How Streaming Platforms Can Harness Predictive Analytics for Better Retention
The streaming video ecosystem is shifting focus from maximising subscriber growth at all costs to maximising profit. The goal for over-the-top (OTT) businesses has become less about getting new customers in the door, and more about keeping the most high-value customers satisfied. Whether it’s a subscription or advertising-based business, average revenue per user (ARPU) growth is positioning itself as the most important metric for any direct-to-consumer (DTC) company. However, in an oversaturated market where consumers can hop from platform to platform with ease, securing a portion of your audience’s attention and wallet has never been more difficult.
Streaming platforms are rising to the challenge by turning retention into a data-driven art. DTC businesses have a treasure trove of first-party behavioral data at their disposal, and on a foundational level, they are analyzing this data to paint a clearer picture of why people are coming to a service, how they interact within the service, and why they might be leaving.
In recent years, however, the advantage of having first-party data has expanded far past basic data analytics, giving way to the ability to create far more advanced predictive machine learning (ML) models. After all, when it comes to data, if we can measure it, we can probably predict it.
By harnessing the value of advanced predictive analytics, streaming platforms can ensure that they’re using all the information at their disposal to make the most informed decisions possible about marketing, packaging, and content, thus leading to more efficient acquisition, retention, user engagement, and monetisation.
The Power of Prediction
Data scientists can create ML models that output propensity scores that indicate a user’s likelihood to commit a certain action; for example, their likelihood to churn, stream content, or buy pay-per-view. These machine learning algorithms are constructed by joining both first-party data (like streaming behavior and transactional activity) with third-party data (external customer attributes, country-specific features, etc.), resulting in attributes like minutes spent streaming on-demand vs. live content, location, device, how long a user has been a subscriber, and/or any number of additional factors. Every DTC platform will leverage different data points depending on the specific nuance of the platform, as well as what distinct action it’s trying to predict.
The beauty of leveraging machine learning in the streaming ecosystem is these models get better and better as more data is collected. As users continue to stream and transact, the algorithm can assess the accuracy of prediction and improve based on these learnings. This allows for more precise segmentation and, ultimately, more useful predictions.
Using Data to Engage Viewers and Mitigate Churn
There is no shortage of use cases when it comes to prediction. Below, we’ll walk through two areas where DTC businesses can derive value from these practices: marketing and user experience.
- Targeted Marketing
It’s clear that real time data analytics helps businesses better understand their audiences and know how to message to them but predictive data algorithms enable businesses to get even more calculated when segmenting audiences. If a streaming platform knows that a certain group of users is likely to churn based off an ML model, it can communicate to those users differently than it would to users not likely to churn. For example, it could offer those users a complimentary month of the service, buying the business 30 more days to communicate directly with the user and reestablish the value of the platform. Alternatively, it could offer these likely-to-churn users a financial incentive, such as an enticing discount on an annual subscription. By only communicating a discounted offer to those most likely to churn, the business can save a substantial amount of revenue they may have lost by offering that financial incentive to users who were not at risk.
- Personalised User Recommendations
Personalised content recommendations can increase engagement within a platform, resulting in habit formation that helps mitigate churn. By leveraging historical user event data, platforms can suggest new content to pull the user in. You watched The Last of Us? Check out these four similar shows. Liked The Fabelmans? Here are three other movies directed by Steven Spielberg. Using machine learning, platforms can derive learnings from the entire population of viewers and personalise the user experience accordingly.
Recommendation engines are built off predictive models to determine a user’s likelihood to stream a certain piece of content in the future. For example, if a UFC fan has watched every Conor McGregor fight and interview they could find on UFC Fight Pass, there’s a high likelihood they’ll watch another. UFC Fight Pass could then proactively recommend new or undiscovered McGregor content based on these models. Using machine learning algorithms to build content recommendation engines is an effective way to increase user engagement, which in turn can help to lower churn rates.
The Future of Streaming is Data-First
Sustainable user engagement and churn mitigation will continue to be both an art and science for streaming platforms. As the industry continues to shift from maximising subscribers to maximising high-value user ARPU growth, the platforms that harness the power of data and machine learning will be best positioned to cultivate an engaging—and profitable—experience for users.
[Editor's note: This is a contributed article from Endeavor Streaming. Streaming Media accepts vendor bylines based solely on their value to our readers.]
Related Articles
Traditional, controlled broadcast environments are being replaced by dynamic, complex ecosystems where multiple applications and services must coexist harmoniously. The challenge is to maintain the quality of a provider's main service, like linear TV with dynamically inserted ads, while accommodating the diverse requirements for and potential interference from third-party applications. Yoann Hinard of Witbe discusses the challenges faced throughout this process, including remote monitoring, dynamic ad insertion, and QC testing across devices.
26 Mar 2024
This article by Lindsay Van Kirk of Operative spotlights recent developments in revenue management technology, and how cloud-native and modular approaches are enabling media brands to unify their existing systems and deliver faster ROI. Kirk offers practical insights and advice for media professionals navigating this dynamic landscape.
14 Nov 2023