San Francisco Is Replacing Duplicate Property Images Faster Than London or Tokyo — But the Work Is Far From Done

San Francisco's Department of Building Inspection has been quietly working through one of municipal government's least glamorous but increasingly consequential tasks: purging duplicate images from the city's digitized permit and property record systems. As of late June 2026, the department had cleared more than 340,000 redundant image files from the Accela permit platform — the same software used by planning offices in Chicago, Seattle, and several European municipalities — freeing up processing capacity and reducing the error rate on digital permit applications by a measurable margin.

The timing matters. San Francisco's Planning Department is in the middle of its state-mandated Housing Element implementation, which requires it to process thousands of Accessory Dwelling Unit applications and multifamily upzoning reviews before a 2029 state deadline. Duplicate images — copies of the same site photograph or architectural drawing filed multiple times by contractors or automated scanning systems — slow down reviewer queues and, in some cases, have caused incorrect parcel data to be pulled into the wrong permit file. The city's own technology office flagged the problem in a 2025 internal audit, noting that roughly 18 percent of image assets in the property database had at least one duplicate entry.

How SF Compares to Peer Cities

London's Ordnance Survey and the Greater London Authority have grappled with a similar problem across the capital's 33 borough planning portals, where duplicate imagery from aerial surveys conducted between 2019 and 2023 created conflicts in the city's Local Plan review system. As of early 2026, London had not completed a unified deduplication pass across all boroughs, leaving each district to manage its own cleanup on separate timelines. Tokyo's Geospatial Information Authority completed a national-level deduplication of cadastral imagery in March 2025, finishing roughly eight months ahead of its original schedule — a benchmark that San Francisco's DataSF team has cited internally as a model, according to agenda documents from a January 2026 City Hall working group meeting.

Closer to home, Los Angeles County began a deduplication project for its Assessor's database in 2023 but the effort stalled mid-2025 amid budget cuts to the county's IT division. San Jose, which uses a different permit platform called Citizenserve, completed its own image audit in October 2024. SF's effort, by contrast, is ongoing and tied directly to the city's broader 2025–2027 Digital Services Modernization Initiative, a program that allocated $4.2 million across participating departments for data hygiene and backend infrastructure upgrades.

Where the Backlog Persists

The hardest-hit neighborhoods are the Tenderloin and South of Market, where older building stock and decades of overlapping permit filings — some dating to post-Loma Prieta retrofits in the early 1990s — produced some of the densest clusters of duplicate records. The Department of Building Inspection's permit center on台 Stevenson Street processes a disproportionate share of those legacy files. Staff there have been using a combination of hash-matching software and manual review to work through the queue.

The SF Planning Department's Case Management Unit on Seventh Street is separately handling duplicates that entered the system through the ePlans electronic submittal portal, which launched in 2017 and initially lacked automatic deduplication logic. That gap was patched in a 2023 software update, meaning new submissions no longer generate the problem — but the historical backlog predating that fix still runs to tens of thousands of files.

Property owners and contractors filing for permits should verify that all image attachments are uniquely named and under the department's 10-megabyte-per-file cap before submission. The DataSF team publishes a monthly data quality report at datasf.org that tracks deduplication progress by neighborhood and permit type. The next report, covering June 2026, is scheduled for release the week of July 13. Officials have indicated the Tenderloin backlog should be substantially cleared by the end of the third quarter — though the SoMa legacy files are expected to take into early 2027 to fully resolve.