As an astronomy PhD student at Harvard University in Cambridge, Massachusetts, Nathan Sanders learnt statistical modelling to analyse supernova explosions. Now, he works for Legendary Entertainment in nearby Boston, applying those quantitative skills to predict which stars and story lines can make a film into a commercial success.

When did you consider leaving astronomy?

I had learned a new computational framework in a statistics course. As I applied those techniques for my thesis, I realized that I loved what I was doing, and the reason had more to do with the statistical models than the astronomy applications. That made me open to new opportunities. I thought I would be doing a disservice to myself if I didn't explore them.

What appealed to you about this position?

When I was hired in 2013, Legendary had just launched its applied-analytics division in Boston. It felt like an opportunity to rethink and reinvent the way that companies pick which films to make and how to build support for them. The goal was to be the first in Hollywood to make decisions on the basis of data and evidence rather than on intuition.

Besides technical skills, what do you look for in candidates when you recruit?

Communication is key. You have to be comfortable with diverse concepts, and talk to business people, filmmakers and technical colleagues.

How did you hear about this position?

I emphasize the importance of volunteering and getting out into the community. As a first-year graduate student, I started a project called Astrobites, a collaborative writing project that creates a Reader's Digest version of astronomy literature. I also volunteered with an organization doing live science demonstrations. The executive director was a friend of the chief analytics officer at Legendary Entertainment. It was one of those random connections that so often creates a job opportunity, but that can be hard for scientists to foster if they are completely focused on their thesis work.

This interview has been edited for length and clarity. See go.nature.com/2bix4y7 for more.