The conference is focused on data science, analytics, and machine learning, specifically as they relate to media and entertainment. The topics discussed include a variety of techniques, such as causal inference, social data analysis, and synthetic control models. The talks explore the challenges and opportunities associated with using data to drive decision-making in media and entertainment companies.

​

While the talks are clearly targeted towards data scientists, there is a focus on making the material accessible to non-technical stakeholders within media and entertainment organizations. Some of the talks discuss the challenges of communicating technical concepts to non-experts and advocate for greater collaboration between data scientists and other teams within the organization. Overall, the conference appears to be aimed at individuals with a background in data science or analytics, but also seeks to provide insight and guidance for non-experts who are involved in data-driven decision-making in the media and entertainment industry.

​

Entertainment Analytics 2023

 

Talks included:

• “The Editor vs. the Algorithm" - University

• “Estimating Incremental Acquisition of Content Launches in a Subscription Service" - Streaming Service

• “Forecasting Long-Term Streaming of Emerging Artists" - Music Industry

• “Combining AI and Storytelling Expertise to Systematically Describe Content" - Tech Startup

• “Measuring the Value of a Title in a Subscription Bundle" - University

• “Entertainment Economics" - E-commerce Company

• “Large Language Models Will Revolutionize Entertainment Analytics" - Analytics Company / Panel

• “The Effect of Linear Television Airing on Digital Channels" - Entertainment Industry

• “Enhancing Film and Television Advertisements Through Genomic Targeting" - University

• “A Structural Equation Modeling Approach to Optimizing Release Windows" - University

• “How Playlists Affect Off-Platform Behavior" - Music Streaming Service

• “An Insights-Driven Narrative to Support Talent Initiatives in Hollywood" - Talent Agency

​

What we learned: 

• ⚡️ The Editor and the Algorithm 🧠: A speaker unpacked insights on human editors and algorithms in online news. Although algorithms excel at certain tasks, human expertise remains indispensable in complex situations.⚡️ The Editor and

• 📈 The Modelling Method 📊: Interesting use of modelling methods and data cleaning techniques to forecast long-term streaming success for emerging artists. Adaptive filtering and feature selection helped produce encouraging early results.

• 🤝AI and Storytelling 🤖: A promising approach merging AI and storytelling to better understand consumer tastes. Clearly defined traits and a human-coded dataset helped train models to predict complex movie traits with high accuracy. Identifying trait combinations provided more valuable insights than simply examining genres.

• 🧪The Data Puzzle 🧬: Thoughtful discussion on the challenges of quantifying the impact of new content launches, given the many variables at play. Models were proposed to adjust for baselines and noise in the data.

• 🔬The Big Picture 🎬: A cool discussion on broader strategic questions for digital news platforms when adopting algorithms, covering goals around clicks versus retention and tweaking algorithms accordingly.

• 📈Estimating Incremental Acquisition 📈: Interesting framework presented for estimating the incremental customer acquisitions resulting from a content launch within a subscription service. Usage behaviour and consumption patterns were used.

• 💲Innovative Allocation💲: An innovative methodology was discussed for allocating subscription revenue among the benefits of a bundled service, utilising discrete choice modelling and counterfactual simulations.

• 🔭Entertainment Economics 🔭: One company shared how data science and economics sit at the heart of decision-making in parts of their organisation. While some business units integrate data well, others still see a 'battle' between gut instinct and data-driven approaches.

• 🤑The Value of a Title 🤑: An interesting framework was outlined for measuring the value of individual content within a subscription bundle, debating the utility of using time spent as a proxy and examining substitution effects between content offerings.

• ⚡ Streaming on Linear TV 📺: Airings of movies on linear TV lead to a small but positive lift in streaming service sign-ups. Yet co-licensing films to similar streaming services causes cannibalization of viewer engagement on the platform. Managing the linear-streaming synergy requires careful consideration.

• 🎬Genomic Targeting in Advertising 🎬: A study found tailoring movie loglines to highlight viewers' genre preferences based on their film tastes significantly increased their likelihood of watching the film. The results suggest personalised advertising based on genomic data could enhance effectiveness.

• 💿Optimizing Release Windows 💿: A structural equation modelling approach explored the impact of streaming availability on movie revenues, uncovering insights into revenue changes caused by different release strategies. A decision support tool was also introduced.

• 🎵 Playlists and Off-Platform Effects🎵: An econometric analysis found that playlisting emerging artists leads to a small but noteworthy increase in their on-platform activities and off-platform live concert bookings. The persistent effects suggest playlisting is a potent tool for amplifying emerging artists.

• 🔎 Entrepreneurial Analytics 🔎: Data-driven storytelling was shown to influence major Hollywood decisions by transforming insights into compelling presentations that help convince clients, even initially sceptical ones. The ability to anticipate stakeholder perspectives and tie insights to revenue generation was emphasised.

• 👥 The Economics of Content 👥: Ad-supported streaming platforms was discussed, with considerations around tailoring content to match advertisers' audience preferences, necessitating a shift in content creation strategies.

• 🖥 Insights-Driven Decision-Making🖥: There was lots of talk about how to apply data and data-driven storytelling to influence complex decisions. From restructuring a sport’s competitive system to podcast forecasting and from talent prioritisation to greenlighting decisions. We learned lessons about how and where data played a pivotal role in driving narratives that informed business decisions.

