How to reduce OTT churn rates with Machine Learning

Kirstin White | Fri May 12 2023 | Industry insights

How to reduce OTT churn with machine learning

Over-the-top (OTT) streaming services have become an increasingly popular medium for viewers to stream their favorite movies, sports, TV shows, and other entertainment content.

Statista estimates that VOD services will likely touch 1.636 billion users by 2027. With this bullish sentiment, it’s natural that the number of streaming platforms is growing.

While this means more options for viewers, the rising competition poses numerous challenges to broadcasters. It makes tackling OTT churn or paying customers canceling subscriptions significantly trickier than before. After all, it negatively impacts market share, revenue, and growth.

What broadcasters can do to keep growing

The answer lies in first understanding your customer base and why they leave. This is a crucial first step because you can’t improve what you don’t know. 

Once you have relevant insights, you can then take proactive steps to prevent churn from happening and maximize customer lifetime. But when you have so many subscribers, where do you even begin?

Why tackling OTT churn matters so much right now

OTT churn, or subscriber churn, is among the key performance indicators for OTT subscription services businesses to assess their health. This metric measures the rate at which OTT service subscribers leave a service. This may include leaving out of a voluntary choice or even involuntary churn due to external circumstances.

Now, customer churn is becoming a real roadblock. In 2023, Parks Associates evaluates it at 47%!


Tackling churn is super important because it directly leads to revenue loss from churned customers. It additionally counteracts the effort and expense required to acquire new customers, and ensures a smooth training and onboarding process, alongside provisioning, and supporting them. 

Both hit profitability. 


Therefore, controlling churn is critical to driving growth for OTT platforms.

How to calculate churn?

You can calculate OTT customer churn as the percentage of streaming customers who cancel their subscriptions or stop using the video service during a given period.

Customer Churn Rate Formula

Formula to calculate churn rate

Why do users churn from video streaming services?

Of course, there can be several reasons for streaming service subscribers to churn.

Involuntary churn usually happens due to declined payments from expired credit cards or other payment failures. Reducing involuntary churn could therefore be tackled with an optimized payment and checkout solution.

On the other hand, some of the most common reasons for voluntary churn include:

  • Content-related concern: If the platform content isn't interesting enough for the users, or if they've binge-watched whatever they like, the likelihood of churning increases.
  • Pricing: Pricing is a significant factor that influences user churn. With rising inflation, competitive pricing is gaining importance as people increasingly consider their expenses. They are quick to pass if they feel the subscription fee is too high or they're not getting enough value.
  • User experience: A poor user experience, such as a messy interface, slow loading times, or even poor customer service, annoys users and motivates them to turn to other platforms instead.
  • Subscription fatigue: Alongside specific reasons to churn, there's an overall force of subscription fatigue operating across the economy. In general, customers are feeling tired of paying for multiple subscriptions. This is feeding into churn rates for all subscriptions, including streaming services.

Machine learning can turn OTT churn into an afterthought

Data science and machine learning present a robust and objective way for OTT businesses to go into the root causes of churn.

Machine learning is essentially an application of AI that can analyze and learn from enormous amounts of information in a fraction of the time it would take a human brain to complete the same task.

For OTT companies, the use case for machine learning is the same: AI will digest data and turn them into relevant churn metrics for OTT players.

By simplifying the analysis of vast amounts of subscriber data, modern machine learning tools empower broadcasters to identify:

  • Exact customers' patterns
  • Customers preferences
  • Behaviors at risk of churn
Combining these insights puts you in an advantageous position to take meaningful steps for churn reduction.

With that, let’s explore actionable ways to leverage data intelligence to overcome OTT subscription churn.

Machine learning can decypher OTT churn for you. Here’s how

First and foremost - when you start with churn management, you must realize that it seldom occurs without warning.

Voluntary churn is always preceded by signs like a fall in customer engagement or a rise in inactivity. You may also find sudden overconsumption of your content, usually followed by unsubscribes (commonly called "burn before churn.”)

Machine learning and artificial intelligence are helpful here. It helps to identify these patterns from data and act on them timely. If you use the right data analytics tools or churn prevention software, you can identify the warning signs highlighting at-risk customer segments. With this foresight, you can quickly re-engage the customer with specific churn management strategies before they churn.

For instance, if you find your customers turning price-sensitive, you can offer coupon discounts to entice them to stay. If a growing amount of customers are upset with the service and provide negative customer feedback, make sure you have the capacity to 1. hear the complaints and 2. deal with the root cause.

Churn prevention tools like ChurnIQ provide dashboards with metrics highlighting your churn rates and the associated subscriber cancellation reasons. 

Cleeng churn metrics dashboard

An example of ChurnIQ’s retain dashboard

It presents a clear picture of what’s happening at your baseline level. You also get a synopsis of upcoming renewals and an analysis of renewal performance.

The analytics tool uses a machine learning algorithm to track accurate churn probability scores with a breakdown of the number of at-risk subscribers and their reasons for leaving your platform.

