Past Events

Twenty-two data and analytics leaders from British and European media organisations spent two days at MediaCityUK in Salford. Twelve talks ran to a strict format: fifteen minutes of presentation, then fifteen of discussion, under Chatham House Rules. Two speakers dropped out late and both Friday gaps were filled from the floor. Where the previous edition asked whether language models would change this work, this one took that as settled and argued instead about what comes after: who owns generated software, where corporate memory goes when everyone can produce analysis, and how anyone proves a probabilistic system still works.

Talks

• “Building a Demand-Tracking Product from Public Web Data” – An analytics consultancy reconstructed the demand curve for a seasonal live-entertainment operator from seven public web sources, built in roughly four weeks of spare hours and sold only once it worked. Almost nobody opens the dashboard that came with it: what gets used is a recurring email and a messaging thread.
• “Adventures with Conversational Analytics” – A news publisher built a routing agent over four data sources and found that ordinary data governance did most of the work, while the fashionable context layer was worth 10 to 20 per cent at the margin. Adoption reached 70 per cent of users because the same agent sits on four surfaces rather than one.
• “Democratising Software Development Across a Media Company” – A public service broadcaster opened a platform where any employee publishes a working tool by dragging a folder into a browser, and 17 apps landed in the first weeks with no promotion. Regional journalists built audio editors better than anything they could buy, and nobody could agree who maintains them.
• “Measuring Incremental Reach from Off-Platform Distribution” – A public broadcaster used copulas and tetrachoric correlations to turn daily, global social figures into weekly domestic reach. A festival week that an independence assumption put at 32 per cent came out at 13 per cent, and one correlation matrix taken from the top of the funnel still held two levels further down.
• “Designed to Advise: Building a Centralised Insight Function” – A former insight director at a commercial broadcaster described merging six scattered teams into business partners sitting over three specialist pools, after which the function became the only team besides the chief executive that thought about the business as a whole. The regret was building on data infrastructure that was not ready, having flagged it and pressed on anyway.
• “Informing Content Exploitation Decisions with Scientific Models” – A studio and streaming group spent four years modelling every licensing and scheduling decision, carrying more than 50,000 interpretable parameters for a single metric. Against a consultancy prototype it doubled accuracy on less data and cut running costs sharply, by connecting to data where it already sat rather than copying everything into one lake.
• “Making Ad Effectiveness Work” – Three advertising measurement veterans argued that the industry's favoured triangulation of models, experiments and attribution fails in practice, and that what separates teams who get it right is organisational rather than technical. One client's own credible evidence supported anything between spending nothing on television and spending seven figures a year.
• “Go Big, or Stay Home: The Economics of Stadium Concerts” – An economics consultancy built the dataset that did not exist: headlining artists up 16.5 per cent in a year, concerts up 12.4 per cent, and the number of stadiums flat. Italian artists play 93 per cent of their stadium shows at home, Korean artists 92 per cent of theirs abroad.
• “Our New Viewing Segmentation” – A public broadcaster cut a year of single-source panel data roughly 25 ways and kept getting the same four groups, defined by what people watch on rather than who they are or when they watch. Day of the week and time of day made almost no difference, and the finding the business used was that moving a show from Wednesday to Friday changes nothing.
• “Analysing a Century of Parliamentary Language on Immigration” – A newspaper's data science team classified the stance of five million parliamentary speeches across a hundred years, using a two-stage pipeline that judged relevance first and sentiment second. General sentiment models cannot do this, because they score the vocabulary rather than the position, and a passage quoting hostility in order to attack it reads as hostile.
• “Evaluations: Getting AI into Production” – A consumer data group shipping customer-facing generative features described the harness around the model, and the two slow qualitative stages that consume the effort: defining what success means, then gathering the evidence. Nobody in the room had a source for a golden question set.
• “Drivers of Frequency: Causal Inference Modelling” – A subscription publisher used a three-month propensity-weighted design to find which content format drives frequency, because withholding the thing people paid for is not a testable option. Every format came out the same size, because subscribers use several at once, so the ranked list the business wanted does not exist.

What we learned

The argument has moved on from whether these tools work to who is accountable for what they produce, and how anyone would notice when they quietly stop working.

