At Wicket Labs, we believe you can when it comes to subscribers and your service. That’s why we’ve developed a new way of thinking about user health, our Customer Happiness Index, or CHI Score. This concept expands beyond traditional consumer satisfaction indices, which are typically based on interviews and surveys. Instead, we’re using data + artificial intelligence to uncover unique, predictive, and most importantly, actionable insights into your customer cohorts.
Our approach is to leverage one of the core value propositions of the WIcket Scorecard which is to build a data set from many sources. Using data harmonization and applying machine learning to this data allows us to look at consumer behavior across many dimensions. CHI derives correlations between each factor and churn probability, which, not surprisingly, grows inversely to happiness.
This approach gives our customers powerful tools to work with. Every subscriber receives a CHI Score, an aggregate of scoring across four weighted categories used to track dozens of features.The categories we use are:
This includes features like subscription length, acceptance of price increases, consecutive months with usage, renewals, etc.
Here we’re focused on the number of series watched, types of content viewed, whether a user is a fan of a series, whether they watch new content upon release, etc.
Features in this category focus on frequency and intensity of viewing. In other words, how often does a user engage with the service, and when they do, how much do they watch? Are they using multiple devices? Do they watch in a discernable pattern?
This category can include data about customer service interactions, correlations between primary viewing device and overall ratings for the app on that device, percentage of session time spent watching content vs. browsing for something to watch etc.
The Power of CHI
With a unique CHI Score for every subscriber, we can unlock several new insights for our customers regarding their users’ current and future behavior.
As alluded to above, there is a direct relationship between CHI Score and churn probability. A glance at the CHI Score for a given subscriber is a good indicator of health and likelihood that they will be around next month and beyond. The power in this comes from understanding why someone is satisfied, or likely to churn and then exploring ways to act in either scenario.
To make this concept more actionable, CHI Scores can be viewed in multiple ways. You can pick a dimension like primary viewing device, and look at CHI Scores for each device broken out by category, enabling you to identify what is driving churn risk, isolate the issue and have a meaningful impact on retention. The Scorecard also allows you to look at CHI scores across other dimensions like sales channel and pricing plan to uncover and address additional areas of concern or opportunities to improve your experience.
To take a deeper look at CHI on a particular device class, we illustrate the variation of user scores vs. ideal score across each category, and how the mean score for each device compares to the overall mean score. This is a great way to identify areas of improvement (below average) and where they might be most effective (concentrated distribution of scores).
Churn Prediction & Personas
In addition to exploring CHI through the lens of devices and channels, you can visualize your subscriber base relative to churn prediction. This would be very difficult to understand and act on if every subscriber was modeled individually (imagine thousands or millions of dots!). To simplify, we’ve introduced the concept of personas, using machine learning to cluster sample sets of users with similar scoring attributes across the dimensions we measure. This paints a clear picture of behavioral patterns, which in turn makes taking action more effective.
The churn prediction charts illustrate personas and devices with their CHI Score and attendant churn probability. On the chart below, personas are broken out across devices, which can show interesting variation, and sometimes personas that only exist for certain devices.
On the following chart, personas are shown broken out by CHI category so that it’s easy to see how each category contributes to the overall score. To understand what actions might make sense for a Persona it’s important to understand the composition of the CHI Score.
To make this easier and more actionable, we have developed a user-friendly description of what is driving the CHI Score of each persona, and reasonable actions to reduce predicted churn. This is interpretable machine learning at work, and the promise of it has us very excited.
Combining this kind of insight with straightforward tools to get the list of subscribers within a persona, sets the stage for more targeted, effective communication with customers, improved product enhancement, and better content promotion.
We’re excited to put this functionality in front of customers, and our first step is to introduce it in a new section of the Wicket Scorecard called the Lab. It’s a place where we’ll debut new functionality letting customers interact in advance of it going live on their Scorecard. This allows us to gather feedback and react to it, resulting in a stronger product for our customers.
Contact Us to learn more about the CHI score and how it may reduce your churn.
Tags: audience lifetime value • churn • content engagement • lost customers