​

​​

Entertainment Analytics 2022
 

Industries included:

• Movie studios and theatre chains

• Video and music streaming services

• Music

• eBooks

• Social media

 

Analytical methods included

• Thompson Sampling approach

• Markov Chain

• Structural Modeling

• Self-organizing maps

• Data-mined variables

• Random forest models

 

Business problems included:

• Optimizing movie ticket prices

• Measuring the value of content

• Estimating the lifetime value of a subscriber

• Optimizing release windows and simulating counterfactuals

• Understanding audience preferences and marketing films

• Analyzing the impact of Spotify playlists on the music industry

• Tailoring messaging and targeting to TV show fandoms

• Communicating data science findings to executives

• Extracting implicit knowledge from large AI models

• Addressing piracy in eBook pricing

• Reducing endogeneity bias in econometric models

• Determining audience affinity for brands on social media

• Navigating the transition from SVOD to AVOD for streaming services

​​

Talk-by-talk, here’s what we discussed: 

• Pricing Movie Tickets: Optimizing Revenue with Elasticity Testing: This presentation discussed how one chain of movie theatres used a Thompson Sampling approach to understand elasticities along the distribution and determine how to optimize movie ticket prices for each theater. The team ran field tests to ensure consumer acceptability, and found significant revenue improvements and fill rates at the same time, ultimately rolling out the approach more broadly with tweaks for inflation.

• Measuring the Value of Content on Screens and Streams: Using Parrot Analytics to Determine the Worth of Shows: This presentation discussed how Parrot Analytics uses data on consumption and interest to estimate the value of content in different contexts, such as estimating the lifetime value of a title to a streaming service, taking into account global rollout. Parrot also uses affinities between shows to calculate the value of a show for each streamer, and tracks talent demand to estimate the value that a lead actor brings to a show's value.

• Measuring the Value of Acquisition and Retention for Subscription Services. Using a Markov Chain to Estimate the Lifetime Value of Member: This presentation discussed how a video streaming service uses a Markov chain to estimate the lifetime value of a member, taking into account retention and acquisition probabilities, as well as the impact of pricing changes on the transition probabilities in the chain. While the model is complex and requires assumptions about a steady-state subscriber base, it allows them to properly measure the impact of pricing changes and optimize policies across the customer journey to maximize revenue.

• Optimizing Release Windows and Simulating Counterfactuals. Using Structural Modeling to Understand Consumer Behavior: This presentation discussed how structural modeling can be used to optimize release windows and simulate counterfactuals for the entertainment industry. By using a consumer model to understand how consumers make choices between different formats, and extending the model to include piracy as a format option, the model can examine alternate release scenarios of interest to the business and assess the impact of piracy on home release. However, the model's complexity can make it challenging to explain to decision makers, and the quality of data sources can vary.

• Understanding audiences and content using genomic metadata: The presentation discussed how self-organizing maps and genomic metadata can be used to analyze the content of films and understand the preferences of audiences. By quantifying the characteristics of films and identifying the genomic traits that resonate with audiences, this approach can provide useful insights for marketing films to specific demographics and forecast audience resonance for new films.

• The Power of Spotify Playlists: The presentation discussed the impact of Spotify playlists on the music industry and the declining market share of major record labels. The speaker shared research findings indicating that playlists account for 50% of streams and have caused a decline in the share of payments going to major labels. The presentation also examined whether Spotify has power over which songs succeed and the impact of playlist inclusion decisions.

• Marketing to Fandoms: Analyzing social media data using audience segmentation can help tailor messaging, targeting, and ad spend for TV shows. By identifying fandoms and understanding audience structure, marketing efforts can drive tuning and boost ROI by driving conversations among others.

• Communicating Data Science to Executives: A panel discussion on challenges and approaches: Data scientists still face challenges in communicating their findings and recommendations to executives, who may be more interested in other areas such as marketing or portfolio optimization. While progress has been made, more work needs to be done to integrate analytics into decision-making processes and cultivate a technical and business level of communication. Data scientists should focus on advocacy, education, and collaboration with stakeholders.

• Unlocking the Hidden Value of Large AI Models: This talk discussed ways to extract implicit knowledge from large, task-specific AI models to help people in an organization make better decisions. By using these models in more creative ways, businesses can gain insights that can be used to optimize marketing and content decisions.

• Pricing Against Piracy: Lowering eBook prices may reduce indirect piracy but not direct piracy, according to a study that used a synthetic control model. While it is difficult to determine the magnitude of the effect, the results suggest that lowering prices may be effective in reducing piracy.

• Improving Causal Inference using Data-Mined Variables: This talk discussed the issue of endogeneity bias in models that use predicted quality data, and proposed using random forest models to generate potential instrumental variables that could help reduce this bias in econometric models.

• Social Data for Unlocking Net Audiences and Marketing Techniques: The presentation discussed how to use engagement on social media as a better metric than followers to determine audience affinity for brands and to evaluate the incremental reach of actors for casting purposes. The speaker also emphasized that understanding the landscape of brands with the highest affinity is separate from how to make the best recommendation for business objectives.

• The challenges of transitioning from ad-free to ad-supported video streaming: In this talk, Roku shared insights on the analytics and challenges of moving from subscription video on demand (SVOD) to advertising video on demand (AVOD). They found that the average revenue per user (ARPU) was higher for ad-supported content, but survival rates were lower, so mixed tiers with reduced fees and some ads are worth testing. Adding linear channels was also found to be an important growth driver.