San Francisco Is Quietly Digitizing Its Building Stock — Here's How It Compares to Cities Doing It Better

San Francisco's Department of Building Inspection has been running a low-profile cleanup of its digital property record system since early 2025, targeting a specific and stubborn problem: thousands of duplicate, mismatched, or outdated building images lodged inside the city's permit and parcel databases. The backlog, spread across records covering neighborhoods from the Tenderloin to the Outer Sunset, has slowed permit reviews, confused title searches, and muddied the city's own housing inventory counts at a moment when every unit on the books matters.

The timing is not accidental. San Francisco is under a state-mandated housing element obligation to plan for tens of thousands of new units through 2031, and accurate parcel data is foundational to that work. When a building's file contains two conflicting photographs — one showing a vacant lot from 2009, another showing a four-story residential building from 2022 — planners and permit technicians have to stop and manually reconcile the discrepancy. Multiply that across a portfolio of roughly 200,000 parcels and the drag on the system becomes real.

What San Francisco Is Actually Doing

The Department of Building Inspection partnered with the city's Department of Technology to run an automated deduplication pass on the property image library beginning in January 2025. The effort uses hash-matching software to flag exact duplicates and a secondary review layer for near-matches — photographs taken of the same building from different angles or in different years that may represent legitimate historical data or may simply be redundant uploads from contractor submissions.

The San Francisco Planning Department's DataSF platform, which publishes open datasets on permits and parcels to the public, has been a secondary beneficiary. Cleaner image records feed into cleaner parcel data, which in turn makes the public-facing datasets more reliable for researchers, developers, and neighborhood advocacy groups like the Tenderloin Housing Clinic, which relies on those records when tracking single-room occupancy buildings and their habitability status.

City Hall officials have not publicly announced a completion date for the deduplication project, and the Department of Building Inspection did not respond to a request for comment by press time. But the DataSF changelog shows the property image tables were last updated in May 2026, suggesting the work is ongoing rather than finished.

How Other Cities Have Tackled the Same Problem

San Francisco is not alone in confronting this. Amsterdam's municipal cadastral office completed a full photographic deduplication of its building registry in 2023, working from a baseline of roughly 500,000 structures across the city and its inner suburbs. The Dutch project used an AI-assisted image classification layer on top of conventional hash-matching, cutting manual review time by an estimated 60 percent, according to documentation published by the city's digital infrastructure directorate.

Singapore's Urban Redevelopment Authority has taken a more aggressive approach, integrating street-level imagery directly from its own municipal capture fleet — rather than relying on contractor-submitted photos — which largely eliminates the duplication problem at the point of ingestion. That approach costs more upfront but has kept Singapore's building image database largely clean since the program launched in 2021.

London's approach has been patchier. The 32 London boroughs each maintain their own planning databases under a fragmented system, and duplicate image problems have been repeatedly flagged in audits of the Planning Portal, the national online system that channels applications. A 2024 review by the Planning Advisory Service found inconsistent image standards across boroughs, with some accepting uncompressed files and others imposing strict size limits, creating a mismatched archive.

San Francisco's effort sits somewhere in the middle of this global spectrum. It has the institutional will and the DataSF infrastructure to make the cleanup meaningful. What it still lacks, according to urban data specialists familiar with the project, is a mandatory image submission standard for new permit applications — the rule that would prevent the duplicate pile from rebuilding itself over the next decade.

The practical consequence for residents and developers: cleaner records should mean faster over-the-counter permit approvals at the DBI's Counters at 49 South Van Ness Avenue, where backlogs have drawn consistent criticism from contractors and homeowners alike. Whether the deduplication effort translates into measurably shorter wait times is something the city's own performance metrics, published quarterly on DataSF, should begin to reflect by the end of 2026.