Cities use long-ago data to plan far-away futures. Let’s fix that
A Sidewalk Talk Q&A with Replica CEO Nick Bowden.
This Sidewalk Talk Q&A is part of a series of conversations with leaders at Sidewalk Labs incubated and portfolio companies.
Replica is a Sidewalk Labs spinout company that makes complex, rapidly changing urban environments easier to understand through the power of data — data that can inform city planning in a far more responsive way than it currently does.
“If you went back and looked at any city-wide or region-wide comprehensive plan from the year 2000, which oftentimes were aimed at a 2030 or 2040 kind of horizon, you wouldn’t find any mention of Uber, or Lyft, or ride-sharing, or Amazon, or Prime, or any number of things,” says Replica CEO Nick Bowden. “Just that small list of five things has fundamentally changed how cities operate. So, there’s this obvious, like: ‘Wait a second! We don’t have any flexibility or adaptability to adjust to the rapidly changing innovation that’s happened in the built environment.’ ”
Replica works with public-sector agencies across the U.S. to provide just that flexibility for long-term planners. Bowden spoke to Sidewalk Talk editors Eric Jaffe and Vanessa Quirk about Covid’s dramatic impact on urban systems, how Replica embeds privacy into its approach, and the ways the company collaborates with governments to help them model a greener, more liveable future.
Watch a video of our conversation above or read an edited transcript below.
Eric Jaffe: For those unfamiliar with Replica, can you describe what you do and your mission?
Nick Bowden: We would describe ourselves as a data platform for the built environment, and we use that terminology pretty specifically. We provide insight about mobility (how and where and why people are moving) about economic activity (how, where, and why people are spending money), about land use (how is land use changing, how is the land being used in a particular city). And then we have additional metrics around things like public health. With Covid going on, how does that affect things? I would describe a lot of our mission as: how do we provide data that gives insight into the system that is a city, or the system that is a region, versus just a very specific set of data around mobility or a very specific set of data around economic activity.
Vanessa Quirk: Can you describe how public agencies have traditionally drawn their insights around long-term planning, what the limitations of those traditional approaches are, and how Replica advances those efforts?
I started my career in the public sector, and I saw this a bit firsthand. The business model, if you want to use that terminology, historically, has been to use what we would describe as “long ago data” to forecast the far away future. A lot of planning processes look at the last Census — which could be three, four, five, 10 years old — and then try to project out 25 years or 30 years into the future. I don’t know if there’s anything fundamentally wrong with that kind of approach, but the obvious gap is that it doesn’t have a lot about what’s happening today.
The specific example that I always give is, if you went back and looked at any city-wide or region-wide comprehensive plan from the year 2000, which oftentimes were aimed at a 2030 or 2040 kind of horizon, you wouldn’t find any mention of Uber, or Lyft, or ride sharing, or Amazon, or Prime, or any number of things. Just that small list of five things has fundamentally changed how cities operate. So, there’s this obvious, like: “Wait a second, we don’t have any flexibility or adaptability to adjust to the kind of rapidly changing innovation that’s happened in the built environment.”
That’s a bit of a description of how it works today. The world is probably changing faster than it ever has, and cities structurally have not historically been in a place where they can adapt quickly to those changes. Covid is the most extreme example of this. Nearly overnight transit ridership plummets, call it 80 percent. There just aren’t the workflows or the systems in place to adjust to that.
Some of what we’re trying to do from a company, or product, perspective, is actually shrink those time horizons in, and say, “What about if we looked at last quarter, or last month, or last week?” And instead of thinking about it, 25 or 30 or 40 years in the future, we think about it in one-year, or two-year, or five-year increments, to allow for adaptability.
Vanessa Quirk: It’s important to emphasize that you’re analyzing population data, not personal data. Can you talk more about how you think about protecting privacy in the work that you do?
It’s a great question. I’d say privacy is at the cornerstone of everything we do. To give a little bit of context, I’d say that, historically, there has been a tradeoff for the public sector, which is: either do we want high-fidelity data that comes with significant privacy risk, or do we choose lower fidelity data that doesn’t have privacy risk?
That feels like a false tradeoff in the modern world of data and computing. Neither we nor our public-agency customers are ever interested in the movement of one person. You don’t develop a policy framework around Joe or Mary. You develop a policy framework around how the system operates, and it goes back to this systems approach. There’s very little incentive and/or reason to have individual data. It just doesn’t make a whole lot of sense.
A lot of our approach has been designed to take and understand a large composite of input data sources, train a set of behavior models that mimic the things that real people do, and then to represent that in a model, the output of which doesn’t actually sacrifice fidelity for privacy. And the use cases are always about groups of people. So, how is a group of people using this particular transit line? How is a group of people using this particular part of downtown? How is a group of people spending money?
I’m glad that you asked. It’s a really important part of our approach. I think it just hasn’t been possible, historically, from a technical perspective.
