- A Cure for the Common Meeting
- Big Data and Hollywood:
A Love Story
- The Race to Probe the Twittersphere
- Watson Takes The Stand
There is a core tenet in the study of theatrical performance which says that an audience is more than a just a passive receptacle. It lives. It breathes. And as such it possesses the power to inform creative pieces as they unfold.
Hollywood, and movies in general, interrupted this relationship. As a packaged medium, film robs us of our feedback outlets. And as a result, we no longer sling putrescent produce at stages. Nor do we usually clap after a good movie, not because we’re indifferent to what we’ve seen but because there is no one there to applaud.
But it would be very wrong to think that audiences have no role in the Hollywood machinery. Rather, the role has shifted. We now throw our tomatoes online—in blogs and twitter feeds and Facebook timelines. A quick scan of the Internet will tell us that audiences are just as vociferous as ever.
And now, according to those who are plotting the convergence of big data and big entertainment, Hollywood is using technology to try to recapture what was lost in the relationship between creator and consumer. Newly emerging tools are empowering big studios to convert massive quantities of movie-goer reactions into meaningful, actionable insights into what works and what doesn’t. With big data analytics, movie executives are keeping one ear to the audience and the other to the craft in a way that is dramatically altering how movies get made, marketed and distributed.
Many, especially those on the business side, welcome the change. “My dream is when Hollywood really starts looking at data, and using data in a big way, and it’s driving business value,” says Richard Maraschi, global leader of advanced analytics at IBM.
In that role, Maraschi is helping the studios leverage the wealth of feedback that consumers put online to make better creative and marketing decisions.
No matter what step in the process you look at, Maraschi says, making more successful movies requires getting a high-resolution picture of your audience. And until recently, this snapshot was very soft-focus.
“In the movie industry we use the quad segment,” says Maraschi. “Male. Female. Over 25. Under 25. Boy, that’s very granular isn’t it? That’s basically what the studios use in terms of segments when they do panels and polls.”
That blunt instrument is fast giving way to computers that can render us in fine detail by picking up the trails of digital breadcrumbs we leave online and building them into predictive models of what we like and don’t like. As a result, the quad segment approach to audience profiling is being replaced by a profusion of very fine-sliced profiles that more accurately represents the sensibility spectrum of a modern audience.
A Conversation with Richard Maraschi, Global Solutions Leader, Advanced Analytics, IBM
Can you explain what IBM is doing in Hollywood?
There’s a joke in Hollywood where they say, “How do you market a movie?” And all the studio executives get on a rooftop and scream, “Come see my movie!”
That is how a movie is marketed. If you see a movie coming, you see it on every street you’re driving on around New York. You see it on every possible TV channel. You see it online. It’s definitely a mass marketing type task.
That’s because they’re so paranoid and scared that they won’t get enough people to come see it on the first weekend, which is very predictive of the life of the film. And you can’t necessarily change a film. But you can do some marketing and advertising to get people to see it.
And that’s where IBM comes in?
We are one of the leaders in predictive analytics. That could be to predict what the box office is going to be in three months for a new movie coming out. That could be creating a “people that like this also like that” kind of scenario. It could be creating clusters and segments of movie audiences.
With a blockbuster film, 50 or 70 million dollars of the marketing is spent within 10 to 12 weeks. That’s a lot of money. So [IBM helps Hollywood] get smarter about how you spend that money and become more targeted and effective.
Can you give an example of how it does that?
Lots of people can tell you [audience] sentiment and use keywords and see what the sentiment is based on that. We can do what we call deep sentiment analysis, which is parsing and categorizing sentiment into its features.
If you think about a film, it would be certain characters, the music, certain plot elements, a feeling about the film, a scene in the film. You can start to get down to that level and then tie that back to who said [what on social media], what kind of people said that. Was that on target on our audience? Is there a new audience that’s talking about this film?
