Questions	1. What are some case studies which can demonstrate the positive potential of data analytics on reducing subscription churn? 2. What are some risks to consider regarding the application of data analytics to customer understanding customer preferences? 3. What are priorities for consideration (organization-wide) in implementing customer analytics?
Turnaround Time	9 days (average 3 days for each question)
Delivered Content Description	24 content slides, with overview of methodology, an executive summary, insights on each of the 3 questions with detailed case studies and subscription data, and detailed analysis on steps to implement best practices, good governance and protective measures on the management of customer data.
Approach	Analysis focused on specific areas of concern to the client, with respect to data analytics and its role in supporting growth and reducing churn.  Top case studies of subscription-based revenue companies were selected based on client fit.  The report was developed using publicly available information, with additional research on company-specific data points.  Annual Reports and company website data formed a basis for much of the strategic and financial analysis, where available. This summary was developed with permission.  Parts of the analysis have been left out for confidentiality.
Client	Subscription-based Media Company

 

Big Data and Netflix

Often touted as a victory for data analytics, business schools frequently highlight the clever use of data analytics by Netflix to improve customer satisfaction and drive growth through personalized recommendations.  Netflix was faced with significant challenge in retaining subscribers and ensuring they remained engaged in the platform.  Although the streaming giant boasts a vast library of content, users were fatigued by the process of finding suitable shows and movies which were aligned to their viewing preferences.  As viewer boredom set in, churn turned upward.

For Netflix viewers, the main problem in using the platform was meaningful content discovery.  Netflix needed to find a way to ensure users could find good content.  As part of its solution, Netflix turned to data analytics and machine learning to develop a sophisticated recommendation engine. Netflix was already recommending content, in a de facto manner.  The menu screen itself was limited in size, and early attempts at highlighting attractive shows relied more on broad user preferences and rough profiles of customers based on limited demographic data. 

Given what we have learned about predictive analytics through the years, the solution to use the large customer viewing datasets is obvious.  Looking back, Netflix had only its imagination and a couple of key assets:

• Asset 1 – Customer Data: Netflix collected massive amounts of data on user behavior, including viewing history, ratings, search queries, and even the time and device used for watching content.
• Asset 2 – User Profiles: The company created detailed profiles for each user, capturing preferences and viewing habits, and overlaying this information with more personal details (birthdate) made available via credit card subscriptions, overlaid zip code census data.

The success in imagination was made manifest through a series of data filtering techniques used to develop more detailed recommendation algorithms.

The first, collaborative filtering, was a step through which developed recommended content based on the preferences of users with similar tastes in programming. 

How did Netflix define taste in programming?  A second step of content-based filtering was applied, whereby the content itself was analyzed with respect to characteristics such as movie genre, actors, and directors, to enable improved recommendations and better leveraging of user profile data.  By using machine learning algorithms, Netflix continuously improved its recommendations. These algorithms analyze patterns and correlations in the data to predict what content users would select.  That, combined with rigorous A/B testing on similarly-scoring content, enabled the Netflix growth story that we know today.

Improving Customer Retention with Analytics

As we know from history, the implementation of personalized recommendations had a significant impact on both users and Netflix’s long run sustainability potential:

• Increased Engagement: Users spent more time watching content on Netflix, as they could easily find shows and movies that matched their preferences.
• Reduced Churn: Improved content discovery and user satisfaction led to lower churn rates, as subscribers found the service more valuable and enjoyable.
• Growth in Subscriber Base: With reduced churn and a better reputation, subscriptions continued to grow.

What is demonstrated in this analysis is some often-touted lessons on data analytics and how to use this information to generate growth.  Wrapped up in a neat package and slide presentation, we learn – correctly so – that personalization is key for mass market business models like Netflix. This personalization was enabled by its wealth of user data dutifully collected over the years.  The successful outcome was also made possible by the machine learning technology that continuously refined and improved recommendations to viewers, effectively keeping the recommendations relevant as user preferences evolved.

Some Contrarian Perspectives – Potential Downsides of Analytics

While the case study presents itself as, literally, a textbook example of an effective application of data analytics, there is, as always, more to the story which adds a touch of gray to the otherwise black and white solution.

