses large amounts of customer data but does not leverage it effectively.

The bank needs to identify customers that intend on leaving so that the marketing department can provide targeted-oriented advertisement means to retain these customers.

Furthermore, the bank wants to understand the influencing factors for churn so they can be more proactive towards addressing such issues instead of just reacting after the fact.

The real problem & need is to reduce customer churn, stabilize the business and increase profits.

Solution: With Tellius, teams at the bank started discovering characteristics that caused customer churn. These patterns were revealed automatically by the underlying sophisticated machine learning algorithms. Furthermore, customer churn profile identified high-worth risky customers. Proactive campaigns are now being run at regular intervals to ensure that they can retain such customers before they leave.

Approach: Here’s the three-step approach that worked really well for the bank:

• Identify and connect to the right set of data: Data on customers including demographics, assets, credit scores, complaints, accounts, tenure, etc.
• Self Service data transformation: Unimportant columns were removed, Missing/Invalid data fixed using quick options, Data from various sources joined together.
• Single Click Automated Insights: Insights were discovered that generated customer churn profile. Characteristics of such customers and their propensity to churn were predicted. This predictive model is now used to predict and reduce churn by proactively reaching out to them.

Key findings discovered by using Tellius platform

Interesting insights were automatically surfaced by the Tellius Genius AI engine. These discoveries amazed the multinational bank’s teams. Here are a select few:

1. Customer Age, Number of Complaints, Account Balance and the Number of Products they use have the highest influence on whether a customer will leave or not.
2. Customers in the younger age band and with low account balance are more likely to close their accounts.
3. Female customers in the 25-35 age band are more likely to quit.
4. Most of the customers that churn were not using any product besides the bank account.
5. Credit Score doesn’t have a high impact on customer churn.

Summary

As we saw in the case study above, one of the top multinational banks was able to utilize Tellius’ platform to understand why their customers were leaving. They can now proactively address churn. Note that all of the insights were discovered automatically without spending 100’s of hours on manual data discovery or without writing a single line of code.

Tellius platform utilized it’s underlying machine learning algorithms to answer specific business questions and discover hidden insights in the data.

Teams at the bank can now clearly understand profiles of customers that churn. They can reach out to existing customers that fit such profile and take preventive steps to reduce churn.

The multinational private bank can now:

• Adjust its overall strategy to reach out to high-value customers that are likely to churn.
• Reduce cost of incentives as they have a better understanding of churn profiles.
• Improve profits by retaining customers.
• Achieve higher ROI – time and cost/savings, and increase in revenue.

Hope this post was useful to understand how advances in machine learning can be applied to solve real-world problems. Stay tuned for some more use cases/customer case studies!

Please feel free to reach out directly for feedback, questions, and comments:

hardik.chheda@gmail.com twitter.com/hardikchheda linkedin.com/in/hardikchheda

Note: Any references or images used in the post are only for informational purposes. The ownership and copyright remain with the original creator. Please let me know if any content violates your rights – I will take it down as soon as possible.