Matthew Marchetti was among thousands of Houstonians motoring through the turbulent, murky stormwaters of Hurricane Harvey as the sun set on August 27th, 2017. He was sitting in a stranger’s dinky motorboat, attempting to rescue neighbors and shuttle them to safety. The problem was that he didn’t know where to look — neither did the police and fire departments.
“It was a big ol’ mess,” Marchetti says. As he disembarked to high ground, he phoned his friend Nate Larson, and said, “You know, we should develop a little website. Just for our neighborhood.” His idea was to create an application where a family in distress could quickly submit a call for help containing their location and information, which would instantly appear on a map. A responder could pull the location in order to execute the rescue. Once the family was safe, the information would be taken down so rescuers could focus on those still in need.
This idea — the “little website” — would balloon into something much larger.
Marchetti and Larson built a web-based geolocation service that collected data from social media, centralizing and visualizing the calls for help. It became a clearinghouse for volunteers and their boats, who could then be dispatched to help with rescues. As they worked in their company’s office through the night to build the site, the hurricane’s gusts pounded the windows. Water seeped in from above. The power flickered. But they finished, put about 25 people into the system, and went to bed.
When they woke, their service had blown up: there were over 1,000 entries.
“We thought to ourselves, ‘We’re screwed.’”
Marchetti says. “We’re going to be on the news tonight… ‘Open developer idiot develops a website that doesn’t rescue anybody.’” The site gained traction as it was shared across social media and through word of mouth. As the storm crested, the amount of rescues swelled. At any given time, there were 40,000 to 60,000 people on the website.
At least 25,000 people were rescued in Houston using the app, Marchetti says. High-density rescue locations — such as a nursing home, an apartment building, or a block with multiple potential rescuees — might have been undercounted. For this reason, the number of rescues could be even higher. “It’s kind of hard to pin that number down concretely, but we’re working with a lot of universities going through that data trying to figure it out,” Marchetti says.
The service — now known as CrowdSource Rescue (CSR) — was meant to fill the deficit of public services during a time of immense, dizzying catastrophe. CSR reduced the redundancy created by reposting and sharing across multiple platforms. It crowdsourced every part of the operation: posting, dispatching, rescuing, and updating. It allowed Houstonians and outside volunteer organizations such as the Cajun Navy to work hand in hand with public officials.
Marchetti cites numbers to back this up: CSR has spent about $25,000, in addition to donations, to conduct 37,000 rescues to date. FEMA, by comparison, spent $90 million and conducted just 9,000 rescues in the wake of Harvey. The Houston Fire Department owns just 17 boats, but thousands of boat operators were using CSR.
As Harvey subsided, Marchetti prepared to return to his day job in real estate. Days later, another supersized storm was barreling toward the US: Hurricane Irma, which triggered the evacuation of millions. So, Marchetti and Larson quickly retooled and improved CSR to make what they say were a couple thousand rescues.
Weeks later, Hurricane Maria pummeled Puerto Rico, knocking most services offline and displacing most of the island’s residents. Later, an earthquake rocked Mexico City, injuring thousands. Then, wildfires swept across Northern California, charring hundreds of thousands of acres. With each disaster, Marchetti and Larson added features, released new versions of the app, and increased the reliability and ease of access.
“I guess this is just my life now.”
CSR has guidelines for when to deploy its service: are 911 resources outstripped? Are volunteers dispatching themselves into the field? Do they need a system to support their efforts? From there, “We then try to determine if we will be effective and what our overall strategy will be,” Marchetti notes. If CSR can help coordinate an organized and effective response, then it will go in.
“I guess this is just my life now,” Marchetti says. His motorboat epiphany has turned him into a kind of unofficial spontaneous volunteer rescue czar. As he and Larson gear up for the next big storm, they are on the brink of releasing a new version, a process Marchetti describes as “very cathartic” since he hasn’t been forced to develop it in the throes of a hurricane.
