The Age of Data in Real Estate
Sarah Popelka is a Research Assistant at Urban AI and is currently pursuing a Master’s in Urban Governance from SciencesPo. In this article, she analyzes the insights of our Urban AI Conversation #12 with Katya Letunovsky: “The Age of Data in Real Estate”
Contemporary cities are often conceptualized and enacted as a grid — blocks of built areas, separated by roads and other linear flows, distributed in two-dimensional space. Yet, beyond this planform representation, a number of factors contribute to shaping and defining the built environment. Katya Letunovsky, in conversation with Urban AI, describes how her team at Habidatum, enabled by artificial intelligence and a unique data platform, aims to capture the underlying real estate market dynamics that serve to iteratively reshape the city through space and time.
The data scientists at Habidatum maintain the notion of the city as a grid, but in their conceptualization, the grid exists in three dimensions, rather than two: data on visible and invisible urban characteristics are layered to give a deeper understanding of the city at a given instant, with each of these momentary slices stacked vertically in time. Habidatum has termed this data structure, which simultaneously represents temporal and spatial dynamics, the “Chronotope Grid.” Every “voxel” in this Chronotype Grid (i.e. each unit of data representing a discrete location at a discrete point in time) conveys information about commercial viability, occupant and user mobility, and transportation network accessibility. Achieving this complex data model requires the use of sophisticated algorithmic techniques to merge and assure spatial and temporal parity between thematically disparate datasets: cell phone location data, financial consumption data, as well as data about the built environment. All these data are aggregated and anonymized.
According to Letunovsky, the unique, data-laden model of urban space that Habidatum provides allows urban stakeholders to unlock new insights about the cities in which they operate. Rather than utilizing a simple high traffic/ low traffic binary to assess market potential, Letunovsky and her team calculate a multivariate “location risk score,” based on metrics calculated for a context-specific service area around the target property. If the property supported localized commercial uses, for example, the service area might be calculated as all locations within a 10–30 minute walk from the business. On the other hand, if the property supported more broad-scale industrial uses, the service area might be calculated as all locations within a 10–30 minute drive of the business. Habidatum adds value to the traditional service area calculation by using cell phone location data to validate the network-derived, potential service area, highlighting the true locations, from which people travel to reach the target property. The validated service area is used to calculate the commercial density (the number of commercial buildings) and commercial diversity (the distribution of commercial use types) within an accessible distance of the target property, providing a risk (or potential) score relative to the risk associated with comparable buildings within the service area. The unique data analytic capabilities unlocked by the Chronotope Grid data model provide a deeper understanding of market dynamics, especially in the absence of available data.
Letunovsky gives a number of examples highlighting the utility of her team’s approach. In instances where occupancy data are incomplete or unavailable, Habidatum analyzes temporal variability in the volume of cell phone location data at a target property to estimate the volatility of the building’s tenure. Using a similar analytical approach, Habidatum can flag empty buildings soon after vacancy. These use cases, Letunovsky says, particularly interest banks and insurers, who use neighborhood occupancy characteristics to run risk analysis and improve risk modeling accuracy. This vein of location-based building occupancy analysis also proved particularly useful in estimating and characterizing the return to normal economic activity in various neighborhoods, following COVID-19 pandemic-related shutdowns and restrictions.
Although Habidatum’s analyses do not venture into the realm of future forecasting, the team has developed interpolation and imputation algorithms, which both provide a more complete picture of the present potential of a property and allow for scenario modeling, even in the case of a new or proposed development that might not have any associated data. With these tools, urban planners and property developers can experiment with the effects of introducing new properties or altering the balance of use types within a neighborhood. In doing so, they have the opportunity to equip themselves with a more data-driven view of how their planned interventions might shape broader market dynamics and, in turn, a target property’s own potential, prior to moving forward with any changes. In this way, the conceptualization of the city that Letunovsky and her team champion- made possible by the unique data model that they have developed and enabled by artificial intelligence- has the potential to provide a novel view of land use, building occupancy, and real estate market dynamics, equipping urban stakeholders with the means of taking a more proactive, data-enabled planning and decision-making approach.
By Sarah Popelka
Urban AI Conversations are webinars during which members of the Urban AI Community present their latest research, solutions, or ideas.