SF City Records Swamped by Duplicate Images: What Officials and Experts Are Saying

San Francisco's municipal data systems are carrying thousands of redundant image files—duplicate photographs, scanned documents, and permit attachments stored two, three, or more times across city servers—and department heads are starting to talk openly about the cleanup bill. The problem, specialists say, has worsened since the city accelerated its digital permitting push after 2022, when the Department of Building Inspection moved the bulk of its paper-based intake to an online portal.

The timing matters. San Francisco is simultaneously trying to cut administrative overhead and speed up housing approvals—a pressure felt acutely at 49 South Van Ness Avenue, the Civic Center complex that houses Planning and DBI staff. Redundant image data slows search functions, inflates cloud-storage contracts, and, in at least some cases reviewed by city technology staff, has caused version-control errors where inspectors pulled the wrong photograph from a case file.

Why Duplicates Accumulate—and What It Costs

The core problem is workflow design, not human error, according to technologists familiar with government records systems. When applicants upload a site photograph through the city's Accela permitting platform, the file can be saved at the point of upload, again when a planner opens it for review, and a third time when it is exported into a legacy document-management system. Multiply that across tens of thousands of permit applications filed annually and the redundancy compounds fast.

San Francisco's Department of Technology, headquartered on Seventh Street in SoMa, manages the city's enterprise data contracts. Officials there have not publicly released a figure for the total cost of duplicate storage, but comparable mid-sized municipalities that have completed deduplication audits—including several in the Pacific Northwest—have reported storage savings of 20 to 35 percent after systematic cleanup campaigns. The city's overall IT budget for fiscal year 2025-2026 was set at roughly $140 million, making even a modest percentage reduction meaningful in a year when the mayor's office has been hunting for administrative savings.

The Tenderloin Housing Clinic and the San Francisco Human Services Agency have both expanded their client-data digitization over the past two years, partly in response to the fentanyl crisis response mandate that required faster intake documentation at shelter sites along Eighth Street and in the Mid-Market corridor. Staff working with those databases have flagged internally that intake photographs—used to verify identity and track individuals across multiple shelter placements—are among the most frequently duplicated file types in the social-services stack.

Pressure to Act Before the Next Budget Cycle

City technologists and outside specialists are coalescing around a few concrete recommendations. First, they argue the city should implement automated deduplication tools at the point of ingest—before a file ever hits permanent storage—rather than running periodic manual audits. Second, they want a single canonical document-management repository that all departments write to, eliminating the parallel saves that current inter-department workflows produce. Third, they say the city needs a formal image-retention policy with defined deletion timelines, something San Francisco currently lacks for most non-legal-hold records.

The Board of Supervisors' Government Audit and Oversight Committee has the authority to require a formal technology audit. No such audit has been publicly announced as of July 4, 2026, but members of the committee have signaled interest in a broader administrative-efficiency review before the next budget cycle opens in January 2027.

For residents and contractors who interact with city permitting daily—architects filing plans on projects in the Mission District, developers chasing entitlements in the Dogpatch, nonprofit housing builders working through the Mayor's Office of Housing and Community Development—the practical upshot is straightforward: duplicate images slow the queue. A permit reviewer sifting through four copies of the same site photo to confirm a single data point is a reviewer not moving on to the next application.

Technology staff expect to present preliminary deduplication options to department heads before the end of September. Whether that produces a funded action plan or another shelved report will depend largely on how aggressively the mayor's budget office presses the efficiency case in the months ahead.