Zurich's Duplicate Image Problem: What Happens Next and the Key Decisions Ahead

Zurich's municipal digital infrastructure is carrying a growing weight nobody planned for. Duplicate images — identical or near-identical photographs stored multiple times across separate city databases — have accumulated to the point where administrators at the Stadtarchiv Zürich and the cantonal IT office are being forced to decide how to clean house, and who pays for it.

The problem crystallised this spring when a routine audit of digitised records held across platforms managed by Stadt Zürich's Departement der Präsidentin flagged thousands of redundant image files. Some photographs appear stored four or five times across different departments, the result of years of uncoordinated uploads by offices ranging from urban planning in Oerlikon to cultural services based near the Haus Konstruktiv on Selnaustrasse.

Why This Matters Right Now

Storage is not free. Enterprise-grade server space of the kind Zurich's public administration relies on typically costs between CHF 0.05 and CHF 0.20 per gigabyte per month depending on redundancy tier — costs that multiply fast when image libraries run into the terabytes. Beyond budget, the duplication creates a governance problem: when the same photograph exists in multiple versions with slightly different metadata, archivists cannot reliably establish which copy is authoritative. That matters enormously for legal records, planning documents tied to neighbourhoods such as Altstetten or Schwamendingen, and heritage files submitted to cantonal preservation bodies.

The timing is also awkward politically. Zurich's city council has been pushing a broader digital transformation agenda, and the image duplication issue risks becoming a symbol of the patchwork approach that preceded it. ETH Zurich's Chair of Information Management, which has advised the city on several data governance questions, has previously flagged metadata consistency as a structural weakness in Swiss municipal archiving — a concern that now looks prescient.

Two tools are at the centre of the technical debate. Perceptual hashing — software that generates a fingerprint for each image and flags near-identical files — has been tested by the Stadtarchiv on a pilot batch of roughly 12,000 photographs from the Kreis 4 and Kreis 5 redevelopment documentation. A competing approach, AI-assisted deduplication, promises higher recall rates but requires a procurement process that could extend well into 2027 given Swiss public tender rules under the Bundesgesetz über das öffentliche Beschaffungswesen.

The Decisions That Cannot Be Delayed

Three choices are pressing. First, city departments must agree on a single canonical image repository. Right now, files live simultaneously in the Stadtarchiv's own system, in SharePoint environments managed by individual offices, and in a legacy platform dating to a 2014 infrastructure contract. Consolidation requires political agreement between at least four departmental heads — a process that has stalled before.

Second, administrators must decide what to do with flagged duplicates once identified. Automatic deletion carries risk: a duplicate may carry unique metadata or a different crop that makes it legally distinct. Manual review is thorough but expensive in staff time. A hybrid workflow, where software flags and humans adjudicate borderline cases, is the middle path most archivists are said to favour, though no formal proposal has yet gone to the Stadtrat.

Third, there is the question of citizen access. The Stadtarchiv offers public search access to digitised holdings via its online portal, and any major restructuring of file identifiers risks breaking existing links shared in academic papers, journalism and genealogical research — a constituency that tends to complain loudly and specifically. Redirect mapping, which preserves old URLs while pointing to new canonical files, adds cost but avoids that disruption.

The Stadtarchiv is expected to present a recommendation paper to the relevant municipal committee before the summer recess ends in mid-August. If the committee endorses a procurement path for deduplication software, a public tender could be published by October, with a contract awarded no earlier than the first quarter of 2027. Meanwhile, the existing duplicate files will keep accumulating — at a rate archivists describe as several hundred new redundant images per week — until a system-wide upload policy is formally enforced across city departments.