A separate US edition, organised and funded by the two academic partners rather than by the full organising group. Between thirty and forty people met for a single day at the Marina del Rey Marriott. The format was group discussion rather than presentations, with the focus on film and television. It was the first US edition since 2023.

Talks

Thirty data and analytics leaders from major UK and European media organisations gathered for two days at The Lowry in Salford. This was the first Entertainment Analytics conference to comprehensively address the impact of language models on media businesses. Sessions ran 30 minutes each, split equally between presentation and Q&A under Chatham House Rules.

Talks

• “Estimating Cross-Media Reach Using Copulas” – A public broadcaster showed how copulas (a multivariate technique from finance) de-duplicate audience reach across TV, radio and online for 42 local areas, feeding an automated weekly pipeline that local newsrooms use in near real-time.
• “Optimising YouTube Distribution Through Genre-Specific Channels” – A public service broadcaster found 72% of YouTube views came from algorithmic recommendations but only one or two shows per channel ever broke through. Restructuring into niche genre channels using demographic clustering drove 77% year-on-year growth.
• “Outcome Measurement for Television Advertising” – A commercial broadcaster modelled 640+ brands to show TV ads produce a large short-term spike in website visits plus a tiny long-term effect lasting six months that roughly equals the short-term impact in total value.
• “Structuring Unstructured Recipe Data with Classical ML” – A major newspaper’s data team extracted structured data from 30,000 free-text recipes using fine-tuned named entity recognition, achieving 96–99% accuracy while deliberately avoiding generative AI so they could diagnose every error precisely.
• “Multi-Agent Systems for Audience Insights” – An assessment of where multi-agent AI systems genuinely work in the audience insights pipeline. The standout finding: persona simulation generated approximately $12 million in net profit across 58 projects in four months, where companies simulate their existing segmentation answering hundreds of additional questions.
• “Z-Score Audience Analysis in Sports Media” – A European media group with 230 million monthly devices used Z-scores to analyse reader behaviour across content categories. The simplicity of the metric was its greatest strength: editorial teams immediately understood and acted on it. A parallel “zero Google” analysis showed 80%+ of sports traffic arrives directly.
• “The AI Transformation Chain” – Interviews with 22 insight leaders revealed 19 are failing at AI transformation. Success requires a six-link chain: executive sponsorship, app approvals, foundational skills, standardised tasks, process orchestration, and sustained reinforcement. One bright spot: a consumer goods firm delivering 36% more insights with 20% fewer people.
• “A Completion-Based Model for Fairer Streaming Royalties” – A former chief economist proposed adding a completion threshold to music royalties. Analysis of 15,000 songs showed completion rates are consistent across song lengths (85–95%), so the model does not punish longer songs. It also eliminates click-farm fraud engineered to last just 31 seconds.
• “Customer Segmentation Across Linear and Streaming” – A pay-TV operator simplified 16-segment clustering into four primary consumption cohorts to enable cross-platform comparison for the first time. The key finding: ~25% of the base relies heavily on linear partner channels, making channel-dropping decisions riskier than assumed.
• “AI-Generated Long-Form Sports Content at Scale” – A sports publisher produces 15,000 AI-generated match previews per year across 49 football leagues. The architectural insight: turn structured data into text strings first using traditional ML, then use the language model purely for fluency. AI content showed ~20% lower engagement on major leagues but near-parity on obscure leagues where no human alternative existed.
• “AI-Powered Product Discovery in Luxury Retail” – A luxury retailer demonstrated text-to-image search, image-to-product matching, and an LLM outfit stylist, all built by a single data scientist. The image search required bespoke training to bridge the domain gap between clean e-commerce shots and distorted real-world photography.
• “Visual Exploration of Content Catalogues Using AI Embeddings” – An interactive tool maps 20,000 streaming titles into navigable 2D space using language model embeddings. Filtering by date reveals the emergence of genre clusters over time. The presenter argued chat interfaces alone are insufficient for exploration when you do not yet know the question.
• “Estimating Impact of LLMs on Traffic” – A major international news publication shared data on millions of weekly visits lost to AI platforms, with scrape-to-referral ratios reaching 60,000:1 for some AI services.
• “AI in Academic Publishing” – A global academic publisher explored how language models are reshaping scholarly research workflows and discovery.
• “Data Democratisation in News Media” – A European media company described building a self-service analytics culture across newsrooms.
• “Using LLMs to Analyse Survey Responses at Scale” – A UK broadcaster applied language models to process a massive volume of open-text survey responses, dramatically reducing the time from collection to insight.
• “Candid Truths about AI in Insights” – A practitioner guide to the specific risks of using language models for data analytics, where compounding errors through iterative analysis can be far more costly than a poorly written email.

