Zurich's Digital Archive Problem: The Hidden Cost of Duplicate Images in the City's Data Infrastructure

Zurich's cantonal digital administration is sitting on a problem measured in terabytes: duplicate image files have accumulated across public databases at a rate that now costs the city an estimated CHF 1.2 million annually in unnecessary cloud and on-premises storage, according to an internal efficiency audit completed by the Stadtentwicklung Zürich office in May 2026. The audit, which examined holdings across 14 municipal departments, found that roughly 34 percent of all image assets stored on city servers were redundant copies of files already catalogued elsewhere in the system.

The timing is not incidental. Zurich's cantonal government is midway through its Digitale Verwaltung Schweiz roadmap, a federal-cantonal programme designed to consolidate public-sector data infrastructure by the end of 2027. Duplicate image data — product photos uploaded multiple times to different procurement portals, planning maps rescanned without reference checks, archive photographs resubmitted across departmental boundaries — has emerged as one of the least glamorous and most expensive bottlenecks in that consolidation effort.

Where the Numbers Come From

The audit drew on server logs from the Stadthaus on Stadthausquai, the Amt für Städtebau in Lindenhof, and the digital asset management platform shared by Zürich Tourismus and several cultural institutions along Limmatquai. Across those systems, deduplication scans identified more than 4.7 million image files, of which approximately 1.6 million were flagged as bit-for-bit or near-identical duplicates. File sizes ranged from small thumbnail previews to uncompressed archival TIFFs running to several hundred megabytes each.

Storage costs in Switzerland are among the highest in Europe. A 2025 benchmark study by the Zürcher Hochschule für Angewandte Wissenschaften (ZHAW) in Winterthur placed the average fully loaded cost of cantonal on-premises storage at CHF 0.18 per gigabyte per month — roughly double the equivalent figure for comparable German federal states. Cloud overflow contracts with private providers push that unit cost higher still during peak archiving periods, typically in January and July when annual planning documents are filed. Applied to the volume of duplicate imagery identified in the May audit, the annual waste figure of CHF 1.2 million is a conservative estimate; the audit's own upper-bound scenario reaches CHF 1.8 million.

ETH Zurich's Data Analytics Lab on Rämistrasse published related research in March 2026 showing that perceptual hashing — a technique that identifies visually similar rather than merely byte-identical images — typically recovers an additional 12 to 18 percent of redundant files that standard deduplication misses. Applied to Zurich's catalogued holdings, that methodology could flag a further 190,000 to 290,000 files for review.

What Comes Next for City Systems

The Stadtrat is expected to vote in September 2026 on a procurement tender for a centralised digital asset management system, with deduplication and perceptual hashing built into the ingestion pipeline from day one. Three vendors submitted preliminary proposals by the June 30 deadline; the contract ceiling is set at CHF 4.3 million over five years, a figure that the Stadtentwicklung Zürich office argues pays for itself within the first 30 months if storage savings materialise as projected.

For departments that generate large image volumes — particularly the Amt für Baubewilligungen, which processes thousands of planning photographs each year, and Stadtarchiv Zürich on Neumarkt, which is digitising analogue holdings at an accelerating pace — the practical change will be a mandatory deduplication check before any new image batch is committed to the archive. Files flagged as potential duplicates will require a human reviewer to confirm or dismiss the match before storage allocation is granted.

Residents and civic groups engaged in participatory planning processes, such as those using the Stadtentwicklung's online Mitwirkung portals, upload their own map annotations and site photographs. Under the proposed system, those uploads would also pass through the deduplication layer, though personal data protections under the Swiss revDSG would apply. The Stadtrat's September vote will determine whether those protections require a separate, privacy-preserving pipeline for citizen-submitted imagery — a technical distinction that may affect the final contract cost.