By the Numbers: How Duplicate Images Are Quietly Draining San Francisco's Digital City Records

San Francisco's city agencies collectively manage tens of thousands of digital records, from building permit photos logged with the Department of Building Inspection on Otis Street to street-condition snapshots filed through the SF311 app. An emerging technical audit problem — duplicate image storage — is quietly compounding those data burdens, and local IT officials are under pressure to quantify and fix it before a new round of budget cuts forces harder choices.

The issue sounds mundane. It isn't. Across municipal databases and the nonprofit and tech-adjacent organizations that plug into city data systems, duplicated image files can account for anywhere from 15 to 40 percent of total storage consumption, according to industry benchmarks published by the Cloud Security Alliance in its 2025 Data Hygiene Report. For a city government that already faces a structural budget deficit projected to exceed $800 million over the next two fiscal years — a figure the San Francisco Controller's Office has flagged repeatedly — redundant storage isn't just an IT annoyance. It translates directly into contract costs with cloud vendors and the staff hours burned managing bloated archives.

What the Local Numbers Actually Show

SF Digital Services, the city's in-house technology unit headquartered at 1 Dr. Carlton B. Goodlett Place, has been piloting a duplicate-detection sweep across several departmental databases since March 2026. The program — internally called the Data Integrity Initiative — targets image repositories attached to three high-volume departments: Public Works, the Department of Homelessness and Supportive Housing, and the Recreation and Parks Department. Early internal estimates, shared in a March 2026 budget committee presentation, suggested that Public Works alone carried more than 120,000 duplicate or near-duplicate images in its field-inspection database, files generated by crews documenting pothole repairs, sidewalk violations, and encampment clearances across neighborhoods from the Tenderloin to the Excelsior.

The Recreation and Parks Department's situation is illustrative in a different way. Staff photographers and contracted vendors documenting work at Golden Gate Park's Botanical Garden and at Dolores Park have, over several years, uploaded event photos through at least three separate content management platforms — none of which cross-referenced the others. The result: thousands of images stored in triplicate, each copy drawing on the city's Microsoft Azure contract, which the city renewed in 2024 under terms the Department of Technology has not publicly detailed beyond confirming the multi-year commitment.

Private-sector organizations operating at the intersection of tech and social services face the same problem. Tenderloin-based nonprofit Glide, which runs one of the city's largest drop-in service operations on Ellis Street, uses image documentation to track facility conditions and client intake records. Digital asset management consultants who work with Bay Area nonprofits estimate that organizations of Glide's scale — handling thousands of interactions per month — can accumulate duplicate image rates above 25 percent within 18 months if no deduplication protocol is in place. Glide has not publicly commented on its specific data practices.

Why 2026 Is a Turning Point

The pressure to clean up these records has intensified for two reasons. First, San Francisco's Proposition B oversight reforms, passed in November 2024, require city departments to produce cleaner data audits as part of expanded public accountability dashboards set to go live in late 2026. Second, the AI tools now being layered onto city systems — including computer-vision programs being tested by SF Public Works to automatically categorize street-damage photos — perform significantly worse when trained on datasets polluted with duplicate or near-identical images. Garbage in, garbage out, as data engineers put it. Redundant images skew model training, inflate false-positive rates, and can cause automated systems to misclassify damage severity.

For residents and city watchdogs, the practical upshot is this: demand that your district supervisor ask SF Digital Services for the full results of the Data Integrity Initiative before the fall 2026 budget cycle closes. The duplicate-image problem is neither glamorous nor politically charged, but it sits upstream of nearly every data-driven city service San Francisco is betting on to stretch its thinning budget further. Cleaner data costs less to store, processes faster, and produces more reliable results. That math holds whether the image in question is a cracked sidewalk in the Mission or a bloom in the Conservatory of Flowers.