People are paying attention. Companies, including Google and Accenture, have chipped in resources and spotted technical advice and training. (Google did not respond to a request to comment; an Accenture spokesperson declined to comment.)
Though Nextdoor “does not have a partnership with CSR,” Nextdoor spokeswoman Jen Burke said that the company confirmed that a senior employee worked with CSR to make it possible for CSR users to also share rescue information on Nextdoor. That person also introduced Marchetti and Larson to Nextdoor’s chief architect, Marchetti says. State and local governments are also in conversations with Marchetti to interface with CSR, he says. He and Larson have partnered with academics and universities to mine the vast tranche of data they’ve acquired for insights and suggestions for improvement.
During Harvey, the US Coast Guard instructed storm victims to only call 911 or a list of designated numbers if they needed to be rescued. “If busy keep trying. Do not report distress on social media,” they tweeted. The agency has since taken a measured step toward reform. “The Coast Guard is examining the lessons learned in order to find the best way to incorporate social media data into our disaster operations response,” said USCG spokeswoman Lisa Novak. But “social media has not been monitored continuously for distress calls, and this has not changed. Calling  should be the first action for emergencies.”
CSR is an imperfect tool, and superstorms are chaotic by nature. Also, CSR is just a five-person organization. Marchetti and his team cannot verify the legitimacy of some calls for help that are posted on his site. (This is also the case with 911 calls, he says, Local law enforcement officials declined to comment on this remark.) To address concerns about false posting, CSR takes stock of a post’s metadata. “If you’re sharing that someone needs to be rescued here in Houston, but the IP address is out of Russia… we’re going to be a little suspicious,” Marchetti says. But he concedes that there is no way to always know for sure.
“Responding to a disaster as a civilian is not without its risks.”
Another issue is the safety of the spontaneous volunteer rescuers. With vigilante rescues, some individuals will inevitably get in over their heads and become part of the problem rather than the solution. “Responding to a disaster as a civilian is not without its risks,” Marchetti says. To minimize these risks, CSR uses a geofencing technology that blocks out dangerous areas. If emergency management officials have declared an area dangerous for civilians, a volunteer won’t be dispatched to it. The app also lets users mark hazards such as downed power lines or swift water.
“We had amazing volunteers, truly amazing. But not one of them had a Black Hawk helicopter,” Marchetti says. Now, CSR is teaching first responders and local and state agencies how to use its service. When command and control centers communicate with CSR, rescue and response efforts can be coordinated and real time.
With “over four-hour wait times on our 911 lines… [and when] the storm gets to that point where we can’t satisfy the demand, [and we] need to flip the switch and get outside resources, now we can do that,” says Jesse Bounds, director of innovation for Houston Mayor Sylvester Turner. Next time “we release the cavalry,” Bounds says, referring to spontaneous volunteers, “they will be pre-vetted and automatically plugged into our networks. You won’t have the need as much for this person with the kayak who thinks he can go off and rescue a whole family.”
CSR will provide the government with a “better, more coordinated response once their resources have been strapped,” Marchetti says. Cooperation, communication, and coordination win out over mass chaos, he says, and with good teamwork, “we can triage better, respond quicker and do it more effectively.”
As the water receded from Houston back into the bayous, reservoirs, and Gulf Coast, Marchetti pulled the Harris County death records to see how victims of the storm died. “Half of the stuff could have been avoided,” he says.
The 2018 hurricane season is just around the corner. It’s expected to be “worse than usual.” When the next superstorm hits, authorities will likely find themselves overwhelmed, short on people and in need of help. With a new-and-improved application, an army of spontaneous volunteers, and a real-time dialogue with public officials on the front lines, “we’re going to be in a really good position,” Marchetti says. The “community will respond and bounce back quicker. And lives will be saved.”
Correction: This article originally gave the wrong first name for Matthew Marchetti in a caption, and wrongly identified Harris County. Also, after publish, Marchetti explained he’d hired two more people — so CSR is now a five-person org.