The open-topic project commenced with a hunt for publicly available large data-sets. Each of the team member divulged into areas of personal interests before converging on "Entertainment" as topic of common interest for the group.
We explored video games data, TV-shows and movies data. With each of the topic we brainstormed possible hypothesis and problem that can be addressed or solved better with visualization. Then we went back into the data sources identified the hypothesis can be answered from the available data and cycled back-and-forth.
Once we had a set of hypothesis for each of the topic, we debated on the data sufficiency for each of them and then the team voted for each other favorite hypothesis, before converging on a few interesting ones.
The project would visualize the team of casts, directors, producers, studio and genres and any combination of them and see the trends in the success of movies in terms of budget, popularity, profit or ratings over time. The visualization could possibly provide insights into attributes that could potentially produce a more bankable team.
The visualization would also be interesting for addressing issues of diversity in movies. For example, how many roles does an actress play in her twenties, and how do the number of roles, genre of movies, and the success of those movies change over the course of her career?
There are many groups who may find this visualization interesting. Movie buffs would want to follow their favorite actors and be able to see trends in actors’ success rates. Producers and casting director and agents could explore certain insights into what makes an actor/actress successful. It could also serve as a useful resource for journalists interested in researching actors' careers and associations.
Extensive APIs for retrieving meta-data about the movie. Data: Release date, cast, director, writer, producer, studio, keywords, genre, reviews and ratings.
Information about movie budgets for the 4,326 films with the highest budgets of all time. Data: Movie release date, genre, production budget, domestic box office
Detailed statistics on domestic and international box office sales. Data: Opening weekend box office, theaters, total gross box office
Information about critic ratings and reviews. Data: Name of critic, date of the review, and the publication the review was from
With a few favorite ideas bookmarked, we spent a 3h session on iterative design exercise for visualization design. In the first iteration we sketched 8 quick ideas for each of the hypothesis in about 5 min. After this, the team entered into a design critique discussion on the sketches and talking about the strengths of each.
In the second iteration building on each other sketches and combining ideas, a more elaborate but fewer sketches were done and critique was done again. After this, we dispersed with the task of coming up with one single collaborated visualization sketched in detail. This would be the pre-cursor to the three independent ideas for the poster session.
This design focused on the problem of building a “bankable” team. We wanted a user to be able to search the database of casts, crews, and production companies and build a team around them. Then the user can select what aspect to compare: Budget, profit, critic rating, and user rating amongst the players in the team.
The visualization presents interesting ways to quantify what makes a team “good” for a movie. However, the goal of finding a “bankable team” was too much of a machine learning or computing problem rather than a visualization problem. Instead of focusing on a bankable set of actors and producers, we would instead be able to explore trends of casts, crews, and production companies to create a “desirable” team. The visualization will provide information about the cast, crew and production companies and the user can make his/her own deductions from it and attach sentiments to the information and thus find out what is the best team for him/her.
Insights: Budget, revenue, rating trends, bankability of the team
This design looked at the trends of casts and movies that have won Oscars and Golden Globe awards. Originally we wanted to also be able to compare with movies that have been nominated or won Golden Raspberry Awards, as it would provide an interesting contrast between highly acclaimed movies and what is objectively considered the “worst” in movies. This was seen to be more of a machine learning problem rather than an information visualization issue. We realized that this was a machine learning problem because here all our hypotheses were objective and could be solved or proved by bring the data together from different places, rather than involving the subjective choice of the user.
We decided not to move forward with this visualization because, along with being a computing problem, it would prove to be difficult to retrieve the awards data from various sources and make sure it is all consistent.
Insights: Differentiate award winners vs nominees Correlation between number of awards won and box office success
This design was indeed to examine historical trends, both in general and by genre of the colors used in poster. This could be useful for graphic artists interested in what palette to use for creating a poster for a particular movie that emulates past trends. The idea was based on Vijay Pandurangan’s visualization of poster hues from 1914 to 2012, which shows that movie posters have been getting darker as well as incorporating more blue. We wanted to investigate these trends more specifically, e.g., by frequency and by genre, and allow for comparisons by features such as movie rating.
Although some viewers thought that it was the most novel idea discussed, others found it more obscure or difficult to interpret. As one commenter put it, “I was confused as to your motivation for the color of the movie poster. Are you asking the question of understanding how the colors used in a movie poster influence box office success ? Or is this another way to understand the temporal history of movies, and their choice of color on the posters over time ?” Ultimately, we decided that this design had a relatively narrow scope and catered mostly to graphic designers and movie fans who were interested in color use.
Insights: Trends in movie poster colors overall, over time, and by genre, Correlations between colors and revenue
In our designs, our visualization emphasizes that making movies is about personal tastes and preferences by offering a variety of variables to choose from to accommodate those different subjective preferences. For example, instead of a ML problem of creating an ‘optimally profitable’ team that consistently brings in the highest revenue out of of a pool of actors, we allow users to make comparisons based on variables that represent subjective choices, such as the observation that an actor who has worked with another many times in the past might enjoy working with that person again.
In the poster presentation, we presented three initial designs for a system involving movie trends and data. The first design stuck closest to our initial proposal and is ultimately the design we chose moving forward based on the feedback gathered from the poster session. We found it to be the most compelling visualization problem (of the three) as the definition of a ‘good’ team is very subjective and it is this problem that everyone faces when making a movie. Allowing users to save different teams to compare and contrast on various attributes was another suggestion from the poster session that we implemented in our final design.
There are many groups who may find this visualization interesting. Movie buffs could follow their favorite actors and be able to compare works of the actor along with other actors for different production companies. It could also serve as a useful resource for journalists interested in researching actors' careers and associations. Producers and casting directors could find this tool especially useful when choosing stars who have worked together successfully in the past, or who have experience in a given genre.
Making movies is a matter of taste. When casting a movie, for example, casting directors might be interested in a variety of variables: What projects an actor has participated in in previously, how large those projects were (in terms of budget), and the success of those projects on critical and economical levels (ratings and revenue). The proposed system visualizes all of these dimensions in various frames.
Casting directors and other interested parties might also be looking for a specific mix of experience. The system includes frequent and successful co-stars in a Sankey diagram, displays trends in an casts or crew’s career, as well as in individual movie details. The system can cater to a number of different needs. If a budget for an upcoming project has been negotiated, it would be useful to find actors or directors who have done similar projects previously, or if a budget is currently too low but a certain actor would otherwise be a great asset, the tool might be useful for helping a producer convince studio executives that it is necessarily to raise the budget somewhat to attract the most appropriate lead.