Eric Jaffe: Let’s get into one of those use cases. I know you’re working in California, which recently became the first state to adopt a new measure, called vehicle miles traveled, for how it measures the impact of new development on the transportation system. Can you talk about that work and what goes into working directly with government?
California has a long history of being one of the more progressive states pushing the boundaries on the key metrics for policy frameworks, regulatory frameworks, and even funding frameworks. They have recently switched from a “level of service” measurement — effectively, “level of service” is a measure of congestion — to a framework around vehicle miles traveled [VMT] — which is, how many miles does a person over the course of a total day accumulate? So, when you go from home to work, or you go from work to school, or you go to lunch, or you go home, that total measurement. There’s been a lot of work done to connect VMT to emissions and climate change.
We’ve been working really closely with several parts of the governor’s office in California, Caltrans, and several of the largest regional planning agencies in the state to provide high-fidelity VMT measurements. For the last couple of years, as California prepared for this regulatory change, they’ve only been able to look at VMT at a census track-level of fidelity.
Oftentimes, these policy frameworks require you to distinguish between residential VMTs, the residents and how many miles they drive, versus call it commercial VMT, which is going up with things like Amazon delivering to houses, versus worker VMT, which is created by people going to their workplace. Because of some of our technical processes, not only can you distinguish between those three different kinds of VMT types — residential, commercial, and work-driven — but you can look at it at the network level.
The state of California has something called state bill 743, it’s a big regulatory framework. A specific example is if you’re a real estate developer and you want to do a new project. So let’s say you’re taking a piece of land that currently has a couple of single-family homes, and you’re converting it to a 10-story apartment complex. You’re not asked to measure the level of impact on the streets, like, does it create congestion or not. What you’re actually asked to measure is: what kind of new VMT will be generated from the residents that live there, and how will that distribute across the network? It’s been a really fun piece of work for us because it’s not only pushing us to have better metrics, but it’s also giving them the level of insight that they just haven’t had before.
Vanessa Quirk: You mentioned worker VMT. I would imagine that, during the pandemic, some of those numbers decreased. Thinking about Covid more broadly, it really has the potential to change the way cities think about their commuting networks. Would you mind talking about some of the trends that you’ve seen across U.S. cities during Covid, and how data can help cities adjust for this new future?
This is a question I wish more cities were prominently asking, because I think that the magnitude of change that’s possible is significant. I think it could really alter urban frameworks as we know them in a lot of ways.
Specifically to your question though, one of the most interesting trends over the last year — because we process data for the whole country every day and release it every week — is that there are still a lot of U.S. cities today, in May of 2021, where travel behavior is still down 50 percent from pre-pandemic levels. In the Bay Area, I saw something this week that BART ridership is still only at like 20 percent of pre-pandemic levels. Mobility has not recovered in a way that maybe people would have expected. I think, in part, that’s because people are not yet going back to work regularly or they have permanently altered their schedules.
However, spending has actually exceeded pre-pandemic levels. People are spending more money across almost every category, even though travel has not returned. The conclusion that you could reasonably draw is that things like Amazon and Uber Eats and Door Dash have created significant trip substitution — that’s the term that you’d hear folks in the transportation industry use — where people are no longer driving to Target, but Target is coming to them. There’s nothing to suggest that that’s going to go back to where it was before.
That’s a significant change in how the network gets used, right? If everybody’s going to the place they have to go to, and then coming home, in that case I would have two trips, right? I’d have a trip to Target and a trip back home. But if Amazon is distributing goods across a whole neighborhood, it’s a radical shift of the use of the network. I think that that should be a profound, top-level question for a lot of agencies: how permanent is this? The data suggests that behavior has stabilized in a way that we should expect to see it continue in the future.
Eric Jaffe: Replica just raised another round of funding. As you look forward for Replica, what are the big challenges to making the impact that you hope to make in cities?
Company building’s hard. Building a company, building a product, it’s a hard task. It’s a fun task, but it’s a hard task. I think there’s a lot of work on that front that we have to do independently to be able to provide the impact that we want to provide.
On the product side, we have something we internally refer to as a 40-quarter plan. I think it takes 10 years to actually build something meaningful, and I think having a long-term perspective is a really important part of that. We intend to be doing this work for a long time. I think that biases us towards thinking about our partnerships with public agencies in a long-term way. We want to do a lot of work and be able to help them more and more over time.
Specifically, from a product perspective, we’ve started work on Scenario, which allows a user to go into the product and actually manipulate the conditions of today to a future state. An example of that might be like, we’ll go back to the example used earlier, this piece of land that is currently two single-family houses, and I actually want to put a 10-story, 40-unit apartment complex on it, and I want to run a simulation that shows what’s going to happen when that comes to fruition. It’s a really fun problem to solve on a technical level and on a policy level and across the board of what we do.
So, we come full circle, which is this long-ago data to forecast a far-away future. And we’re now getting to a place where we’re excited to have the tooling to be able to do near-past data into a shorter-term future. A lot of the work that we’re putting in right now is to be able to provide that offer of our product to the agencies we work with.