The Science of Blockbusters: What Makes a Good Story?

Have you ever watched a great movie and thought it was great, only to find out that your companion thought it was terrible? Writing feels subjective, which makes it hard to evaluate objectively. Is that intuition misguided?

The code on what makes a good story is being cracked by a Wharton marketing professor. He and two colleagues devised a way to measure the success of a narrative. They compared movies, TV shows, and academic papers, and the results were published in a study.

Berger was at Wharton talking about the study. You can listen or read the transcript of the conversation below.

What inspired this study and how did you measure it?

We consume narratives constantly. We are reading books, watching movies, and reading online. We are also creating stories. When we make a presentation, when we give a talk, even when we write an email, we are creating content that is like a story. Some of these things are more successful than others. Some movies are big and some are small. Some books are popular and others are not. Some written content and online articles are so engaging that we can put them down. What makes a hit? Why does something do well while others don't? Can we use language to understand that?

“When we make a presentation, when we give a talk, even when we write an email, we are creating content that’s like a story.”

What was the key lesson from Wharton?

There is a science here. We often think that movies are just magic, creative process where things gel together and there is no way to know if it will succeed or fail. That is not right. It feels like magic. When we are watching a great movie or reading a great book, we are focused on the narrative and don't think about anything else. There is a science. Understanding the progression of ideas can help us understand the science of stories. Computational linguistics and natural language processing can shed light on some questions that would otherwise be impossible to uncover.

You measured three things: speed, volume, and circuitousness. What did you find about them?

Think about traveling in a car. You can drive faster or slower. What does it mean to drive faster? You cover more distance in the same amount of time. When you travel at a greater speed, you go a few extra miles.

We can say the same thing about stories, content, or narratives. Think of a story as a collection of ideas that come together over time. Maybe there is a wedding, and there is another scene in a wedding, and there is a scene after that about something else. The distance between those ideas can be measured.

A narrative, a story, or a piece of content can go faster or slower than a car in the same amount of time. It can talk about two things that are related, or it can talk about another thing that is unrelated. This can be seen in textbooks. Imagine opening up an Earth science textbook from your high school days, and the beginning of the chapter relates to what happens next in the chapter. If you went from the beginning of a chapter to two or three chapters down, it would be further away because the content is not relevant.

Embedded tools allow us to take text, put it in a multidimensional space, and measure distances. We can look at the first chunk of the movie and the next chunk to see if they are closer together or further apart. Are they traveling faster, moving before other ideas, or slower, moving from one idea to the next?

The idea of volume is the same as the idea of moment to moment. Some stories cover a lot of ground. They cover a lot of different things. The stories are a bit more narrow. There are a few things that are closer together. You can also measure the volume. A story has a set of points. All of those points can be taken together and asked if they are closer together or further apart. If someone goes for a 4-mile run, they can either go around a 1-mile track or go on one 4-mile loop.

“We can understand the science of stories, of content more generally, by understanding the progression of ideas.”

How direct are the ideas? Are we moving through the shortest path possible, or are we taking a more indirect route? The fastest way to get through the twelve points on the clock is to go around the circle and hit each of them. You could go from twelve to six to one to seven to three to nine. The volume would be the same if you covered the same points. It is still inside the clock, but it would be a more indirect route.

It's the same thing with stories. Do stories follow a linear path from idea to next idea to next idea, or do they double-back, touching on similar things that they have touched on before, before moving on to unrelated ones in the future? We measured each of these three things in a variety of ways.

What can marketers learn from Wharton's knowledge?

Whether we are a leader or a marketer, we are always creating content. We don't think of ourselves as speakers or writers, but we spend a lot of time speaking and writing. Almost everything we do on a daily basis involves language, either producing it through writing or speaking or consuming it through reading or listening. The findings help us think about how to better lay out the content in a way that will impact the audience. Is it better to cover a lot of ground or relate the ideas more closely to one another? If we are covering the same ground, should we use a very direct path or more of a spiral, where we go back to the same ideas again and again to deepen our understanding?

I think they have implications for content engineering. These tools can be used if I am trying to figure out which movies to green light, or if I am trying to figure out if a book will make the New York Times bestseller list. Our research shows that we can predict how successful movies and TV shows will be. These findings have useful implications for us. We think about telling stories as a way to have fun, but we are not. This work has implications for how we should lay out our ideas, because we are constantly laying out narratives to explain them.

“Almost everything we do on a daily basis involves language — either producing it through writing or speaking, or consuming it through reading or listening.”

You covered movies, television shows, and academic papers in this study. I'm curious about social media, which you didn't study. Mass communication on social media has become very convenient. What do you think could be done to apply this study to social media?

Academic papers are not the same as TV shows or movies and that is why we wanted to look at them. While we can measure the speed and volume of both movies and academic papers, the features that matter are very different. Speed is bad for academic papers. Volume is bad for TV shows.

To think about your social media question, we need to think about what we hope the language, ideas, and arguments will do. Are we hoping they will be fun and interesting? Speed can be good and volume can be bad. A different set of features might be valuable if we are trying to impart knowledge or convince people of something.

We are doing some work on social media. Think about yourself as a brand or aninfluencer. You can think about the volume across multiple pieces of content because of the speed or volume of one piece. I'm going to post about X today and next week. Is it better to post two buckets of things or one? How should I plan on increasing the impact of my content? We are calculating things like speed and volume over time across the content they produce to look at how different ways of laying out ideas may be more or less impactful.

Is there anything else that marketers and advertisers should know about this study?

Content is created by marketers and advertisers. We are making sales pitches, whether it is ads, content marketing, or as a salesperson. The findings have implications for how we should present our ideas. Should we jump from benefit to benefit or should we focus on one thing? Should we try to talk about everything a product can do, or should we stay focused? As marketers think more and more about content marketing across a variety of platforms, I think these findings have some clear implications for improving content marketing, making ads more effective, and helping salespeople sell a bit better.