Customer intervention matrix: a new look at customer churn scores
Customer Churn Predictive Models are becoming table-stake nowadays. It's a good practice to look at the evolution of predicted churn scores when considering strategic sales and marketing customer retention initiatives.
One could simply look two churn benchmarks in two different periods for the same customers.
Sometimes it makes sense to look at the retention score over the past year and churn score calculated based on current year interactions. Let's call the first static score because it is calculated one time at the end of the year, and the second dynamic score because it can be calculated on a daily basis, comparing YTD (year to date) features.
In the education business, you can measure churn based on the number of digital learning platform student activations generated indirectly by college professors. In the current year, we may see no or very little activations related to an instructor (we categorize these customers as "pending" as the verdict on their retention will be known at the end of the year). Nevertheless, we can predict the current year churn scores based on the use of the digital platform or the number of service tickets.
The pharmaceutical industry can fit the same framework described above, think of doctors as instructors, and patients as students.
The framework of a static and dynamic score doesn't necessarily need a business model with "influencers". Amazon could very well compare my past year retention or churn score based on my web and mobile interactions, and yes sales (guilty as charged!), to my current year activities.
Assuming that we now agree on the usefulness of looking at the evolution of churn scores, utilizing at least two data points, how do we present them in a way that is actionable for the Business? Which is always the challenge for machine learning models, isn't it?
Introducing the Customer Intervention Matrix
You can interact with a sample Tableau workbook I published on Tableau Public below.
Tableau Public
If you set the thresholds above 50%, you will see a “gray area”, which is the difference between the churn score threshold and the retention score threshold. Each bubble is a customer, the bubble size represents the Customer Lifetime Value to Date.
The blue bubbles indicate potentially Saved Customers, They went from high churn to low churn scores.
One could simply look two churn benchmarks in two different periods for the same customers.
Sometimes it makes sense to look at the retention score over the past year and churn score calculated based on current year interactions. Let's call the first static score because it is calculated one time at the end of the year, and the second dynamic score because it can be calculated on a daily basis, comparing YTD (year to date) features.
In the education business, you can measure churn based on the number of digital learning platform student activations generated indirectly by college professors. In the current year, we may see no or very little activations related to an instructor (we categorize these customers as "pending" as the verdict on their retention will be known at the end of the year). Nevertheless, we can predict the current year churn scores based on the use of the digital platform or the number of service tickets.
The pharmaceutical industry can fit the same framework described above, think of doctors as instructors, and patients as students.
The framework of a static and dynamic score doesn't necessarily need a business model with "influencers". Amazon could very well compare my past year retention or churn score based on my web and mobile interactions, and yes sales (guilty as charged!), to my current year activities.
Assuming that we now agree on the usefulness of looking at the evolution of churn scores, utilizing at least two data points, how do we present them in a way that is actionable for the Business? Which is always the challenge for machine learning models, isn't it?
Introducing the Customer Intervention Matrix
You can interact with a sample Tableau workbook I published on Tableau Public below.
Tableau Public
The business will define the retention and churn thresholds, in this dashboard they are set to 70% but you can change them.
If you set the thresholds above 50%, you will see a “gray area”, which is the difference between the churn score threshold and the retention score threshold. Each bubble is a customer, the bubble size represents the Customer Lifetime Value to Date.
The purple bubbles indicate Unhappy Customers. They were unhappy customers last year and they continue to be unhappy.
The orange bubbles indicate customers whose churn score is deteriorating
The red bubbles indicate customers that were happy until last year and now they show signs of predicted discomfort (From Happy to Unhappy). They should be the primary of the Customer Retention program.
The blue bubbles indicate potentially Saved Customers, They went from high churn to low churn scores.
The green bubbles indicate potentially Happy Customers, They maintained their lowe churn scores.
If you now add your Sales Organization and geo-location dimensions, you can drill down and present a Customer Intervention Matrix by Region, District, State, City, Rep, or any combination of thereof.
In conclusion, you can use the Customer Intervention Matrix as a bird's eye view of your customers "changing mood" to help you defining priorities and tailored programs/messages to the different "segments" defined above. I am a big fan of dashboards that can be used by senior leadership and salesforce operations at the same time, and this dashboard one seems to serve the dual purpose well.
Thank you Raj! I am glad you enjoyed it
ReplyDelete