About and method
Last updated
What this site is
An independent screen of which empty commercial buildings could physically and financially become housing, computed building by building from data each city publishes itself. It currently covers three study areas: Downtown San Jose, Greater Downtown Denver and Center City Philadelphia.
It exists because information about vacancy was never the bottleneck. Cities have published vacancy reports for years and conversions still rarely happen, because what actually decides a conversion is the shape of the floor plate and the gap between what the work costs and what the finished building is worth. Both can be estimated from public data, and this site estimates them in the open, with the method and its failures written down.
The RESIDE Act — Section 210 of the 21st Century ROAD to Housing Act — is modeled on top of that, because it is the first federal program aimed squarely at these conversions. What the Act does and what else is in the law are covered separately.
Who publishes it
An independent project, not a government body, a consultancy or an advocacy campaign.
This site is not affiliated with HUD, the U.S. Congress, any member of Congress, or any city government. The name refers to the statute it analyses. Nothing here is an official source for the law, and no part of it has been reviewed or endorsed by anyone named in it.
It takes no money from building owners, developers, brokers, city agencies or anyone else with a stake in a particular building’s result. It carries no advertising, no sponsored placements and no affiliate links, and it sells nothing. There is no paid tier and no client work behind it, so no building’s score is anyone’s deliverable.
The engine and the site are one codebase, and the published dataset for every study area is downloadable under the same licence as the text, so the analysis can be rerun and disagreed with rather than taken on trust.
How a building gets its score
Geometry first, because geometry is what most often kills a conversion, and no subsidy fixes it.
Daylight geometry. The city’s building footprint is eroded inward by the daylight depth — the distance a habitable room can sit from a window. What survives that erosion is windowless dead core; the rest is the daylit band. Two numbers come out: the share of the floor within daylight reach, and the distance from the deepest interior point to the nearest facade. Past roughly 80 feet, a conversion needs a carved light well and effectively never pencils.
Yield, cost and gap. The daylit area, less circulation, divided by an average unit size, gives an estimated number of homes. Conversion cost is a base cost per square foot plus the code work a change of use triggers — sprinklers, egress, plumbing risers, glazing, and seismic bracing where the city requires it — applied across every converted floor, including the dead core, which must be built but can never be rented. The gap is that cost less the completed value.
The grant, then the score. A modeled RESIDE award is applied against the gap, within the statute’s per-award ceiling. The published score weighs daylight efficiency, interior depth, scale and finance, and a building that fails the depth test or yields no homes is capped regardless of how it scores elsewhere. Each study area page sets out the full method and every parameter, with the source for any default that area replaces with a local figure.
Where the data comes from
Building footprints, heights, zoning and addresses are the cities' own published GIS, not a commercial dataset.
Nothing here is scraped from listings or bought from a data vendor, which is why the study areas are the ones they are: the model only works where a city publishes building heights or floor counts alongside footprints. Where a city records floor counts, those are used; where it does not, floors are derived from published height and the page says so.
- City of San Jose, “DPW Basemap MapServer (layers 21 and 33)”
- City and County of Denver, “Denver Open Data Catalog: Building Outlines 2022, Addresses, Statistical Neighborhoods, Zoning”
- City of Philadelphia, “OpenDataPhilly: Building Footprints, Address Points, Zoning, Planning Districts, OPA Properties”
Statute facts are read from the enacted text, Public Law 119-101, rather than from coverage of the introduced bill — a distinction that matters here, because the final law dropped the funding mechanism and the program name that most published summaries still describe. They were last checked against the enacted text on .
How the numbers are checked
Every claim on the site is either computed by the engine or carries a source; the two are checked differently.
The floor-plate geometry is validated against shapes whose answers are known analytically, and computed floor areas are checked against the cities’ own recorded areas, which they match to within 0.01%. The engine that produces the published data and the model that re-scores buildings live in the browser are separate implementations, so a parity check asserts they agree on every building in every published study area before the site can build at all. Each study area also carries regression tests over its own data, so a change that quietly moves one city’s results fails the build.
Facts the engine does not produce — what the statute says, a city’s vacancy rate, an incentive a council passed — are carried with a publisher, a title, a link and the date the claim was verified, and they are listed at the foot of the page that uses them.
What it cannot tell you
A high score means a building is worth a professional feasibility study. It never means the building is convertible.
The model sees an outline and a height. It cannot see interior columns, existing cores, ceiling heights, post-tension slabs, mechanical systems, leases, asking rents, or which buildings are actually empty. It does not know which buildings meet the RESIDE Act’s vacant-and-abandoned test, which turns on code-enforcement and court records that no city publishes as open data, so every “unlocked by RESIDE” result assumes a building would qualify. It values every home at market rate, although the Act requires attainable housing, which makes its grant results optimistic. And there is no HUD funding notice yet, so the per-building award formula is an assumption, not a rule.
Each study area page states its own limits in full, including the ones specific to that city’s data.
Corrections
If a building on this site is described wrongly, write to contact@resideact.org and it will be looked at.
The most useful corrections are specific: the page, the figure, and what it should be, with a source if there is one. Owners, architects and city staff who know a building’s interior know things this model cannot see, and that is exactly the kind of correction worth making.
Where a correction changes a published number, the underlying data is corrected and the page’s update date changes with it. Where the disagreement is about an assumption rather than a fact, the assumption is a slider in each study area’s survey: it can be changed, and every building re-scores against it.
Citing and reusing this
Text and data are licensed CC BY 4.0. Quote it, republish it, rerun it.
Every study area publishes its full scored dataset as GeoJSON, with the geometry, cost, gap and score for each building. Cite this site as “RESIDE Act Feasibility Engine, resideact.org” under CC BY 4.0. If you are writing about a specific building, please link the building’s page rather than the dataset, so readers land on the figures with their caveats attached.