Can an Alphabet spinoff use phone location data to transform urban planning?

Replica's technology is so new, urban planners don't know what to do with it. To understand its biggest selling point to communities across the country, consider the laborious transportation studies Kansas City has done for the last 50 years or so.

About once a decade, planners ask residents to fill out a travel diary in a low-tech effort to assess how people move about the metro area. Participants jot down details about the mundane trips they take in a given week — commuting to work, dropping the kids off at school, and that jaunt to the grocery store that's never quite as short as planned — helping decision-makers identify choke points and which public infrastructure is most in need of attention. The last survey was completed in 2019. It covered nearly 4,000 households in the metro area and cost taxpayers $800,000.

Replica thinks it can do better with the help of millions of smartphone location data pings.

In a recent pilot project, Kansas City transit planners tried the high-tech solution from Replica, a spinoff from Alphabet's Sidewalk Labs. It's easy to see the appeal. For a fraction of the cost of a traditional study — about $230,000 for a year's worth of data — planners got an expansive computer-generated model of how people move through their community, based on a combination of cell phone location data and census information. Unlike surveys that are based on years-old data by the time they are put to use, Replica's models can be updated quarterly.

"We're very excited for the potential about the types of questions we can answer with this data that we couldn't even two years ago," said Ron Achelpohl, the official who oversaw Replica's pilot in Kansas City.

Because of this potential, Replica has also sold its technology to transit authorities in Chicago, Portland and Sacramento, for about $5 million total, records show. The company says it expects to double its footprint this year.

Still, Replica's future is uncertain. In interviews, planners say that Replica's data is powerful but can be overwhelming. Replica needs to convince them it can provide actionable information. Meanwhile, the source of the models — smartphone location data — is a significant concern for officials wary of privacy controversies. The success of Replica's business hinges on delivering insights planners can use while addressing the privacy concerns of their communities.

More data, more questions

The Kansas City pilot culminated in a 54-page report in late December, which found that Replica was better suited to fine-tune models generated by traditional survey data, rather than be the basis for the model itself.

The document also addresses a key aspect of Replica's expansion into other metro areas: whether transit agencies will have the use cases — and the capacity to execute them — to justify buying the data. In Kansas City, Achelpohl said his team didn't have enough staff "to take advantage of all of [Replica's] capabilities."

"It's such a rich menu of data we haven't had before," Achelpohl said, citing the value of more accurate vehicle traffic simulations for non-highway driving, which historically has been left out of survey research, as well as better estimates of pedestrian traffic, which he said are almost nonexistent with traditional planning tools, as two examples.

Replica's models are more comprehensive than survey data, providing planners with significantly more detailed information on shorter trips, such as a walk or specific bus ride route, that a resident might forget to include on their survey. Replica data is also updated in near-real time, a boon for metro areas where changes in workplace and residential density often outpace traditional planning tools.

Despite the possibilities, Achelpohl said he wasn't sure what the next step would be in discussing a long-term contract with Replica. "We're going to need to take some time to figure out how to best use these tools," he said.

Replica CEO Nick Bowden said the Kansas City pilot provided the company with valuable feedback to apply to models in other cities. He expects to tout new customers as soon as this spring.

Early projects in Kansas City, Portland, Sacramento and Chicago have shown Bowden that inking lucrative deals with large would-be customers can be tricky, however. "There's enough nuance and intricacy, that it certainly takes longer than we anticipated originally," Bowden said. "There's probably not a single source of data that will fulfill every need they have. The question now is, 'What is the right place for Replica?'"

Replica's dashboard allows urban planners to see where commutes begin.Image: Obtained via a public records request

Other learning experiences from Replica's pilot in Kansas City include how company employees represent their relationships with existing customers when trying to woo new ones. Bowden fired a former head of business development after he solicited a meeting with a neighboring city by name-dropping Achelpohl's planning agency, the Mid-America Regional Council, emails show. The exchange appeared to make the city manager in nearby Lee's Summit uncomfortable, prompting a swift response from Bowden.

"This isn't ever going to be acceptable or tolerable to be misleading at Replica," Bowden wrote to Achelpohl. "I'm sorry you were caught in the middle of this situation and our relationship with you all was misrepresented."

In an interview, Bowden said he stressed to his staff to be mindful of the incident when pitching potential new customers. "How we represent ourselves, and think about what we do and communicate with customers, is the most important thing," he said.

