São Paulo's Duplicate Image Crisis: The Key Decisions That Will Shape the City's Digital Archives

São Paulo's sprawling municipal digital infrastructure has a growing and largely unacknowledged problem: duplicated images are filling government databases, slowing systems, inflating storage costs, and — in at least one documented case — causing outdated property photographs to appear in official Prefeitura de São Paulo cadastral records used for urban planning along the Marginal Tietê corridor.

The issue cuts deeper than routine IT housekeeping. São Paulo operates one of the largest municipal data ecosystems in Latin America, with agencies from the Secretaria Municipal de Urbanismo e Licenciamento to the São Paulo Turismo company each maintaining separate image repositories that were never designed to talk to each other. When images get uploaded multiple times — through system migrations, user error, or poorly integrated platforms — duplicates pile up silently. Nobody owns the clean-up. Nobody set a deadline. That is about to change.

Why This Moment Matters

The Prefeitura de São Paulo is six months into a broader digital modernization drive that Mayor Ricardo Nunes announced in January 2026, committing to consolidate the city's fragmented IT systems ahead of the 2028 municipal budget cycle. Part of that consolidation involves auditing image repositories across more than 40 secretariats. According to a procurement notice published on the city's Portal de Compras in May 2026, the administration is seeking vendors capable of automated duplicate-detection and image deduplication services, with a contract value ceiling of R$4.2 million.

That tender closes on July 31. Whoever wins will inherit a data estate that, by the city's own internal estimates cited in the procurement document, contains an image library exceeding 18 terabytes spread across legacy servers in the Centro Administrativo building on Viaduto do Chá and newer cloud instances managed through the Prodam-SP municipal technology company headquartered in Lapa.

Prodam-SP has been the quiet center of the debate. The company, which handles IT infrastructure for the city, has been running pilot deduplication tests since March using open-source tools on a subset of the Geosampa urban mapping platform's image layer — the same layer that cartographers and urban researchers at the University of São Paulo's Escola Politécnica rely on for flood-risk modeling in districts like Itaquera and Capão Redondo. Early results from that pilot have not been made public, but the procurement document references a benchmark of removing at least 30 percent of redundant files in the first 90 days of any new contract.

The Decisions Ahead

Three choices will define what happens next. First: whether the city opts for a centralized image repository — one master library that all secretariats pull from — or a federated model where each agency keeps its own database but deduplication runs automatically across them. Urban planning experts have argued for years that São Paulo's size and bureaucratic culture make federation more practical, but it is also more expensive to maintain long-term.

Second: what to do with images flagged as duplicates but not confirmed as such. Automated detection tools can misidentify slightly edited or cropped versions of the same photo as unique files. A wrongly deleted image in the city's heritage documentation archive — maintained partly through the Instituto do Patrimônio Histórico e Artístico Nacional's São Paulo office on Rua Sete de Abril — could mean losing the only digital record of a demolition or renovation in a protected zone.

Third: who audits the process. Civil society groups including Transparência Paulistana have already written to the Câmara Municipal requesting that any deduplication contract include an independent audit clause, arguing that image purges in public databases have, in other Brazilian municipalities, been used to quietly remove inconvenient visual documentation from planning disputes.

The July 31 tender deadline is the first hard date in what is shaping up to be a months-long process. If procurement is completed on schedule, a vendor could begin work by September — putting the initial 90-day deduplication sprint squarely in the middle of São Paulo's rainy season, when the city's drainage and flood-response systems, many of which rely on geolocated image data from the Córrego do Ipiranga basin monitoring network, are under maximum stress. Getting the sequencing wrong is not just an IT problem. It is a public safety question.