What we learned

Language models and AI have moved from experimental to existential for media organisations. Technical capabilities exist but human adoption, political barriers, and fundamental business model questions remain largely unresolved.

The UK conference returned after a five-year gap. Two days of 50-minute talks followed by 50-minute discussions, with attendees from major broadcasters, publishers, music companies, data consultancies, and regulators. A suggestion was made to create a penalty jar for every mention of “Gen AI” during the two days.

Talks

• “Building a Data-Driven Culture at a 180-Year-Old Publisher” – A media analytics leader joined a major global publication and found it roughly five years behind a comparable title in data maturity, despite nearly identical tech stacks. The gap was cultural, not technical: the fix was getting data people into leadership meetings, not buying new tools.
• “Principles for Responsible AI Deployment” – A major financial newspaper warned that organisations experimenting with generative AI were creating a “Pandora’s POC box”: small experiments proliferating with no central visibility, each carrying infrastructure costs, and in one case duplicating work the internal team had already completed.
• “AI Applications in News Publishing” – A data science leader returning from parental leave found a third-party consultancy had been contracted to rebuild a recommendation model that duplicated her team’s own work, using the organisation’s own content taxonomy. The contract was signed before the data science team knew about it.
• “Cross-Platform Audience Analytics and Attribution” – All tests optimised for a single metric: total customer lifetime value of incoming subscribers. This removed endless debates about conversion rate vs. product mix vs. average order value, making testing fast and decisive.
• “Data Governance and Regulatory Perspectives” – Discussion of how media data regulation is evolving, with perspectives from regulators and practitioners on the challenges of balancing innovation against privacy obligations.
• “Advertising Effectiveness and Outcome-Based Measurement” – Calculating the return on investment for a data science initiative frequently took longer than building and deploying the model itself. One speaker described spending extensive effort just to justify £50,000 of infrastructure spend.
• “Music Industry Data: Streaming Economics” – An examination of how streaming royalties are calculated, distributed, and disputed, with fresh data on the economic impact of playlist curation on independent artists.
• “Regional Broadcaster Perspectives on Data Maturity” – Even a deliberately simplified approach to AI required about an hour a day of experimentation to avoid falling behind. Attendees from major broadcasters described mandatory eight-hour AI training courses as a prerequisite just to access basic tools.
• “Measuring UX Impact Without A/B Testing” – A TV platform provider that doesn’t own the customer relationship developed causal inference alternatives to randomised testing. The key learning: getting product owners to write down hypotheses and decision trees upfront mattered far more than the measurement methodology itself.
• “The Data and AI Playbook from Private Equity” – Research across 35+ PE-backed data leaders found compensation has risen from £250,000 to £400,000 (now including carried interest) even as data teams elsewhere face cuts. Generative AI had produced zero change in enterprise valuations so far because it has only delivered productivity gains, not new revenue.
• “Glocalization: Local Artists Dominating Global Platforms” – On-demand streaming is driving local artists to the top of local charts. In Germany, zero of the top 100 radio songs were German-language, but a third of the top 100 streaming songs were. In Denmark, nine of the top ten streaming tracks were Danish. The UK has not produced a true global pop star since 2017.
• “Simplifying the NLP Stack with Generative AI” – A major publisher found that prompt-engineered named entity extraction outperformed established NLP tools (spaCy, NLTK) in blind editorial evaluations, and reduced training data requirements from 100 articles per topic to 10 using a human-in-the-loop feedback system.
• “Neuroscience-Based Creative Optimisation” – Predictive models trained on 120,000+ neuroscience test subjects (eye tracking, EEG, pupil dilation) evaluate marketing creative effectiveness at millisecond-level subconscious responses, avoiding the self-reporting bias of traditional survey methods.

