The motion picture industry is growing at a rapid growth rate, likely due to the acceleration of online and mobile distribution, lower admission prices, and government policy initiatives. This industry is also rich in data, thus making it extremely exciting for statisticians. The movie industry, which used to rely on traditional conventional wisdom and simple rules of thumb to predict box office outcomes, is slowly seeking new “analytical” approaches.
Stakeholders are looking for a ‘magic formula’ to better understand and predict box office success are turning to statisticians and data scientists to help with this challenge. (Interestingly enough, blockbuster movies are even including analytics as a part of the actual script e.g. Moneyball). To increase their profits, producers and directors need to understand what raises the curiosity of their target audience. This is where analytics can play an effective role. Analyzing the trends from different sources such as Google search, YouTube trailer views, ratings on IMDB/Rotten Tomatoes, weekly collection reports of similar genre, and star cast can help to predict the success of a particular movie.
The analytics around movies can be implemented for both movie-goers and theatre owners simultaneously. The initial goal is to predict the box-office returns of a particular movie and the objective to set up logic-driven relation with little prior information of box-office records and awareness among audience. The data scientist might have a significant amount of information about the movie, which includes director(s) past movie records, casts, story-types and marketing data. This knowledge feeds into the analytics engine, as the main analysis is driven only by the record/data available of that movie.
Apart from these factors, another critical one which will determine initial box office success is the movie trailer engagement. YouTube trailer search will reflect how a particular movie is trending in comparison with other movies releasing around the same time.
A sample step-by-step data analysis can be summarized as:
- First, categorize all the past and to be released movies by techniques such as cluster analysis.
- Post clustering a similarity check can be performed via silhouette plots etc. to judge whether the movie claimed to be “similar” to some movie is true or fiction.
- Based on the model obtained from past data, use similar movies over these models to obtain the net return estimate of the upcoming movie.
- To build an effective statistical model, also incorporate other factors such as awareness of the story which will ultimately help identify the human interaction effect.
In order to produce the more insight results, it is important to target the correct audience. The idea should be to track the moviegoers who are potential customers, and then run the most impactful campaigns. This can include YouTube suggestion paths and basic clustering methodology to create potential customer build-up. Even theatre/multiplex owners can strategize which movies to play, when and where, in order to gain maximum occupancy rate. Theatre owners need to account for the audience’s preference, which will vary, depending on factors like demographics.
In addition, seasonality should be factored into the analysis as appropriate –movies releases around festivals, holidays and even weekends. On a particular week, several movies get released simultaneously. They can have a cannibalistic effect, as people usually go to one movie during the week. Hence, it is extremely important to look for search/view ratios rather than their absolute values. These prior reports will not only help predict profit, but also if more marketing is required.
More and more analytical models will play a greater role in the motion picture industry by contributing towards superior marketing strategies that better predict the overall success of each movie.