Amid college application mania, big data may be able to predict a Harvard "yes"

By Carmel DeAmicis , written on December 17, 2013

From The News Desk

It's December 17th. That means admissions application deadlines are fast approaching. High school seniors across the country are frantically penning version after version of personal essay, stress sweat leaving sticky marks on their keyboards. It's judgment time. Will they get into the college of their dreams? Will they get into even one college out of their 11 safeties?

It's a traumatic coming-of-age ritual that most college-bound only go through once. For these applicants, a magic eight ball would take a big weight off their shoulders. "Will I get into Harvard?" the bespectacled nerd would ask, shaking the sphere and waiting with apprehension for "yes" or "no" to appear.

In the age of big data, many things are possible. You know where this is going. Someone has invented a magic eight ball that predicts whether a specific school will accept a student. It's called StatFuse.

StatFuse's algorithm factors in almost all of a student's application data -- GPA, test scores, percentage of AP's, ethnicity, location -- and returns a percentage probability that she'll get into a specific school. If the percentage is 40 percent or higher, the student will likely be accepted or wait listed. If it's less than 40 percent, the school will probably reject the student.

"Counselors aren't as valuable at a high school as they were in the past because technology does what they used to do," StatFuse founder Jeet Banerjee says. "We're trying to create a more technologically connected space for college admissions."

How does the StatFuse team know its algorithm is accurate? They validated it by plugging in 5,000 profiles of students who have already been accepted to colleges, gathering the information from a website called Parchment. Banerjee says the algorithm was accurate 99.7 percent of the time with the initial 5,000 profile test run.

It's surprising, given that the algorithm can't take into account personal essays. "Obviously there are outlying factors," Banerjee says. "If the student found the cure for cancer there's no way to measure that."

To help concoct the algorithms Banerjee and his co-founder called different retired college admissions officers for guidance and analyzed the profiles of previously accepted students. They gave each factor a weight of importance in accordance to how the school weighted it, essentially creating a different formula for every college.

"The fundamental metrics are the same for every university, like GPA and SAT," Banerjee says. "Other questions have higher or lower ratings based on the specific universities, like if you have a family that's an alumni at the school."

He said the University of California system has one algorithm for all the schools, the Cal State system another, and Ivys each their own, When StatFuse launched, it only offered its algorithm for 80 of the most popular colleges, but since then they've grown it to 1,200.

StatFuse has flown under the radar up till now. The co-founders are based in Southern California and went through the accelerator K5Launch in Orange County. They've been bootstrapping, initially trying to make money through freemium features before realizing the key to prosperity through this platform was to partner with other companies.

Scholarship companies, financial aid organizations, SAT prep companies, and colleges looking to buy student leads are all willing to pay StatFuse a pretty penny for exactly that type of information. It helps that StatFuse can send them exactly the right students, whether it's people with poor SAT scores to the test prep companies, or those in Oregon interested in liberal arts colleges.

"In a public school, one college counselor is responsible for 500 students, everyone's vying for their time and expertise, that's not adequate," Banerjee says. "That market is collapsing. Parents and students are finding out they can do it from their own home."