What we learned

Twelve talks covering content valuation, streaming forecasting, genomic targeting, release window optimisation, playlist economics, and the emergence of language models in entertainment analytics. A live demonstration showed a non-programmer running a full clustering analysis of music industry data through natural language conversation.

Talks

• “The Editor vs. the Algorithm” – A field experiment at a major news outlet found the algorithm outperformed human editors, but a combination of both would have delivered 13% more clicks than either alone. Pure algorithmic personalisation also reduced consumption diversity over five months, creating a feedback loop that narrows the content a platform can justify producing.
• “Estimating Incremental Acquisition of Content Launches” – An econometric model valued individual titles within a streaming bundle by measuring their effect on subscriber churn. Tentpole releases increased total platform consumption by roughly 6%, and subscribers whose consumption dropped below their personal average were disproportionately likely to cancel.
• “Forecasting Long-Term Streaming of Emerging Artists” – A record label’s data science team trained their model on 7,000 songs with eight years of data. One case study: a song released in 2019 doing 16,000 streams per week jumped to five million per week by May 2022. TikTok virality has decoupled a song’s discovery moment from its release date.
• “Combining AI and Storytelling Expertise to Describe Content” – A human-coded dataset of clearly defined traits was used to train models that predict complex movie traits with high accuracy. Identifying trait combinations proved more valuable than examining genres alone.
• “Measuring the Value of a Title in a Subscription Bundle” – A structural demand model allocated tens of billions of dollars of subscription revenue across bundled services. The Shapley value method was abandoned because removing a single benefit entirely produced such extreme counterfactuals that results were not credible.
• “Large Language Models Will Revolutionise Entertainment Analytics” – A live demo showed a non-coder using natural language to run a full clustering analysis across dozens of countries, complete with named segments, marketing recommendations, and visualisations. Of the room, only about six attendees had corporate-endorsed language model access, while at least five admitted to using unapproved tools.
• “The Effect of Linear Television Airing on Digital Channels” – Airings of movies on linear TV produced a small but positive lift in streaming sign-ups. Yet co-licensing films to similar streaming services caused cannibalization of viewer engagement.
• “Enhancing Advertisements Through Genomic Targeting” – Content genome data was applied to film and television advertising targeting, matching ad creative attributes to audience taste profiles at scale.
• “Optimising Release Windows Using Structural Equation Modelling” – A structural model examined how consumers choose between theatrical, home video, and streaming formats, extended to include piracy, allowing simulation of alternative release scenarios.
• “How Playlists Affect Off-Platform Behaviour” – Playlisting emerging artists led to a noteworthy increase in both on-platform activity and off-platform live concert bookings. The persistent effects suggest playlisting is a potent tool for amplifying emerging careers.
• “An Insights-Driven Narrative to Support Talent Initiatives” – Data-driven approaches to talent representation decisions, quantifying the commercial and cultural impact of diversity initiatives in the entertainment industry.
• “Entertainment Economics” – An examination of economic models in entertainment, covering pricing, bundling, and consumer behaviour across digital and physical formats.

What we learned

The conference returned after a two-year COVID hiatus with around 70 attendees, up from 31 in 2016. Topics included Thompson Sampling for ticket pricing, Markov Chain subscriber lifetime value, structural modelling for release windows, playlist power, the SVOD-to-AVOD transition, and the first live demonstration of language models for entertainment analytics.