The privacy problem

As Replica learns to work with public agencies, the agencies, in turn, are learning how to work with the company's sensitive data at a time when people are increasingly warned about their privacy. Replica doesn't disclose specific vendors it purchases cell phone location data from and, so far, cities haven't demanded to know. Of its customers to date, public officials in Portland came the closest, emails show.

"We would like to see more thorough documentation of the Replica data sources, methodology and dataset prior to signing," Portland technology strategist Eliot Rose wrote in an email to Bowden in December 2018. "This pilot has been receiving increased scrutiny from the press and our residents."

In a response, Bowden shared Replica's limited data disclosure, saying, "NDAs prevent us from sharing specific vendor names, but this provides a great deal of detail about the general sources and process. Most importantly, Replica does not handle, process, or store personally identifiable information at any point."

The Oregon Metro agency did not release the data disclosure in response to a public records request from Protocol. In emails released by the agency, Bowden wrote that Replica considers this information "confidential and proprietary but [we] want to make sure you all have a complete understanding of how Replica is created."

The explanation ultimately proved sufficient for Portland to move forward with the pilot. Regional planners there have worked in recent weeks to begin validating preliminary Replica models. Portland officials last completed a comprehensive traditional travel survey in 2011, to the tune of $1 million. The ongoing pilot with Replica cost about $450,000.

It's a sympathetic use for big data. At the same time, a lot of this data shouldn't exist. — ACLU's Jay Stanley

The privacy concerns in Portland underscore growing tension in cities across the country, as government officials and regulators alike reckon with scrutiny of the industry built on a vast collection of smartphone location data. At the same time, numerous "smart city" efforts are underway, creating an appetite for more data.

"It's a sympathetic use for big data," said Jay Stanley, a senior policy analyst at the American Civil Liberties Union. "At the same time, a lot of this data shouldn't exist."

Part of the concern for Stanley and other privacy advocates is that the specific combination of data Replica aggregates — where a person lives and where they work, effectively — has been shown to be especially vulnerable to re-identification.

Bowden conceded it's theoretically possible for someone, with the right tools and access to Replica's internal data, to de-anonymize the information, but he said Replica takes a variety of steps to mitigate opportunities to misuse it. This includes stripping personally identifiable information from the data it purchases from third-party vendors, such as telcos and companies that bundle smartphone app data, as well as requiring vendors to hash device IDs.

"We're not naive — there's a risk to everything," Bowden said. "But we effectively remove a lot of that risk."

Understanding this risk can be difficult for urban planners, particularly in metro areas with limited staff. The American Planning Association frequently fields questions from local government officials on how to best use new data tools, said research program manager David Morley. While the association doesn't provide prescriptive guidelines to planners, Morley said the general advice is to "read the fine print" on any vendor contract.

Morley said that urban planners can sometimes be "caught in the middle" of a desire to utilize new technology to better assess the needs of a community, while at the same time steering clear of privacy concerns.

"A lot of these applications using these large datasets, the way in which they are making sense of that data can be really opaque to the average decision maker," Morley said. "Are local decision-makers able to keep up with this rapid advancement of tech and the implications of that for safeguarding the public interest?"

Uncertain future

While officials in Kansas City turn their focus to other projects as they decide whether to continue their relationship with Replica, other cities are reaching a critical juncture working with the data.

In Chicago, officials from an array of agencies were recently given access to Replica data. The partnership — so far Replica's largest with a statewide $3.6 million contract — will soon extend to Champaign-Urbana.

Officials in Sacramento, meanwhile, say their pilot project with Replica, which cost $600,000 and extends through 2021, has shown early promise. Planners began using the data in December and are currently validating it against real-world measurements and traffic counts.

Replica's software layers census information with cell phone location data.Image: Courtesy of Replica

As government officials in these cities work to maximize the value of Replica's models, the company's investors acknowledge that it's still early on in Replica's development.

"A lot of this is us spending less time talking and more time listening," said Scott Brady, who led an investment by Innovation Endeavors in September, when Replica was spun out of Sidewalk Labs with a fresh infusion of capital. Brady said Replica must "refine and perfect" its tools as it hires additional staff this year, and that tailoring the data to suit the needs of customers is the key to further growth.

"We're on the cusp of something that's so new. We're still learning the questions we can ask of this data," Brady said, adding that, "As a taxpayer, these are the types of tools that I hope cities are using."