SF City Agencies Are Drowning in Duplicate Images — And the Numbers Show How Bad It's Gotten

San Francisco's Department of Technology has flagged a chronic and worsening problem buried inside the city's sprawling digital infrastructure: thousands of duplicate image files clogging shared servers, slowing public-facing platforms, and burning through cloud storage contracts that taxpayers fund. According to an internal review presented to the city's Committee on Information Technology in June 2026, redundant image assets now account for an estimated 34 percent of total unstructured file storage across six major municipal departments — a figure that has roughly doubled since fiscal year 2022–23.

The timing is uncomfortable. San Francisco's tech and administrative budgets are already stretched after a bruising round of layoffs across the private sector pushed down payroll-tax receipts, and Mayor Daniel Lurie's office has instructed department heads to find efficiency savings before the fiscal year 2026–27 budget closes in late July. Duplicate data — images in particular — turns out to be a surprisingly large line item when aggregated across Muni, the Planning Department, SF Public Works, and the city's 311 service portal.

Where the Redundancy Is Worst

The problem is most acute at SF Public Works, which maintains a photo archive of street-condition reports stretching back to 2014. Workers in the field submit images through a mobile app that feeds into the department's asset-management system on Bryant Street. That system, according to the June review, contains more than 1.2 million image files, of which an automated audit found at least 410,000 flagged as probable duplicates — same file hash, different timestamp or filename. Public Works pays for cloud storage through a Master Service Agreement with a vendor whose rates are pegged at roughly $0.023 per gigabyte per month; the department's image library alone occupies more than 18 terabytes.

Muni's situation is different but no less costly. The Municipal Transportation Agency uses image data for everything from signal-camera feeds archived for insurance disputes to marketing photography recycled across platforms. Staff at the MTA's Presidio Division maintenance facility have noted that images pulled from old campaigns regularly re-enter circulation without any deduplication check, meaning the same JPEG can exist under a dozen different filenames across shared drives and the agency's content-management system.

The 311 portal — San Francisco's primary civilian-complaint intake point — is where the problem has the most direct public-service impact. Residents submitting pothole or graffiti complaints on Fell Street or along the Embarcadero sometimes attach photos already in the system from prior reports on the same block. The portal's underlying database, built on a Salesforce government platform and managed out of City Hall's Office of Civic Innovation, does not currently run a real-time deduplication pass at the point of upload. That means storage consumption grows with every resubmission, and case-workers manually sorting reports spend additional time cross-referencing images that are functionally identical.

What a Fix Actually Costs — and What Doing Nothing Costs More

The Department of Technology's June review estimated that deploying a perceptual-hash deduplication tool across the six departments would require a one-time implementation cost in the range of $280,000 to $340,000, covering software licensing, integration work, and staff training. Ongoing annual maintenance was pegged at approximately $45,000. Against that, the review projected annual storage savings of around $190,000 and a reduction in staff time spent on manual image management of roughly 2,100 hours per year — hours currently billed at city pay grades that average close to $78 per hour when benefits are included.

The math is not complicated. Breakeven on the capital outlay arrives in under two years, and every year after that represents net savings in a budget environment where the controller's office is hunting for exactly this kind of operational inefficiency. The Committee on Information Technology is expected to take up a formal recommendation at its August session, with a potential pilot program targeted at SF Public Works and the 311 portal first, before any city-wide rollout.

For San Franciscans filing service requests through 311 or watching for faster pothole repairs on streets like Valencia or Cesar Chavez, the practical upside is a system that routes complaints more accurately and resolves them faster — because the staff handling them aren't wading through duplicate evidence. The data made the case. Now the city has to act on it.