How Google’s Bad Data Wiped a Neighborhood off the Map
Inside the big, twisted industry of neighborhood data collection
Annette Lott sat through the meeting with studied patience, waiting for the moment city officials would open the floor and she could ask them about Google Maps.
It was late spring in Buffalo, New York, in 2015 — a season that was unusually hot that year, and heated. The wood-paneled meeting room at Gethsemane Grape Street Baptist Church hummed with anxious homeowners from Buffalo’s Fruit Belt neighborhood, where a burgeoning, billion-dollar medical complex threatened to displace them.
Poor folks had called the Fruit Belt home for more than 150 years — first German immigrants, then African-Americans. Lott’s parents bought their pale turquoise two-story house in 1955, moving north from Bluefield, West Virginia, to help build up a community that would become the heart of Buffalo’s black working class.
Now that community was under threat — or so it seemed to Lott. The 66-year-old had Googled directions to her neighborhood and found that the app had changed the name of her community from the “Fruit Belt” to something called “Medical Park.”
Lott learned that the issue had been festering for years, and she wanted answers. The 2,300 residents in the Fruit Belt didn’t refer to the community as “Medical Park,” but Google Maps had done so since the late 2000s. Community members argued the designation was a calculated tweak in favor of gentrification, a digital rechristening that would be used to sell houses, market Airbnbs, and wrest the neighborhood’s future from the people who had made a home there for generations.
Lott didn’t know it at the time, but the misnomer also revealed a great deal about the invisible process major tech firms use to put neighborhoods on their maps — and how decisions based off arcane data sets can affect communities thousands of miles away.
Who erased the Fruit Belt? Lott demanded of officials. What the hell is Medical Park? And how did it get on this map?
“There is nobody in City Hall who would dare utter the phrase Medical Park anywhere near the Fruit Belt,” Brendan Mehaffy, the city’s chief planner, told me recently. “So I really can’t tell you how this happened.”
But unbeknownst even to city officials, the city did have a hand in the Fruit Belt’s digital erasure — as did Google, two defunct mapping startups, and an ubiquitous, secretive data broker that claims to keep tabs on 100,000 neighborhoods.
The Fruit Belt was quiet on a recent weekday, hushed under a slick of wet snow. A lone man walked in the street, his hands stuffed into his coat pockets, never checking for cars as he crossed the road.
Fewer people live in the neighborhood now than did even two or three decades ago. And signs of contraction abound: a boarded-up storefront with its shingled roof caved in; a scattershot of empty lots amidst dense houses and gardens. Meanwhile, just west of the Fruit Belt and across Michigan Street, every year the Medical Campus adds new labs, offices, incubators, and classrooms.
“I walk through my community and see amazing prosperity on its edges,” says Dennice Barr, a longtime resident and activist, sitting at a local coffee shop. “None of that is for us.”
Despite its hardships, the Fruit Belt is still a living community, a place where neighbors call out good morning and grandparents survey the kids from front stoops. Some of the homes here date back to the 1840s, when German Lutherans founded this 36-block neighborhood and planted roots on streets with names like Orange and Lemon, giving the neighborhood its name. Today, the houses are interspersed with playgrounds, a charter school, five Baptist churches, and the incongruous, scaled-down suburban homes the city built in the 1980s and 1990s to backfill demolitions. Barr’s own family moved in more than 70 years ago, part of a wave of black families lured from southern states by jobs at Bethlehem Steel or bulldozed out of Buffalo’s other East Side neighborhoods.
By 1970, the Fruit Belt was almost entirely black, the Germans bought out by black bus drivers, machinists, and, on Mulberry Street, a single, retired Negro League baseball player. To residents, the Fruit Belt is a point of pride — a legacy they fought to gain, and then to hold onto.
“This is the most stable African-American community in Buffalo,” says Michael Chapman, the head pastor at the Fruit Belt’s St. John Baptist Church. “This is the community that has been under African-American control longest.”
But in the 1950s and 60s, the Fruit Belt ceded several blocks of residential land to the expansion of Buffalo General Hospital and the Roswell Park Cancer Institute. (Housing conditions made the Fruit Belt “eligible for clearance,” city planners wrote in 1964.) Throughout the 1980s and 1990s, development officials dreamed of reviving Buffalo’s downtown by expanding the two hospitals. Their vision was realized in 2001 when a clique of politicians, planners, and health care executives began plotting a research corridor called the Buffalo Niagara Medical Campus and anchored, in part, by the University at Buffalo.
Even in those early days, the campus unsettled Fruit Belt residents who thought it might undermine the neighborhood. Their concern was well-founded. Buffalo’s white mayor at the time, Anthony Masiello, crowed that the development would “change the face, economically and residentially, of several neighborhoods.”
