It’s Time to Close the Door on Churn

BlogMar 01, 2021

Matt Smith

Remember when you were younger, and you were either coming or going from the family home and left the door open? No matter whether it was winter or summer, you no doubt would incur the wrath of a parent and the inevitable demand to “Close the door! You’re letting all the heat/cold out!”

Today, if you are a programmer and run a direct-to-consumer (D2C) video service, you are likely dealing with a similar situation. But instead of heating or cooling, the things escaping your service are revenue and subscribers.

Churn is one of the largest challenges for today’s D2C services. While growth among subscription services is better than ever (and increasing), retaining those subscribers and fully understanding how to satisfy them as customers remains a mystery for many.

The churn rate for OTT services grew to an average of 41% in Q1 2020, up from 35% in Q1 2019. I’m no statistician, but it sounds like that trendline is heading in the wrong direction. What can be done to mitigate this? Like a map to the proverbial treasure cove or finding the fountain of youth, solving this challenge will take persistence and a lot of data.

To level set, it is worth taking noting that may D2C service providers aren’t sitting on their hands while trying to retain their valuable subscribers. Far from it. Some are continually taking a close look at their user experience, ensuring that discovery and navigation of content is as frictionless and pleasing to the viewer as possible. Others continue to explore tuning the content recommendation algorithms that present additional shows to watch, based on a viewer’s experience and choices on their platform. Some are emailing their users, notifying them of new episodes of some recognizable show that the viewer may or may not find compelling.

But to understand user behavior anywhere online, one key element always emerges as the magic elixir: data. Data will always shed light on what the user is doing, where they’ve been, what they’ve interacted with and how long they did so. This data is instructive, but unless you know how to interpret it and what to do with it, it’s not nearly as valuable as it could be.

Enter machine learning (ML) and artificial intelligence (AI). Here, they become invaluable. However, AI/ML don’t help much unless they’ve been trained to interpret, analyze, and instruct users on how to apply the insights they produce. Many call this “training the algorithm” — teaching it to see trends in data, group similar data sets, and even recognize when something looks irregular compared historically to data sets it has analyzed before.

Let’s apply this logic to a couple of real-world examples.

Joe is a new subscriber to one of the largest name brand OTT services. He signed up last Sunday but hasn’t had the time to sample one single show. The service operator is using their data to analyze all their 1.5 million subscribers. Their platform has been trained to notify an internal team of new trial has inactivity. On Wednesday, the system sends Joe an email, saying they see that he hasn’t watched any shows. They offer Joe 50% off his first month’s subscription fee if he will activate the offer immediately.

Rather than letting a potential subscriber come and go without sampling a single show, the service has used the prescriptive elements of its data analysis platform to convert Joe to a paying customer. Sure, it won’t work all the time. But this case did help to reduce churn.

This same service has been working with a partner to analyze portions of their data to uncover scenarios like Joe’s above, and to also look at other areas of their business where visibility isn’t so great. They know they’re not running optimally and believe the data could tell some stories. After an analysis of some of their data, the algorithm has discovered that roughly 1,500 trial accounts have been set in the system with no expiration date. Further, these viewers have been active on the service for several months. The loss of revenue is calculated in the single digit millions of dollars. Similar to the home analogy earlier, leaving a door open was leading to previously undiscovered losses. At least in the family home, the cause and effect were understood.

So, what’s the moral of our story? Close the door. I know, very funny. For OTT service operators, the answer is to get a handle on their data and understand it. Once that’s achieved, unleashing powerful media-specific AI on that data is the fastest, most effective way to reduce churn and be more informed on their audience’s behavior.