Will computers pick out fashion's next top models?

Body size, height, and even shoe size can influence a young fashion model’s success. But what about her Instagram profile?

According to a new study, “a strong [Instagram] presence may be more important than being under contract with a top agency, or than the aesthetic standards sought after by the industry.”

It’s not news that social media celebrities have become powerhouses. What’s interesting here is that University of Indiana researchers tried to decipher what qualities in a person’s social media profile might hint at future success and came up with a formula, a sort of automated draft pick for models. If it actually ends up working, it could be worth a ton to the $1.7 trillion global fashion market.

“The long term the impact of our work is fundamental: in a landscape that is changing, agencies and casting directors want to have top-notch technology to make the best out of the information freely available online,” Emilio Ferrara, the computer scientist who led the study, told me over email.

For the study, Ferrara and his team focused on a set of newbies—young female models who were just coming up the fashion ranks. They scanned a database called the Fashion Model Directory for girls filed under “New Face,” and came up with 431 girls. They assessed how the girls’ hip and waist size and height—as well as their Insta-presence—affected their chance for success, in this case measured by the number of runway shows they were hired to walk in during the 2015 summer/spring fashion weeks in New York, Milan, London and Paris.

For Instagram, they looked at the number of posts each n00b made in the three months leading up to these fashion weeks, as well as the number of likes and comments each of those posts got. They also ran sentiment analysis algorithms through the English-language comments to figure out what kind of reaction they were getting from their audience.

As expected, taller models with smaller dress sizes, hips and shoe sizes got more jobs. Those with top agencies, also had a better shot at getting hired. Of the 431 girls they found on the Fashion Model Directory, only 253 of them had an Instagram account. More posts and more positive comments were positively correlated with a better chance of walking in a runway show.

That’s interesting enough, but then, they wanted to see whether their software had learned the qualities that make a young model successful. So they tried to see if their algorithm worked to predict the 2015 fall/winter bookings for 15 new models in the database.  Ferrara’s algorithm was able to correctly pick six of the eight “successful” models who got bookings and six of the seven who didn’t land jobs. According to the researchers, a social media presence is as important a factor as having the right body type in predicting a model’s chance for success.

The study is small, a fact the authors readily admit. And according to their data analysis, it does seem like being Insta-famous, in some cases, doesn’t completely equal the playing field between models working with top agencies and those who aren’t (confirming our own findings).

But Ferrara’s work is the latest example in a growing breed of algorithms attempting to quantify the subjective. A few years ago, researchers used Twitter and Wikipedia data to predict how well newly released movies would do at box office. There are algorithms that try to predict box office success by scanning movie scripts. Others try to put a price on fine wines. Recently, Microsoft researchers trained a computer to detect humor using New Yorker cartoons, in the hope that it’ll be able to help judge what’s funny. Software is moving toward the realm of the abstract, driven by access to tons of data, and more powerful computers.

“Up until recent years, there was the tendency to think that some qualities are ‘ephemeral’ and impossible to quantify; beauty standards were considered among the most inextricable ones,” Ferrara told me over email. “The new wave of research under this umbrella [demonstrates that] machine learning can…. successfully capture the dynamics of these systems and make credible predictions.”

So what does this all mean for the future of fashion? Does this signal that scouts will go the way of other professions like call center representatives and drivers? Probably not. If there’s anything I learned from binge watching Project Runway it’s that a model’s success also depends on how well she can strut down the runway. If she’s got a limp walk, chances are she’s not going to make your designs look sexy, which means she’s out, no matter how big her Instagram or Pinterest following is.

Daniela Hernandez is a senior writer at Fusion. She likes science, robots, pugs, and coffee.