Sure enough, developers tore down low-income townhouses and bought out a nearby African-American cultural center to make way for new apartments and medical offices. Worse, some residents reported rent hikes of as much as 50 percent, and one in three homes sat vacant and unused — a common sign that owners planned to resell them at a higher price.
City power brokers cheered the trend: “It is overwhelmingly good news that property values are rising around the burgeoning Buffalo Niagara Medical Campus,” an editorial in the Buffalo News read. “The Fruit Belt area is becoming interesting,” it continued, to new investors and residents.
But to people who grew up around the Fruit Belt, like Lott and Barr, the news was mixed. Their families had fought to buy these houses and maintain their neighborhood over decades of neglect by the city. Now it seemed they might be usurped by the Fruit Belt’s changing fortunes. Henry Taylor, a professor of urban planning at the University at Buffalo who developed a strategic plan for the renewal of the Fruit Belt in the early 2000s, found himself incited to regular “shouting matches” with his colleagues who consulted on the Medical Campus.
“What we’ve seen in the Fruit Belt is what I call type 2 gentrification,” Taylor says. “Type 1 is where whites move in and blacks and other groups are forced out. But there’s also type 2, where low-income blacks are pushed out through a process of land speculation and institutional development.”
By 2008, the Medical Campus loomed — quite literally — over the Fruit Belt, silhouetting its low, gable front houses against cliffs of cement and fritted glass. Neighborhood organizers met quarterly with campus officials to swap proposals and grievances.
It was one of those organizers, Veronica Hemphill-Nichols, who first noticed the naming error. In 2008, Hemphill-Nichols says she Googled directions to her house on her first, recently-purchased smartphone. When the map loaded, she saw it spelled out: “Medical Park.”
“I flipped my wig,” she says now.
A gang interventionist by profession and the mother of four sons, Hemphill-Nichols had lived in the Fruit Belt since 2001, renting a four-bedroom house on Grape Street after the West Side neighborhood got too expensive. And she was determined to prevent that sort of development from displacing people in her new neighborhood — so determined that, six years later in 2014, she’d launch an unsuccessful New York state assembly bid on that platform.
Hemphill-Nichols “blasted” the city planning office, first by email and then at a meeting at Centennial Missionary Baptist Church, asking them to explain what happened. She then took the maps to Medical Campus officials, who said they had nothing to do with it.
As she stressed to officials at the time, the existence of the Fruit Belt wasn’t open for debate. The neighborhood appeared on city maps dating back to the 1950s, bordered by Michigan Avenue, Jefferson Avenue, Cherry and North streets. In 2005, when the University at Buffalo commissioned a neighborhood map based on more than a century of historical records and city plans, it clearly placed the Fruit Belt on a neat scalene wedge between the Kensington Expressway and the Medical Campus.
“Maps don’t just show the world — they change the world,” says the geographer Mark Graham. “They affect how we interact with the world and understand the world. In doing so, they shape the world itself.”
Residents couldn’t prove it, exactly, but they believed the Google Maps error was both a symptom and cause of their displacement. “They took our name from us and no one knew about it,” Hemphill-Nichols says. “Once you take our identity, you plan to take everything else.”
Matt Enstice, the Medical Campus’ chief executive officer, says the facility never requested any sort of name change or geographic designation. Mehaffy, the head of the Office of Strategic Planning, says the city had also made no attempts to “rebrand” the Fruit Belt either, at least since 2010 when he started his tenure.
Nevertheless, residents say references to “Medical Park” started cropping up all over the place. There was the Zillow listing for a vacant lot in “Medical Park” asking for far more money than entire house typically fetches in the neighborhood. There was a “Medical Park” rental aimed at med students and doctors on Airbnb. Then, there was the flood of new students and workers parking and smoking in the Fruit Belt. Residents wondered if the newcomers saw “Medical Park” when they mapped their commute and just made assumptions.
With so little public understanding of how these systems work, it’s tempting to assign intent to these myriad decisions. The process is too opaque to scrutinize in public. And that ambiguity foments a sense of powerlessness.
Worst of all, in May 2018, a young realtor named Kim Santana posted a glamorous Instagram post of herself standing on the corner of High and Lemon Streets, holding a large coffee cup. “Whether you call it the fruit belt or the medical park neighbourhood, this is a great opportunity to be part of the buffalo renaissance!” her caption read. (“I really believe that the Medical Campus is going to bring some attention to the formerly known Fruit Belt neighborhood,” Santana told me by email.)
