A studio is a lot like the CIA Situation Room in Zero Dark Thirty. No one can never be 100% certain of Box Office projections. You have Sony’s Michael Lynton at the head of the table asking what his movie will produce next weekend at the Box Office. Amy Pascal says that she thinks the film could do $20 million, with a 60% probability. Jeff Blake says that when he spends $X million in marketing, it should do $20 million…with 65% probability. Kathryn Bigalow and Mark Boal say it’ll do $25 million, with 100% probability…well 95% probability, because they know certainty freaks everyone out. Really, the process of projecting Box Office and the corresponding implications for a film’s Ultimates (the total revenue stream of a film in all of its distribution windows), as well as the impact for a film on a studio’s slate could benefit from the big data and predictive analytics revolution. Over the last decade, the many slate financing failures indicate a less-than-satisfactory projection process.
Box Office Predictions: Building a Comparable Universe
When predicting the Box Office and the Ultimates for Les Miserables, is the right universe of comparable films more like Chicago ($170.6M) or The Nine ($19.7M) – both musicals with December holiday season releases? Both were Rob Marshall films; is he a good comparable for Tom Hooper? Gosh, Tom Hooper’s last film, The King’s Speech, did $135 million and won the Oscar. How does that inform the analysis? Enter the world of data science. Predictive analytics helps the studio to improve its modeling and financing planning with better analysis. In a quick example, we just named three different variables, all of which could contribute to a prediction of what Les Miserables could achieve at the Box Office. And although this analysis can be done at the aggregate level, it is far better suited for a detailed model that includes a geospatial and time-series component. The exhibitors and distributors know exactly where tickets sold and where they did not for Chicago and The Nine. They know what days of the week and times were best. And, there is a flood of other possible information to cull in order to help a studio and the exhibitors work together to better project how a musical might perform. Let’s take an example…
When predicting the Box Office and the Ultimates for Les Miserables, is the right universe of comparable films more like Chicago ($170.6M) or The Nine ($19.7M) – both musicals with December holiday season releases? Both were Rob Marshall films; is he a good comparable for Tom Hooper? Gosh, Tom Hooper’s last film, The King’s Speech, did $135 million and won the Oscar. How does that inform the analysis? Enter the world of data science. Predictive analytics helps the studio to improve its modeling and financing planning with better analysis. In a quick example, we just named three different variables, all of which could contribute to a prediction of what Les Miserables could achieve at the Box Office. And although this analysis can be done at the aggregate level, it is far better suited for a detailed model that includes a geospatial and time-series component. The exhibitors and distributors know exactly where tickets sold and where they did not for Chicago and The Nine. They know what days of the week and times were best. And, there is a flood of other possible information to cull in order to help a studio and the exhibitors work together to better project how a musical might perform. Let’s take an example…
Who and What is Geospatial Modeling
At its core, Box Office predictions reflect a fundamental transaction. The studio and exhibitor know the price of the ticket – let’s say around $8 on average. All that needs to be predicted is how many tickets will be sold. And, the best way to develop that prediction is to estimate the number of theatres that will show the movie and the number of tickets every theatre will sell. It is a data-rich environment. Every exhibitor knows exactly how many tickets have been sold at any given location for any given movie. That reflects a lot of transactions, a lot of movies and a lot of data that can be applied to a future forecast. And, the more, the merrier! Cinemark owns 299 theatres and 3,918 screens in 39 states. They sold over 150 million tickets in the last twelve months. AMC own 332 theatres with 4,804 screens across the United States, and approximately 200 million movie-going guests each year. And Regal owns 6,880 screens in 540 theatres in 38 states and the District of Columbia, with over 200 million attendees. Big film releases can reach as wide as 4,400 screens. The equations get more interesting with the utilization of rewards data from these chains, data from technology partners (fandango, moviefone), and social listening data (e.g. twitter mentions).
At its core, Box Office predictions reflect a fundamental transaction. The studio and exhibitor know the price of the ticket – let’s say around $8 on average. All that needs to be predicted is how many tickets will be sold. And, the best way to develop that prediction is to estimate the number of theatres that will show the movie and the number of tickets every theatre will sell. It is a data-rich environment. Every exhibitor knows exactly how many tickets have been sold at any given location for any given movie. That reflects a lot of transactions, a lot of movies and a lot of data that can be applied to a future forecast. And, the more, the merrier! Cinemark owns 299 theatres and 3,918 screens in 39 states. They sold over 150 million tickets in the last twelve months. AMC own 332 theatres with 4,804 screens across the United States, and approximately 200 million movie-going guests each year. And Regal owns 6,880 screens in 540 theatres in 38 states and the District of Columbia, with over 200 million attendees. Big film releases can reach as wide as 4,400 screens. The equations get more interesting with the utilization of rewards data from these chains, data from technology partners (fandango, moviefone), and social listening data (e.g. twitter mentions).
And beyond ticket sales, some of that data may reflect information about the geography (amount of a population within drive time of x minutes from theatre – a larger concentration in more dense regions – but people’s willingness to travel longer distances in dense areas may be less, which also needs to be accounted for) or demographics (housewives in a viewing quadrant who may have helped drive “The Help” to a book club success and will likely support the movie). Some data will result from studio decisions. For example, P&A covered local news ads before the opening, driving awareness and Box Office. There could be external noise – like rainstorms that impacted attendance, or the big college team playing its rival on a Friday night. All of this should and can be incorporated into a forecasting model.
A geospatial analysis might be able to tell a studio to expand a movie like Silver Lining’s Playbook to 730 screens rather than 745. But, more importantly, it can help empower specific decisions like which screens yield the greatest likelihood of success, helping the studio to earn $7,000 on average per screen, rather than $5,476. The impact for a single film’s best weekend would be to increase the Box Office from $4.1 million to $5.1 million – or 25%. At its widest release, the film was in 2,641 theatres. Thus, if that 25% improvement in the opening weekend (wide) rippled through the run of Silver Lining’s Playbook, the result would be a total Box Office of $88 million vs. the $69 million currently estimated.
This introduces another fascinating trend in Box Office: how much does the Box Office of a film change from its opening weekend through the lifetime of the film? On Opening Night the studios (and the trades) are already using their traditional equations to predict what the Box Office opening weekend will look like. But data science is specifically equipped to better predict these complex relationships. For the top Oscar contenders alone, Silver Lining’s Playbook had a 73 day run, Zero Dark Thirty is at about 40 days. Lincoln is at an 80 day run. And Argo is at 108 days.
Predictive analytics, geospatial modeling and big data are all valuable tools in this projection process. From predicting opening night numbers when a film release date is set through to correlating that prediction to inform estimates for day 2, week 2 and the Box Office bottom line, to DVD or non-theatrical, data science empowers much smarter planning and execution. And, we haven’t even begun to analyze P&A or development or film slate financing: we will get to those in another post. Stay tuned!
