While governments and businesses have reaped big benefits from the revolution in big data and software analytics, nonprofits sometimes lag behind the public and private sectors when it comes to adopting advanced machine learning techniques.
Gideon Mann, head of data science at Bloomberg is trying to close the technology gap and find ways for nonprofits to collaborate with data scientists.
The aim, says Gideon, is to help nonprofits to structure and analyze their data so they can draw insights from it and achieve greater impact.
It’s an idea Gideon has championed with his work on the Data for Good Exchange, Bloomberg’s annual conference that explores how data science can help academia, industry, government, and NGOs work better together to solve broader societal issues.
“Our business is centered around how to gather, analyze, distribute, and store data—and we’ve been a strong partner in the communities we operate for a long time,” says Gideon. “So when we started the Data for Good Exchange, it fit very nicely into the overall corporate agenda.”
It’s a rewarding effort, but it comes with unique challenges. One of the biggest: growing a data-driven culture in organizations that have not yet embraced the digital age.
Data doing good
Change may be gradual, but Gideon notes that there are already examples where nonprofits are benefiting from data science.
For instance, conservationists in Africa have piggybacked on research done by Tanya Berger-Wolf, a computer scientist at the University of Illinois, who found that zebras have unique stripe patterns akin to barcodes.
“You can estimate the number of wild zebras by looking at the biomarkers on their skin,” Gideon says. “By looking at photos, you can get estimates about the size of the herds, and then you can do conservation plans.”
He also cited the experience of the fire department in post-Katrina New Orleans, which had received a huge donation of smoke alarms from the Red Cross and the Louisiana State Fire Marshal. While the firefighters were eager to hand out the units, they were unsure where to allocate them. Between 2010 and 2014, structure fire led to 22 deaths in New Orleans. In all of those cases, no smoke alarms were present. Fire alarms could prevent many of these deaths.
Director of Performance and Accountability for New Orleans, Oliver Wise, helped address the problem by collaborating with a data analytics firm to build a predictive data model using public data that identified the areas of the city least likely to have working smoke alarms, yet most likely to sustain fire-related fatalities. The New Orleans Fire Department used this data to conduct a targeted door-to-door outreach program.
In both cases, Gideon says, it was about applying “machine learning and data science to problems which are very practical, very immediate and make a concrete good.”
Making sense of a mountain of data
One challenge that both businesses and nonprofits face is finding golden nuggets of information in a mountain of data. It’s especially hard for organizations that don’t have trained computer scientists.
During a recent appearance at the Structure Data 2016 conference in San Francisco, for example, Gideon explained how Bloomberg can help UNICEF track the impact of commodity price fluctuations on the food supply chain in regional populations through tapping country-level statistics stored on Bloomberg terminals.
Knowing ahead of time that there is a commodity shock that affects key resources can enable local organizations to anticipate and respond to a potential crisis.
“Mozambique relies on wheat imports from Russia. If a fire decimates the Russia harvest and causes export controls that drive the price of wheat up 30%, there are going to be riots,” says Gideon.
For Gideon, it’s part of a dialogue he hopes Bloomberg will continue to foster between data scientists and nonprofits, “so that if they have a question, they can at least start a conversation—and that can open the door to greater outcomes,” he says.