Residents who saw the image seethed. India Walton, the co-founder of the Fruit Belt Community Land Trust, ticks off the many ways the Instagram post offended her: the tone, the use of the reviled Medical Park name, the presence of a “chai latte or whatever.” Walton includes the image in a presentation she gives to other activists around the country about the dangers of gentrification. The post proves, Walton says, that developers and realtors seized on Google’s mistake to market the Fruit Belt to rich white kids.
Both City Hall and the Medical Campus blamed Google for the name change, repeatedly telling residents at community meetings and town halls that they had no control over what the technology giant called their neighborhood. But no one ever contacted Google in any official capacity either, says one longtime city planner. For their part, many Fruit Belt residents — one-fourth of whom are senior citizens, and two-fifths of whom lack internet access — had no idea how to contact Google themselves. What phone number would they call? Where could they send a letter?
“I think, today, when people are having conversations around their kitchen tables, talking about gentrification and how the neighborhood is changing, Google comes up then,” Walton says. “But I don’t think people have energy to do much more than that. We can’t afford to fight a name when we’re fighting for our existence.”
It’s just as well no one contacted Google, because Google wasn’t the company that renamed the Fruit Belt to Medical Park. When residents investigated, they found the misnomer repeated on several major apps and websites including HERE, Bing, Uber, Zillow, Grubhub, TripAdvisor, and Redfin.
Under normal circumstances, you’d expect such repetition: maps from different sources often include similar features. But the fact that the error was so random, and yet so widespread, suggested a single common source. Monica Stephens, a geographer at the University at Buffalo who studies digital maps and misinformation, immediately suspected the geographic clearinghouse Pitney Bowes.
Founded in 1920 as a maker of postage meters — the machines that stamp mail with proof it’s been sent — Pitney Bowes expanded into neighborhood data in 2016 when it bought the leading U.S. provider, Maponics. In its 15-year run, Maponics had supplied neighborhood data to companies from Airbnb to Twitter to the Houston Chronicle. And it had also just acquired a longtime competitor, Urban Mapping, which has previously supplied Facebook, Microsoft, MapQuest, Yahoo, and Apple.
Though Pitney Bowes is far from a household name, the $3.4 billion data broker is “a huge company at this point,” says Stephens, with enough influence to inadvertently rename a neighborhood across hundreds of sites.
Data brokers responded by coding hierarchies into their data, so that some neighborhoods would, say, disappear into others, or surface only after several zooms.
Pitney Bowes claims to catalog records for 100,000 neighborhoods located in 925 cities and 83 countries around the world. For its clients, the sell is simple: neighborhood data makes it easier for users to find places, and even large tech companies would rather skip the headache of compiling and vetting them personally.
Reached for comment, Google declined to say whether it uses data from Pitney Bowes or its predecessors, though it has sourced data from both Urban Mapping and Maponics in the past, according to a former Google Maps product manager and a 2011 Maponics press release. In an emailed statement, a Google spokesperson said that the site sources its neighborhood names from “authoritative third-party providers and public sources.” But four other companies using the telltale “Medical Park” label confirmed that they had sourced their data from Pitney Bowes-owned databases.
“It’s not rocket science, but it’s hard and it requires a lot manual curation,” says Ian White, Urban Mapping’s founder and former chief executive officer. “There’s no machine that can do this for you… These are representations of unofficial, social spaces.” Considering the challenges, White says, neighborhood researchers have done a great job.
Some of the data in today’s databases are 15 years or older. In the early 2000s, Urban Mapping offered new college grads $15 to $25 per hour to comb local blogs, home listings, city plans, and brochures for possible neighborhood names and locations. Maponics, meanwhile, used nascent technologies such as computer vision and natural language processing to pull neighborhoods from images and blocks of text, one former executive with the company said.
Such methods may have risked passing over neighborhoods without published documentation or significant online footprints. But the real problems arose later, White says, when the data made its way to mapping applications and other clients.
Few of these programs had mechanisms in place to convey ambiguities, such as neighborhood overlap or fuzzy borders. Instead, they often blindly assigned neighborhoods to geofenced blocks. (Wrote one group of geographers in a manifesto on digital maps: “Coders have little time for catching up on the history of mapping or reflecting on the implications of certain design decisions.”)
On top of that, it turned out few clients truly wanted a comprehensive list of neighborhoods — too much detail could derail navigation on real estate sites, and clutter maps. Data brokers responded by coding hierarchies into their data, so that some neighborhoods would, say, disappear into others, or surface only after several zooms.
Today, Pitney Bowes markets its “robust hierarchy” as a selling point, because it makes the data usable in a range of applications. Every neighborhood in its database comes marked as “macro,” “regular,” or “sub,” the smallest classification. It’s the sort of invisible, technical feature that only comes to public light when it erases a place like the Fruit Belt.