Talks

• “Pricing Movie Tickets with Bayesian Bandits” – A cinema chain abandoned traditional regression models and adopted Thompson Sampling instead, treating each theatre like a multi-armed bandit running 10,000 simulated iterations per pricing decision. Movie demand shifts so drastically week to week that traditional price elasticity requires stable demand cinema never has.
• “Measuring the Value of Content on Screens and Streams” – A content analytics company scored over 2,000 characteristics per film using human analysts. Self-organising maps projected this data onto a 2D plane, creating a topographic map of cinema. One superhero film independently clustered apart from its franchise peers because it was actually a spy thriller in disguise.
• “Subscription Lifetime Value Using Markov Chains” – A streaming platform showed that the standard industry method for valuing subscriber acquisition overstates the true value by approximately five times. The error: it ignores the baseline probability that people would have subscribed anyway.
• “Optimising Release Windows with Counterfactual Simulation” – Structural modelling examined how consumers choose between formats, extended to include piracy, allowing simulation of alternate release scenarios of interest to the business.
• “Understanding Audiences Through Content Genome Data” – Overlaying audience taste profiles on a content genome map revealed that marketing a film’s core genre might alienate the target demographic, while a secondary content thread could be a stronger hook.
• “The Power of Streaming Playlists” – Research on the 4,000 most-followed new music playlists found that playlist inclusion causally determines roughly 50% of a song’s total streams. Major-label share on platform-curated playlists fell by ten percentage points, largely explaining their declining streaming revenue share.
• “Marketing to Fandoms” – Data-driven approaches to understanding and activating fan communities across entertainment properties.
• “Communicating Data Science to Executives” – The gap between analytical sophistication and business impact was framed as a four-link chain: right data, right method, right recommendation, right adoption. Almost every conference presentation addressed only the second link.
• “Unlocking the Hidden Value of Large AI Models” – A live demonstration where a non-coder used natural language to run a clustering analysis of music industry data. Only about six of 70 attendees had corporate-endorsed language model access. Five admitted to using non-endorsed tools that could get them in trouble.
• “Pricing Against Piracy” – Economic modelling of how pricing strategies interact with piracy rates across different release windows and geographic markets.
• “Improving Causal Inference Using Data-Mined Variables” – Novel approaches to strengthening causal claims in observational entertainment data by identifying and controlling for algorithmically discovered confounders.
• “Social Data for Unlocking Net Audiences” – Using social graph analysis to identify and reach net-new audiences beyond existing marketing databases.
• “Transitioning from Ad-Free to Ad-Supported Streaming” – Average revenue per user was higher for ad-supported content, but survival rates were lower. Adding linear channels proved to be an important growth driver.

What we learned

The fourth US conference continued to build the bridge between entertainment industry practitioners and academic researchers. Two days of presentations and discussion covered content valuation, audience segmentation, causal inference, and platform economics.

Talks

• Content valuation methodologies for subscription services
• Machine learning approaches to audience segmentation
• Causal inference methods in entertainment industry experiments
• Platform economics and marketplace dynamics
• Social media analytics for content marketing
• Streaming behaviour and subscriber retention

The third US edition, over two days at the Marina del Rey Marriott. Nine sponsors backed it, spanning streaming services, a film studio, an online education company, content analytics firms and two universities. Thirteen sessions ran, tilted further towards academic economics than the UK editions: digitisation and the incentives to create, theatrical window length, social influence, microscheduling and forecasting. A panel closed the first day by asking whether machines or people have better taste, with a broadcaster, a university, a retail platform, two streaming services and a studio answering.

Talks

• “Digitization, Revenue Shock and Incentives to Create: Evidence from Bollywood” – From a university researcher.
• “Maximizing Users for Live Streaming Gameplay” – From an analytics firm.
• “Quantifying the Social Influence of Actors and Actresses” – From a social analytics firm.
• “The Causal Effect of Reducing Theatrical Window Length on Box Office Ticket Sales” – From a university researcher.
• “Predicting Film Box Office Using Search Data” – From a large technology platform.
• “Trusting Your Forecasts: Marketing Applications of Structural Time Series Models” – From a marketing analytics firm.
• “A Risk-Retention Model for Subscription Video Services” – From a streaming service.
• “Revenues, Incentives, and the Creation of Artistic Works” – From a university researcher.
• “Panel: Who Has Better Taste, Machines or People?” – A broadcaster, a university, a retail platform, two streaming services and a film studio, closing the first day.
• “Content Genome Applications: Launching a Global Streaming Platform” – Jointly from a content demand analytics firm and a video platform.
• “Application of Causal Econometrics to Product Marketing and Optimization” – From a streaming service.
• “Strategic Microscheduling of Movies” – From a university researcher.
• “Brand Tracker Research on TV Show Fandom” – From a research firm.