First off, critics have argued that an over-reliance on algorithms and data-driven recommendations can lead to a narrow viewing experience on the part of the users.  The recommendation system, by design, pushes users towards a limited subset of content, causing viewers to miss out on diverse or serendipitous discoveries.  This repetitive cycle of content over time brings users to experience, once again, boredom. Defeating its own purpose, the algorithm itself begins to show bias over time. 

As certain genres or types of content are more popular within specific user demographics, the recommendation engine can preferentially promote those types, further entrenching existing preferences and potentially marginalizing niche but potentially popular content. This bias is further reinforced by content homogenization. 

The data-driven approach to content creation and acquisition can lead to a narrowing of the creative forces that go into the development and production of new content.  By focusing on data trends and what has previously worked well, Netflix may invest more in formulaic content that is likely to be popular according to their algorithms, potentially stifling creativity and innovation.

On top of this effect, the predictive nature of recommendation systems reduce user autonomy by subtly influencing their choices. While users might feel they are making independent decisions, their viewing habits are significantly shaped by the algorithm’s suggestions, which can be seen as manipulative. At the end of the day, success of the algorithm depends on the success of data interpretation. 

The recommendation engine also depends heavily on the quality and interpretation of data. Poor data quality, incorrect data, or flawed assumptions in the algorithm can lead to inaccurate recommendations. There is also a risk that the algorithm might misinterpret nuanced user preferences, leading to dissatisfaction.

Taking these points together, there is an argument to be made about the long-term effectiveness of such systems of recommendation. Initially, they can – as the data rightly shows – boost engagement. 

Over time, however, people are people.  Viewers grow dissatisfied in seeing the same type of content again and again.  This subconscious impact brings about a feeling of boredom.  If the user is aware that their content has become seemingly stale, they will notice that the recommendations are repetitive or not evolving to highlight content that captures attention and inspires curiosity.

The Netflix Story – Big Picture on Big Data

While the Netflix example highlights the potential of big data and predictive analytics in enhancing user experience and driving business success, it is not without its drawbacks and criticisms. A balanced view acknowledges both the achievements and the limitations, considering the broader implications on user experience, privacy, creativity, and ethical considerations.

Nonetheless, the user experience is only part of the story.  Subscription churn is impacted by a wide range of factors that go well beyond the AI capabilities of its recommendation engine.

• In January of 2022 for example, Netflix rolled out another increase in its subscription service (previously increased in 2020) from $13 to $14 in North American markets for a standard plan.
• Netflix lost 700,000 in Russia following the outbreak of the Russia-Ukraine war.
• Changes to password sharing terms and conditions has also been a bone of contention between viewers and the company, and tightened restrictions have had an impact.
• Continued growth of online competitors, some of which are more specialized in capturing niche audiences, has put given viewers more choices, putting upward pressure on churn (an effect experienced by all platforms).

An Alternative Case – FarmVille’s Attempt at Data Analytics

An alternative business case, one which demonstrates the potential strategic danger of predictive analytics, involves the online social gave developer, Zynga, best known for games like FarmVille.  Like Netflix, Zynga used big data to help reduce churn, but faced significant challenges along the way. At its peak, Zynga was a leader in the social gaming space, primarily through its integration with Facebook.

The company’s business model heavily relied on user engagement and in-game purchases. To maintain its revenue stream. As competition increased and the initial novelty of social games began to fade, Zynga faced declining user engagement and rising churn rates. The company needed to leverage big data to identify patterns in user behavior and develop strategies to retain players.

The Zynga Solution

Zynga invested heavily in big data analytics with the goal of understanding player behavior and predicting churn. They collected vast amounts of data on how players interacted with their games, including gameplay patterns, frequency and duration of play sessions, types and timing of in-game purchases, Social Interactions between players, and the all-important customer feedback analysis. Zynga attempted to develop predictive models to identify users at risk of churning and to implement targeted interventions to retain them. Interventions included personalized in-game offers, tailored content updates, and direct engagement through notifications and emails.

