Fighting Churn with Data: How it Works

FightChurn_MentalModel1

What is this about, Fighting Churn with Data?  I’m not an artist (disclaimer), but I drew a picture AKA “the mental model” for the book.   My publisher told me that it’s okay to draw pictures on a napkin and they’ll have a designer touch it up later, but I drew stick figures in PowerPoint instead (believe me, you wouldn't want to see me make even a napkin drawing...)   If you work in an organization that offers a subscription product with recurring user interactions, your situation probably looks something like the one shown on the left side of the mental model drawing.   The right side of the drawing is what the book is about.  The goal is hiding down there on the left hand side the end of the flow..

 Why You Are Here  (The Goal)

A primary goal for any product or service is to grow.  Growth occurs when new customers are added through marketing and sales, but when customers leave it counteracts the growth and can even lead to contraction.  For those in the business of providing such services, customers quitting a service has become known as “churn”.  Most service providers focus on acquisitions, but to be successful a venture must also work to minimize its churn.  If churn is not addressed in an ongoing and proactive way the product or service won’t reach its full potential. It may come as a surprise, but despite the wide variety of products and services with recurring user interactions and subscriptions, a single set of techniques are applicable when using data to fight churn.    This skills of fighting churn with data are effective using data in any kind of service with recurring user interactions.

Customers sticking with a service can also be framed in a positive sense, if you prefer to see a glass as half full.  In that case people talk about “customer retention”. Reducing churn is equivalent to increasing customer retention and the terms are interchangeable to a large degree.  When the goal is stated as retaining more customers for longer, it is clear that in addition to “saving” customers who are at risk of churning there should also be a focus on keeping customers more engaged generally. There is even the possibility “upselling” the most engaged customers to more advanced versions of the service, typically for more money.  So saving churns, increasing engagement and even upsells are all important goals for services with repeated customer interactions. The difference between these is a matter of the area of focus and not the strategic intent, and the posts in my blog and the forthcoming book will give you the ability to address engagement and upsells with data as well.

The Usual Situation

Turning to the situation side of the mental model, the key ingredients here are:

    1. A Product or service that is used on a recurring basis by users interactions with the service.  Often the product is paid for on a subscription basis.
    2. Customers or Subscribers who are using the product
    3. The customers have (usually) entered into Subscriptions to receive the product or service. Subscriptions often (but not always) have a monetary cost associated with them.
    4. Subscriptions can be ended or canceled, known as Churn.   Some subscriptions must be renewed periodically, while others simply last until the subscribers chooses to churn.
    5. Information about subscriptions is being capture in some kind of database.  Typically a transactional database.
    6. When subscribers use or interact with the product or service these events are tracked in a data warehouse.  Typically the event data warehouse has less guarantee of data quality than the subscription database.

If your scenario is not quite like this but has some of the elements that’s fine too - there are other scenarios where Fighting Churn with Data also works, which is really very broad: Any product with recurring user interactions works  - so really the data warehouse is required, while the subscription database is not. What is described above is most typical.  I'll write more about those related situations in future posts.

Fighting Churn with Data

The right side of the mental model drawing shows how Fighting Churn with Data works.  The following describes each step in the progress:

  1. Subscription data is used to identify churns and create churn metrics.  The churn rate is an example of a churn metric. The churn database also allows identification of examples of subscribers who churned and who renewed and exactly when they did so, which are all needed for further analysis.
  2. The event data warehouse is used to create behavioral metrics that summarize the events pertaining to each subscriber.   Behavioral metrics are a crucial step that allow the events in the data warehouse to be interpreted.  To people trained in data science this is often called “feature engineering”.  But because one of the themes Fighting Churn with Data is that the results must be communicated to business users, I will generally stick with terms that facilitate that communication. For this reason I avoid “feature engineering” which is easily confused with software product “features” and software engineering itself.  Instead I call this area "Behavioral Metrics", which is more specific than feature engineering for this area of work.
  3. The behavioral metrics for Identified churns and renewals are used together in a Churn Analysis.  The churn analysis identifies in a rigorous way which behaviors of subscribers are predictive of renewal and which are predictive of churn, and can create a churn risk prediction for every subscriber.
    1. Not shown in Figure 1:  At this stage, additional sources of information apart from the subscriber database and event data warehouse can also be brought into the analysis.   This includes either Demographic information about subscribers who are individual consumers, or Firmographic information about subscribers who are businesses.
  4. Based on their characteristics and risks, subscribers are divided into groups or segments that combine aspects of their risk level, their behaviors and any other characteristics found to be significant.  The purpose of these segments is to target them for interventions designed to maximize the subscriber lifetime and engagement with the service. Interventions can include email marketing, call campaigns, training and discounts.  Another type of long term intervention is changes to the subscription product or service itself and the information from the churn analysis is useful for this as well.

And that’s what it’s all about!  Of course, interventions and service modifications are the final crucial step to achieving the goal of lower churn and longer retention.  But note that unlike the data analysis techniques, interventions to influence subscriber behavior are generally very specific to different types of subscription services. There are some general principles to customer interventions which I will also write about, but there is no one size fits all intervention.    Thanks for reading!  If you find like this work please share it.