Can machine learning help more people access SNAP benefits?

Thanks to a machine learning model built in-house at Center City nonprofit Benefits Data Trust, call-center staffers get extra insights while enrolling users onto the Supplemental Nutrition Assistance Program (SNAP).

Deployed last week for Pennsylvania residents, the tool helps employees understand what level of assistance potential beneficiaries might need.

BDT Director of Data Science Matt Stevens said the model is already in use, helping workers identify cases that may require a more hands-on approach through the application process, specifically for the documents submission phase.

“This isn’t making distinctions on how much benefits people can have access to, but rather what level of support we should provide,” Stevens said.

The model, Stevens said, is in the initial phases of use and is expected to be deployed across all projects. BDT, which provides outreach and enrollment assistance to public benefit programs, has helped 650,000 people access $7 billion in benefits to date.

“So far, they’re finding guidance helpful,” Stevens said of case workers at the 13-year-old nonprofit, which employs 160 in Philly. “[Previously] there were no tools to understand who needed what.”

In August, BDT announced a $4-million grant from Walmart would let it enroll some 45,000 people across six states onto the federal assistance program.

“We hope this will mean we’re delivering a service in the way people want to be served,” said Stevens. “We hope that we’ll be able to help more people because we’ll be able to embed efficiencies in the landscape across the spectrum.”

According to the nonprofit’s CEO, Ginger Zielinskie, tech tools like the algorithm have helped the organization become more efficient in its delivery of service. Internally, the use of business intelligence software Looker has also helped streamline the way BDT presents results to its funders.

“It helps us articulate ROI,” Zielinskie said. “Our program managers can provide real-time data.”