What we learned

The second UK edition. Ninety-one people were invited back to the same room at The Lowry, and thirty-six accepted. Seventeen sessions ran across two days, keeping the pattern of a short presentation followed by an equal slot for discussion. The programme was broader than the first UK edition: public broadcasters from two countries, a television platform operator, a news publisher, a media agency, sports analysis and university research.

Talks

• “Agent-Based Models: What Happened Next” – A return to the simulation work previewed as a proof of concept at the 2017 UK edition, a year on.
• “Behavioural Machine Learning for Media” – Two academics combined decision theory with machine learning, running sentiment analysis over film subtitles at scale and linking the result to box office, then embedding choice models inside a collaborative-filtering recommender.
• “Attribution” – From a media agency.
• “Analysis of TV Viewing” – From a Nordic public broadcaster.
• “Viewing on the Platform” – From a television platform operator.
• “Discourse Analysis for Content Discovery and Automated Metadata”
• “Social Listening”
• “Analysis in Sport”
• “Challenges in 2018” – A closing group discussion on the second morning.

What we learned

The second US conference grew to 49 attendees, with sponsorship from universities, streaming services, and premium cable networks. The two-day event maintained the intimate, discussion-heavy format established in the first year.

Talks

• Content demand measurement and prediction at global scale
• Audience analytics for streaming platforms
• Digital pricing experiments in publishing
• Marketing effectiveness measurement using search data
• Social network analysis for audience understanding
• Release strategy optimisation for film distribution

The first UK edition, and the start of the European series. Sixty-eight people were invited to The Lowry in Salford for a full day of presentations, a dinner, and a half day of deeper discussion. Talks ran fifteen minutes, with an equal slot for argument afterwards. The room mixed public and commercial broadcasters, media agencies, record labels, universities and analytics vendors. Two themes ran through both days: how to break content down into machine-readable attributes, and how far a company's own platform data can be trusted to describe its market.

Talks

• “Introductions and Opening Remarks: The State of Data Science and Analytics in Entertainment” – The organisers, from a public broadcaster and its commercial arm, opened the two days.
• “Content Genome 1: TV Show Genomes to Predict Country-by-Country Success” – A business-school researcher and a script-analytics firm broke 500 seasons of a broadcaster's catalogue into roughly 70 attributes, then modelled licensing revenue across 20 countries. The aim was not to pick hits but to say which themes over-index where, which is a licensing negotiation rather than a commissioning decision.
• “Content Genome 2: Hacking Together Film Data and Lessons Learned Therefrom” – A media agency stitched three years of weekly ticket sales across 11 countries to film identifiers, search volumes and trailer views. The lesson was about access rather than method: one supplier released a partially blinded sample, one title in ten, so the modelling could be checked without handing over the catalogue.
• “Content Genome 3: Metadata Extraction from TV Shows and Lessons Learned” – A broadcaster's research department ran speech-to-text and entity extraction over 70,000 archive radio programmes that had almost no metadata, then let the public correct the output. Combining subtitles, face recognition and speaker identification narrowed a search far more than any one signal alone.
• “What Makes a Personalised Weekly Playlist So Good?”
• “Using Agent-Based Modelling to Predict the Future of UK TV Consumption Behaviours” – A simulation firm showed a proof of concept: 600 synthetic agents, each needing a blend of five content "nutrients", choosing between four channels. The framing came from nutritional geometry, where an animal forages for a combination of foods that no single food provides.
• “Clustering 1: The Power of Clustering Survey and First-Party Data Sets” – A social-graph analytics firm split the followers of one television brand into 20 clusters by who they follow rather than what they post. Only 10 to 15% of users on that platform post at all, so listening tools describe a small and unrepresentative minority.
• “Clustering 2: Behaviour-Based Genres and Genre-Based Segments” – A media agency derived 13 genres from the viewing behaviour of roughly ten million people, using cleaned-up media file names as the content signal. The clusters were handed over unnamed so that the people who had to use them could name them.
• “Clustering 3: The Science of Nudging Viewers Using TV Show Profiling” – Behavioural scientists predicted whether a show would land in the top tail of public ratings with about 70% accuracy using three attributes only: genre, country of origin and the number of countries it had been released in.
• “Box Office Forecasting and the Release Date Optimisation Problem”
• “Centaur Evaluations: Merging Mind and Machine for Better Recommendations” – A broadcaster tested recommendation models by having staff judge head-to-head comparisons, with model identities hidden and the order randomised to strip out framing effects. The proposed upper bound was not a better algorithm but a panel of human editors.
• “What Affects Your Pay? What Drives Salary Increases?”
• “One Second Reach: User-Level Data and Second-by-Second Viewing Data”
• “Survivor Bias” – An analyst used the wartime bomber armour problem to argue that first-party data describes only the people who came back. Even a heavy user gives any one platform a fraction of their total time with that medium, and the rest of it is invisible from the inside.
• “Real-Time Media Analysis”
• “Why Data Science Projects Fail”
• “Forward-Looking Media Trends” – A research firm argued that subscription video was adding to pay television rather than replacing it, with nearly three quarters of a billion of combined audience growth. Messaging platforms had 6.5 billion monthly active accounts, close to twice the number of connected people, and almost no measurement.
• “The Data Behind Influencers” – A creator-search platform had confirmed more than 20,000 brand collaborations and detected roughly 200,000 more. Against the investor view that influencer marketing was finished, the like-to-dislike ratio on branded videos had climbed for a decade to about 96%.