Here are five ways you can use ChurnIQ’s data intelligence capabilities to address churn:

1. Find out how urgent your churn problem is

Example of the "Subscription churn risk" graph from the Churn prediction dashboard

The foremost step to counter churn rates is knowing your exact risk status. The Subscription Churn Risk dashboard uses the graph above to illustrate the immediacy of your churn problem among the existing subscriber base. This information makes it an excellent starting point to assess how severe your churn risk is.

For example, the graph above shows that 85% of the at-risk subscribers display “Very High” risk patterns. From this, you can deduce how quickly you need to start addressing your at-risk subscribers and with what intensity. Intuitively, high risk demands immediate action with greater force. On the other hand, lesser-risk situations can afford lesser resources to be allocated immediately.

2. Find out why customers are churning

Once you know the "what," the next important thing is the "why ." After all, can you come up with a solution without knowing why?

Example of the "Detected Churn Causes" graph from the Churn prediction dashboard

ChurnIQ’s algorithm analyzes the subscriber data and detects churn causes, ranging from past behavioral patterns to current ones. It also notices streaming and pricing plan issues, causing dissatisfaction and churn. This understanding puts you in a position to intervene and address the exact cause.

Campaign Example A

In the graph above, we see that the primary churn causes for this broadcaster are:

  1. Not engaged with the content
  2. Payment is unreliable.

This is a precious insight as it can point you to do the following:

  • Better promotion of quality content OR invest in new, more popular content
  • Launch an email campaign to remind customers of outdated payment details

3. Know which customer segment to prioritize

Once you know the severity of your churn problem and the primary causes of this, it's time to get specific.

It's nice to know the general churn status of your full customer base, but it's not that useful unless you can precisely locate the at-risk customers. An excellent way to do that is by creating segments based on churn risk levels and churn risk causes. You can use ChurnIQ to segment Very High/High Risk of Churning this month. Then, coupling this with the detected churn cause, you can narrow down your audience and develop a campaign to compel that particular segment to stay.

Campaign Example B

Suppose you notice that high-risk subscribers are often linked to price complaints. Whereas low-risk subscribers are often linked to a poor customer experience. This should make it clear that prioritizing the price-sensitive group is a priority. Next, you can develop engaging pricing strategies or coupon campaigns that offer attractive discounts.

In the image below, you can see several examples of crucial segments recommended by the ChurnIQ dashboard.

An example of key customer segments to target, presented within the ChurnIQ dashboard

Using data from over a decade in the video subscription industry, we built a selection of subscriber segment suggestions, all available in the Cleeng Dashboard. This guides you toward the most logical ways to segment your customer base to address the different churn problems best.

4. Follow best practices for effectively targeting at-risk subscribers

Once you've located your at-risk customers, the next step is creating a plan to re-engage them and secure them as loyal subscribers. This is an essential step because a perfectly shaped message will be the difference between a retained and lost subscriber.

To simplify this, we used historical subscriber data and industry best practices to create a catalog of recommended retention actions. These actions are broken down per stage of the subscriber journey (Register vs. Experience vs. Winback etc.). Within each stage, the goal outcomes and recommended segments and actions are defined.

The eight stages of the subscriber journey

For example, see below the recommendations for the Retain stage:

The Retain stage of the Segments Catalogue (full catalog available within the ChurnIQ dashboard).

Campaign Example C

If your Prediction dashboard shows "High Risk" associated with "Poor experience," the Segments catalog recommends an apology with a discount to show the customer that the error has been addressed and that their loyalty is valued. 

Campaign Example D

If your Prediction dashboard shows "High Risk" associated with "Frequent churner," you may find the customer burning through your content quickly and then canceling. In this instance, we recommend focusing on any upcoming content. Clearly, the content attracted the customer in the first place, so remind them that there is more to come.

No matter what your churn concerns are or what stage your customer is at, you can access a selection of recommended actions to give you the best chance of retaining them.

Finally, you also need to implement plans, considering the time element.

5. Set up automated retention campaigns easily

If your execution of retention campaigns takes too long, you can lose customers from your net.

For example, consider a user who signed up for a hockey season pass with an end date in 2 weeks. If you wait for the season to be over before you convince them to spend longer on your platform, you could lose them for good once the season ends.

Therefore, acting quickly is essential. To that end, ChurnIQ Segments offers easy, automated campaign creation—no more hours of tiresome planning and list management. Through our integration with Looker, you can connect to any marketing tool. Simply set up a segment once, set the campaign frequency, and have it run automatically in the background.

Use machine learning to beat OTT churn proactively

As the OTT landscape competition is at an all-time high, you must pull up your churn management performance to stay in the game and grow steadily. To that end, the best way to go is by being objective about the approach.

Machine learning presents a powerful data-backed way to identify at-risk subscribers, understand their behavior patterns, and take proactive measures to reduce churn.

Using Data Effectively for Next-Level Subscriber Retention


Cleeng SRM Product