“In this instance, we reviewed our data and Medical Park is listed as the ‘Regular’ neighborhood level,” confirmed Emily Simmons, a spokesperson for Pitney Bowes. Meanwhile, the company coded the Fruit Belt as a smaller “sub-neighborhood.” Simmons says clients, “depending on their own technology, may choose to only display Macro or Regular level data.” When it came to maps built off this data, Medical Park had literally swallowed the Fruit Belt whole.
Back in Buffalo, that explanation doesn’t settle the matter. It may explain why the Fruit Belt appears on so few maps — but it does not explain Medical Park’s origins.
I visited the Buffalo Central Library to find the source of the error, enlisting two special collections librarians to search old city plans for any references to Medical Park. They pulled down banded folders and heavy scrapbooks compiled by New Deal workers during the Great Depression. Maybe the term actually predated the Fruit Belt, or appears in some newer, more obscure city document. Simmons said that some of Pitney Bowes sources for Buffalo data included a long-inactive planning group and the city itself.
Sure enough, one of the librarians located a single planning office map that used the “Medical Park” label. It was a 1999 report on poverty and housing conditions — long since relegated to a dusty shelf stacked with old binders and file folders. Even longtime city planners barely recall this map: Its purpose, they say, was to display census statistics, and its 54 “neighborhoods” correspond to census tracts. Somehow, likely in the early 2000s, this map made its way into what is now the Pitney Bowes data set — and from there, was hoovered into Google Maps and out onto the wider internet. Buffalo published another map in 2017, with the Fruit Belt clearly marked, and broadcast on the city’s open data portal. For whatever reason, Pitney Bowes and its customers never picked that map up.
With so little public understanding of how these systems work, it’s tempting to assign intent — malicious or otherwise — to these myriad decisions. They could be calculated attempts to push out a community already under threat; but they could just as easily be a string of careless accidents. Bad data, pulled from an outdated map. Either way, the process is too opaque to scrutinize in public. And that ambiguity foments a sense of powerlessness.
“We’ve historically tended to self-identify our communities,” says Aaron Krolikowski, a Buffalo-based geographer and data scientist who sees risks in the ripples of centralized data. “If suddenly we become disconnected from that process, I think there’s a lot of questions that emerge about the ability of a community to determine its future, in some cases.”
Google Maps did correct the name of the Fruit Belt in February, after being contacted for this article. So too did Redfin, TripAdvisor, Zillow, and Grubhub. In an emailed statement, a spokesperson for Google defended its use of third-party neighborhood sources.
“Overall, this provides a comprehensive and up-to-date map,” the spokesperson said, “but when we’re made aware of errors, we work quickly to fix them.”
There is little indication that either Google or Pitney Bowes will make this process easier or more transparent in the future; while Google says it invites user feedback on its maps, it doesn’t always act on those suggestions. And even that process, which involves clicking through three levels of menus, precludes users with limited computer skills. (Says the geographer Matthew Zook: “Communities that are not well represented in the online world tend not to show up as much” on digital maps.) Pitney Bowes offers no method for users to submit corrections.
This victory is partial for the Fruit Belt, too. Residents are glad, of course, to get their name back. But some wonder how long mistaken impressions may linger — and everyone acknowledges the name change won’t address the Fruit Belt’s underlying problems.
Rents continue to rise and developers are circling the neighborhood, looking for investments. Next year, advocates say, some Fruit Belt homeowners will get slammed by property tax reassessments as a result of rising property values, spurred by the Medical Campus.
But as residents continue to fight, says Lott, they’ll have the basic certainty their community is seen — by Google, at the very least.
“The thing with Google, it dampened people’s spirits who have been there struggling all their lives,” Lott says. “But it’s good, it’s good now. It’s a neighborhood where people are really close and concerned about each other. And you don’t see that, probably not anywhere.”
Today, if you tap “Fruit Belt” into Google, it brings up a pin near High and Peach Streets, six blocks from the house where Lott grew up. Coming from City Hall, the app dictates a left, then a right, then a left on Michigan.
Drive under the overpass that first slashed Fruit Belt home values in the 1950s. Past St. John’s Baptist Church, on the right, and the mirrored glass and terra-cotta science center on the left. Right on High Street, past the windowless deli and the overpriced lot and the intersection that realtor Instagrammed. Past wide flats of asphalt, neat, narrow houses, and an old blade sign for New Zion Baptist, rusted almost beyond recognition.
And when you get there Google will say, after all these years: “You have arrived at your destination.”