What we learned

A phrase kept recurring across the two days: the poor man's version. Nobody in the room had the data the platforms had, so every method on show was built to work without it.

The inaugural Entertainment Analytics Conference. Around 30 attendees from TV, SVOD, film studios, music, data and technology companies, and universities gathered for an intimate day of sharing and learning about entertainment analytics and data science.

Talks

• “A Moneyball Moment for Hollywood?” – The keynote framed the conference’s founding question: entertainment has historically been a gut-feel industry, but these firms now compete against technology companies that make quantitative decisions. In the fight between data and gut feel, it’s not even a fair fight.
• “Measuring and Predicting Global Demand for Content” – A method for quantifying audience demand for entertainment content across multiple markets using digital signals, enabling cross-country comparison of content appetite.
• “A Global Cross-Media Buzz Reporting System” – Using Google Trends data to build an automated system for tracking and comparing marketing impact across campaigns, territories and media types.
• “Testing Print and Digital Pricing Using Randomised Experiments” – Randomised controlled experiments applied to media pricing decisions, demonstrating how experimental design can replace intuition in subscription pricing.
• “Global Streaming Distribution and Frictionless Digital Trade” – An examination of how digital distribution is reshaping international content trade, reducing friction but creating new competitive dynamics.
• “Improving the Economics of Measuring Advertising Effectiveness” – Ghost ads (showing a control group what they would have seen without the ad campaign) offered a cost-effective alternative to traditional brand lift studies.
• “Using Content Genome to Predict TV Content Sales” – A system decomposing TV shows into 30+ attributes to predict international demand across 500+ seasons and 20 countries. Synopses predicted accurately; scripts confused the analysis with stage directions and dialogue filler.
• “The Insight Contained in the Social Graph” – Graph-based audience clustering using follower networks (not what people say) revealed distinct audience segments invisible to traditional analytics. Among 300,000 followers of major comic-book brands, superfans skewed 60% toward one publisher while film fans skewed 56% toward the rival.
• “The Impact of Free Streaming on Paid Download Sales” – A public broadcaster found free streaming reduced paid digital sales by 7–8%, but the advertising revenue earned from free streaming almost exactly cancelled out the lost sales revenue, creating a genuine strategic stalemate.
• “Early International Digital Movie Releases and Domestic Box Office” – Releasing films on Chinese streaming platforms triggered high-quality pirated copies within days, causing a roughly 13% decline in US theatrical revenue starting from week four of the box office run.
• “How Messaging Apps Are Changing Online Video” – For Generation Z, the primary way they encounter a song is a 15-second clip. Messaging platforms like WeChat and QQ had 1.6 billion monthly users but offered essentially zero analytics. Ignoring platforms because metrics don’t exist was described as dangerous.

What we learned