If you’re trying to build any kind of analytics or consumer analytics platform, you need to collect all this data. It comes in different formats. It’s sometimes not clean. We have tools that can do all this information ingestion, quality cleansing, governance and integration. And we can do that for structured data and unstructured data.
And what can you do with that information?
Once you have all this, what can you learn? You can actually build what we call marketing optimization models where you actually use historical data to actually build models going forward that tell you, “Okay, next time around how do I want to allocate my marketing dollars by film, by media channel, by segment?” and start to do that.
That would be the holy grail: Starting to build these marketing optimization models that prescriptively have the data make these marketing budgeting allocation decisions—should I put more money into this film or not?—[and] have the analytics help the human to make the decision.
“Now we can get down to micro-segments,” says Maraschi, “like soccer moms in Florida that are really passionate about action films. You can start to get higher fidelity on understanding the audience. You need predictive analytics tools to do that stuff.”
Studios can use these real-time opinion assessments to do all kinds of tweaking after a movie has been made; targeting specific demographics in marketing campaigns, tailoring trailers so that they appeal to the kinds of people who will be drawn to a particular movie, pushing distribution to geographic areas where the target audience lives.
But what about the creative process? Can big data really tell writers what to write, or what kinds of screenplays are likely to sell? Can it tell directors anything about what and how to direct? Is there a secret formula that script scouts can use to vet the slush pile on their desks?
In other words, can audience responses be predicted on the basis of past behaviors and attitudes?
In order to do this, says Maraschi, you need to go beyond understanding who the movie goers are and find a way to assess exactly what qualities of a film are eliciting their reactions—to break down the work itself into its component parts to determine what works and what doesn’t.
As companies like IBM have worked on defining the specific attributes of audiences and tracking their sentiment, others are attempting something similar with the movies themselves. Netflix, for example, has sliced and diced movie content into more than 70,000 characteristics, creating an extremely granular mosaic of movie genres.
The same kinds of data could soon become available as a product to Netflix competitors through the Video Genome Project, which is run by a company called Structured Data Intelligence.
It is the hope of many in Hollywood that by combining deep understandings of both content and audience, studios will be able to choose and tailor their movies from the very start, and perhaps even identify some kind of magic formula to screenwriting.
“The movie is a snowflake,” says Maraschi, who is well aware of the controversy that bubbles up when data analysts attempt to capture and codify creative inspiration. “Well, snowflakes have a pattern to them and everybody says they’re all different. A lot of us don’t believe that’s the case. We haven’t gotten to the pattern. We haven’t gotten enough fidelity to get to those patterns. And that’s what is happening right now.”
Naturally, the people who actually write the movies are skeptical.
“Nobody knows what commercial is,” says Jay Douglas, a film professor at Loyola Marymount University. “It changes from day to day. And so you have to write something that you feel passionate about,”
Moreover, the dream of writing algorithms that analyze scripts and reliably rate their marketability may run counter to yet another tenet of the theater, which is that audiences usually watch a performance in order to see something they’ve never seen before, something they could never have imagined—not to participate in a big-screen version of a choose-your-own-adventure book.
Douglas, who teaches his students to think as little as possible about the market when writing their scripts, invokes the genius of Steve Jobs as a model, noting that Jobs turned Apple products into a global brand not by giving people what they already knew they needed, but by giving them something what they did not know they wanted, something they didn’t even know could exist.
Even Maraschi admits we are still a long way from understanding what makes a movie effective. “I think it’s a really exciting area,” he says, “but I don’t think we’re there yet. And I don’t think any Hollywood executive would tell you that we are.”
That may come as a relief to people like Douglas and likeminded writers, actors and directors in Hollywood, who worry that listening too closely to the market and past behavior is a recipe for mediocrity and stasis.
“Nobody will want to do anything that hasn’t been done in the past that stretches into different parameters,” says Douglas, “and I think that’s a really bad idea.” Then again, he adds after a pause, “this is show business”—with the emphasis on the “business.”