Downtown Brooklyn as a Living Lab for AI-Driven Retail Planning

In Downtown Brooklyn, decisions about where to open a coffee shop, attract a retailer, or fill a vacant storefront could soon be informed by an unlikely source: artificial intelligence trained on how people actually move through the neighborhood.

A new research project from the Resilient Urban Networks Lab, led by Takahiro Yabe — an assistant professor at the Center for Urban Science and Progress (CUSP) and the Department of Technology Management and Innovation at NYU Tandon School of Engineering — is using anonymized mobility data and advanced AI models to better understand how people choose where to spend time and how those choices shape the local economy.

Developed in collaboration with the Downtown Brooklyn Partnership (DBP), the work aims to provide a practical tool for retail planning and economic development.

At the core of the project is a system that models thousands of individuals as “AI agents,” each representing a type of person who lives, works, or visits Downtown Brooklyn. Using mobile phone location data, demographic information, and business attributes, researchers created a “synthetic population” of roughly 20,000 agents that simulate real-world behavior.

“We’re essentially building something like SimCity, but grounded in real human behavior,” said Yabe.

These agents can be used to test “what-if” scenarios, such as how foot traffic might change if a new retailer opens, a grocery store is introduced, or a popular café closes. Earlier research could estimate the economic spillover effects of business closures, but this approach goes further by simulating how entirely new additions, such as parks, retail stores, or entertainment venues, might reshape activity across the neighborhood.

“For retail operators, real estate developers, and property owners, one of the biggest challenges is uncertainty,” said Yabe. “This framework allows us to test scenarios before making costly decisions, whether that’s introducing a new tenant, redesigning a space, or rethinking an entire retail mix. The goal is to turn data into a practical decision-making tool that reduces risk and improves outcomes for neighborhoods.”

For organizations like the DBP, which helped shape the research questions and provided local context, the goal is not just to understand behavior but to support decisions about retail strategy. This includes identifying what types of businesses should fill vacant storefronts to attract visitors and encourage them to stay longer in the area.

“They help us frame the right questions,” Yabe said, noting that the Partnership has been closely involved in identifying real-world use cases and providing information about local business openings, closures, and potential tenants.

"Understanding where people go and why is the foundation of a healthy retail market," said Mark Landolina, Senior Director of Real Estate and Economic Development at DBP. "This research gives us the ability to simulate how foot traffic and consumer demand shift when a new business opens or closes before it happens. That means we can go to landlords and tenants with concrete, data-backed recommendations about what types of businesses belong where, taking the guesswork out of leasing decisions and helping the market respond to real demand. That's how you fill the right storefronts with the right businesses and build a stronger Downtown Brooklyn."

Early findings highlight how interconnected the neighborhood’s economy is. In one simulation, when a popular coffee shop was removed, customer demand did not shift to a single competitor. Instead, it spread across many nearby businesses, with most alternatives located within a short walking distance.

This pattern suggests that Downtown Brooklyn operates as a tightly linked retail ecosystem, where businesses share and redistribute foot traffic rather than compete in isolation. It also points to the importance of proximity and diversity in sustaining neighborhood vitality.

The research also demonstrates that everyday movement patterns, particularly predictable routines like weekday lunch trips, can be modeled with a high degree of accuracy. The model has been tested against real-world business openings and closures, and researchers say the level of predictive accuracy improves significantly when combining behavioral personas with spatial and business data.

“The level of accuracy we’re seeing is really, really high, especially for out-of-sample predictions compared to traditional machine learning approaches,” Yabe said.

While the current analysis focuses on restaurant visits as a starting point, the broader framework can be applied to a wide range of retail and urban planning questions, from tenant mix to long-term neighborhood development.

The project is part of NYU CUSP’s capstone program, a two-semester initiative in which students work with external partners to address real-world urban challenges. It was presented on May 1 at the Urban Data Science Showcase at NYU’s Brooklyn campus. The project, titled Enhancing Downtown Brooklyn’s Retail Market Through Data-Driven Interventions, was led by CUSP graduate students Sizhe (Alex) Xu and Divya Natekar, with Ph.D. students DongHak Lee and Boyang Li (CUSP, TMI) as student mentors, Yabe as the faculty mentor, and the DBP serving as the project sponsor.

About DBP Living Lab:

Downtown Brooklyn is a place where collaboration and innovation come together to solve real problems. Through our Living Lab program, DBP partners with groups to solve quality of life challenges facing cities, using Downtown Brooklyn as a platform to test technologies and generate real-world insights. This initiative is a perfect example of what the Living Lab is all about, bringing NYU Tandon's cutting-edge research and technology to work right in our own backyard.