Method is 80 percent accurate in identifying most popular models for the following season

With Fashion Week kicking off Sept. 10 in New York City, everyone wants to know who will be the industry's next top model.

The answer may lay far from the glittering world of runways, magazine covers and star-making designers. Researchers at Indiana University have predicted the popularity of new faces to the world of modeling with over 80 percent accuracy using advanced supercomputational methods and data from Instagram.

To conduct their analysis, IU scientists gathered statistics on 400 fashion models from the Fashion Model Directory, a major database of professional female fashion models, tracking hair and eye color; height; hip, waist, dress and shoes size; modeling agency; and runways walked.

"Popularity" was defined as the number of runway walks in which a new model participated during the Fall/Winter 2015 season in March. Data for the study was collected in fall 2014.

The team then analyzed accounts of the models on Instagram, using the social media platform to catalog each user's number of followers, number of posts per month, number of "likes" and comments on those posts, and whether these comments were generally positive or negative.

To test their ability to predict a model's popularity, IU researchers narrowed their focus to 15 models listed on the FMD as "new faces." Of the eight models expected to achieve the greatest popularity, six were accuracy identified. Of the seven predicted to score lowest in popularity, six were also accurately identified.

The six most popular new models of the Fall/Winter 2015 season were Sofia Tesmenitskaya, Arina Levchenko, Renata Scheffer, Sasha Antonowskaia, Melanie Culley and Phillipa Hemphrey.

"Our analysis suggests that Instagram is as important as being cast by a top agency in terms of its ability to predict success on the runway," said Emilio Ferrara, a computer scientist at the University of Southern California who conducted the research at the IU Bloomington School of Informatics and Computing's Center for Complex Networks and Systems Research and as a member of the IU Network Science Institute.

Other IU contributors to the research were Giovanni Luca Ciampaglia, a postdoctoral researcher at the Center for Complex Networks and Systems Research and an assistant research scientist at the IU Network Science Institute, and Jaehyuk Park, a doctoral student in informatics.

When analyzing the more established model's Instagram accounts, the IU team found that a high number of likes and comments, as well as frequent posting, were associated with success on the runway, although the tone of the comments did not affect popularity. A higher than average number of posts yielded a 15 percent higher chance of walking a runway, but, surprisingly, more "likes" could lower these chances by about 10 percent.

"Social media is changing the game dramatically," Ciampaglia said. "Traditionally, models don't interact with consumers; but now their online activity plays an important role in popularity and, ultimately, success."

In addition to Instagram activity, Ferrara said the greatest factor associated with runway success was representation by one of 20 top modeling agencies.

Another major correlate of popularity was height, with an additional inch over the average roughly doubling a model's chances of walking a runway.

Larger dress, hip and shoe sizes were all negatively associated with the chances of walking a runway, suggesting that a controversial bias for slender and thin models remains in the fashion industry, Ciampaglia said. Waist size was found to be neither a positive nor a negative predictor of success.

The team also found that the majority of new models do not walk any runway, with only 24 percent given the opportunity, most likely owing to strong bias toward established models.

The focus on "new faces" was an attempt to overcome the powerful influence of "cumulative advantage" on popularity, Ciampaglia added, borrowing a term from the social sciences in which the greatest predictor of future success is past success. Researchers also refer to this as the "Matthew effect," a reference to Matthew 13:12, "Whoever has will be given more."

He pointed to Kendall Jenner, the increasingly famous sister of reality star Kim Kardashian, as a poster child for this effect.

"We chose the fashion industry for this research because it represents a strong 'winner-take-all' mentality," Ferrara said. "This aspect of survival of the fittest, plus the large amount of statistical data on professional models, makes it a perfect subject for advancing research on 'the science of success.'"

This area of study, which originated in the sociology of science, started out trying to measure the prestige of research studies, later branching out to analyze factors contributing to career success in the sciences, as well as popularity of cultural products such as music, art and literature. IU's other contributions to the field include a recent study tracking the path by which some research papers go ignored many years before suddenly "awakening" to widespread citation.

In the future, the IU scientists aim to apply the methods developed for this study to predict success in careers beyond modeling. They also note the potential for a similar analysis of the world of male fashion.