Challenges and Failures in Data Analytics

Despite their efforts, Zynga encountered several issues that hindered the success of their big data strategy:

• Over-Saturation and User Fatigue: Zynga’s games often relied on similar mechanics, leading to user fatigue. The predictive models and interventions couldn’t overcome the fundamental issue of players losing interest in the repetitive nature of the games.
• Privacy Concerns: The extensive data collection raised privacy concerns among users. Some players were uncomfortable with the level of data Zynga collected, leading to negative perceptions and trust issues.
• Technical Limitations: While Zynga had access to large amounts of data, effectively analyzing and acting on this data in real-time proved to be technically challenging. The predictive models were not always accurate, leading to ineffective retention strategies.

The social gaming market was rapidly evolving, with new competitors and changing user preferences. Zynga’s big data approach couldn’t adapt quickly enough to these changes, leading to a mismatch between the data-driven strategies and the market realities.  Even when Zynga’s models correctly identified at-risk users, the interventions were not always well-executed. Personalized offers and notifications often came across as intrusive or irrelevant, failing to re-engage users.

Ultimately, Zynga’s heavy reliance on big data and predictive analytics had little effect on churn.  The company experienced a decline in active users and revenue, leading to layoffs and a shift in strategy. Zynga had to re-evaluate its approach, eventually focusing more on mobile gaming and diversifying its game portfolio. The Zynga case highlights that while big data has the potential to provide valuable insights and drive strategic decisions, its effectiveness depends on various factors including data quality, execution, market conditions, and user perception.

The failure to reduce churn despite substantial investments in big data analytics underscores the complexities and challenges of relying solely on data-driven approaches in a dynamic and competitive market.

Driving the Future of Big Data – Some Big Lessons Learned

To prevent similar mistakes in the future, a company like Netflix and Zynga can adopt several strategies to better leverage big data and improve the overall user experience. Here are some key steps:

Diversify Content

Regularly introduce new diverse content to keep users engaged and prevent fatigue.  Continuously inserting a wider range of content into A/B testing or product beta testing and feedback loops will help to create more compelling and engaging experiences.  Users who have a chance to see content choices they may dislike puts a more positive light on the content they do like.  Developers can also be poor predicters.  Occasional surprise preferences on the part of customers will help to further illuminate opportunities for new creative content.

Engender Trust

For many consumer watchdogs, the extensive data collection required for personalized recommendations raises significant privacy issues. Users might be uncomfortable with the amount of personal information the platform collects and analyzes, including viewing habits, search queries, and even device usage. The depth of data collection can lead to concerns about how this data is stored, used, and potentially shared.

Transparent Data Practices are enabled by clearly communicating how user data will be used. Obtain explicit consent from users and implement robust data security measures to build trust.  Ensure that AI and machine learning models are designed and deployed ethically, avoiding manipulative practices that might harm user trust.  Ensure that the data-driven strategies are fair and inclusive, catering to a diverse user base without reinforcing biases.

Evade Intrusions

Tailor interventions and offers based on the context of the user’s behavior.  Use data to find a balance between engaging users and respecting their autonomy. Avoid overloading users with notifications and offers.

Adapt with Agility

Stay informed about industry trends and competitor strategies to quickly adapt to changes in the market.  Adopt an agile approach to game development and marketing, allowing for continuous improvement based on real-time data and feedback.

Collaborate Across Functions

Interdisciplinary teams enable collaboration, which is key for understanding the context of organizational objectives in light of their impact on strategic objectives.  Sharing insights between data scientists, writers, game designers, marketers, and customer support teams is critical to creating well-rounded strategies and operational implementations.  Ensure that data insights are seamlessly integrated into the decision-making processes across all departments.

Put Users First

Map out the entire user experience, from selection to customer support. Ensure that every touchpoint is optimized to enhance satisfaction and loyalty.  Invest in building a strong community around the media content, encouraging social interactions and user-generated content to foster a sense of belonging and engagement.

Monitor

Monitor user behavior and engagement in real-time, allowing for swift identification of churn risks and opportunities for intervention.  Regularly solicit feedback from users to understand their preferences and to hear their views directly, not through the opaque veil of an algorithm.

Conclusion

With these approaches in place and firmly implemented across the organization, subscription media services can better leverage big data to enhance user retention while lessening inherent risks of bias creep and excess customer interference. A focus on user-centric design, privacy, and ethical use of data will be crucial in building sustainable engagement and long-term success.