SF City Departments Are Drowning in Duplicate Digital Images — and the Numbers Tell a Damaging Story

San Francisco's city government is storing tens of thousands of duplicate digital images across its departmental servers — redundant files that, according to a municipal IT audit completed in March 2026, are consuming an estimated 34 percent of allocated cloud storage across several agencies, at a recurring annual cost that runs into the low seven figures when licensing and retrieval fees are factored in. The problem is not new, but pressure to address it has reached a breaking point as the city accelerates its push to digitize housing permits, homelessness case files, and transit maintenance records.

Why does this matter now? The Department of Technology, headquartered on Seventh Street in SoMa, has been absorbing the operational shock of two converging forces: a wave of AI-assisted document processing tools that city departments began piloting in late 2024, and a court-mandated expansion of the Sunshine Ordinance Task Force's public records disclosure requirements that took effect in January 2026. Both developments dramatically increased the volume of image files — scanned permits, surveillance stills, case photos — flowing through city systems. Neither came with an automated deduplication protocol. The result is a storage sprawl that IT managers inside City Hall have described in internal memos as unsustainable.

The problem is visible in at least two high-profile city operations. The San Francisco Planning Department, which processes permit applications for roughly 7,000 to 9,000 residential and commercial projects annually, relies on a document management system called ProjectDox that ingests scanned site photographs alongside engineering drawings. Staff at the Permit Center on Raburn Place, near Civic Center, have flagged repeated instances where the same JPEG file — a facade photo or a utility diagram — appears three or four times in a single project folder, each uploaded by a different reviewer. Separately, the Department of Homelessness and Supportive Housing, which manages the Navigation Center network including sites in the Tenderloin and on Division Street, stores case-worker photographs of client belongings and unit conditions. Cross-referencing those image libraries has revealed duplication rates that, in one internal department count from February 2026, exceeded 40 percent of total stored files.

What the Storage Bills Actually Show

City budget documents filed with the Controller's Office for fiscal year 2025-26 show the Department of Technology allocated $4.1 million for cloud infrastructure contracts, of which a significant portion covers Microsoft Azure and Google Cloud storage tiers. While the city does not break out image storage as a separate line item, an independent analysis commissioned by the Budget and Legislative Analyst's Office — published April 14, 2026 — estimated that eliminating verified duplicate files across the seven largest departments could free between 18 and 22 terabytes of storage and reduce annual cloud expenditure by somewhere between $380,000 and $520,000. That figure does not account for staff hours spent manually tagging or re-uploading files that already exist in the system. At the city's standard IT hourly billing rate of $127 per hour, even modest redundancy in labor adds up fast.

The city's 311 Service Center, which fields public records requests under the Sunshine Ordinance, processed 2,841 image-related records requests in the first quarter of 2026 alone — up 19 percent from the same period in 2025, according to the Controller's Office dashboard updated June 30. Retrieving a responsive image set from a bloated, unindexed archive can take a records coordinator significantly longer than it should, and delays beyond the ten-day statutory window expose the city to complaint proceedings before the Sunshine Ordinance Task Force.

What Comes Next for City IT

The Department of Technology issued a request for proposals in late June 2026 seeking vendors with experience in automated hash-matching deduplication — a process that assigns each image file a unique digital fingerprint and flags identical copies for removal or consolidation. Responses are due by August 7. Whatever vendor wins the contract will likely start with the Planning Department's ProjectDox archive, which city IT staff have identified as the most acute pain point given permit volume and the legal sensitivity of the underlying records.

For residents tracking housing approvals or filing public records requests through the city's open data portal at data.sfgov.org, the practical takeaway is straightforward: cleaner image archives mean faster retrieval, fewer disclosure delays, and — eventually — a smaller slice of taxpayer money spent storing the same photograph of a Mission District fire escape four times over.