How Nairobi's Digital Archives Became a Graveyard of Duplicate Images — and What It Will Take to Fix It

Nairobi County's digital infrastructure is carrying a problem nobody planned for: tens of thousands of duplicate images clogging government servers, slowing down citizen-facing portals and complicating the very e-governance push that was supposed to make public services leaner. The issue, which has quietly frustrated ICT officers across City Hall for the better part of three years, is now forcing a structured cleanup — one that officials describe internally as long overdue.

The timing matters. Kenya is deep inside an IMF austerity cycle, with the Ruto administration under intense fiscal pressure to demonstrate that every shilling spent on technology actually delivers. After the Gen Z-led tax revolt of mid-2024 torched public appetite for wasteful spending, any sign that government servers are bloated with redundant data draws exactly the kind of scrutiny that no ministry wants. Digital efficiency is no longer just a technocratic goal — it has become a political one.

How the Problem Was Built, Layer by Layer

The roots stretch back to roughly 2018, when multiple digitisation drives ran simultaneously without a unified asset management protocol. The Nairobi Metropolitan Services, operating out of its Upperhill headquarters, ran its own document and photo repositories. Nairobi County's own ICT directorate at City Hall on City Hall Way maintained separate servers. The e-citizen platform, managed at the national level by the State Department for ICT, pulled imagery from both sources and added its own. Nobody deduped.

When the Nairobi Expressway opened in 2022 and generated a wave of infrastructure photography for public communications, the same images — aerial shots of the Mlolongo interchange, ground-level views of the Museum Hill junction — were uploaded independently by at least four separate units, according to public procurement records reviewed by The Daily Nairobi. The problem was structural, not accidental. Each procurement cycle bought storage without auditing what was already there.

The Silicon Savannah narrative accelerated things further. Between 2020 and 2023, donor-funded smart city pilots in Kibera and Mathare produced thousands of geo-tagged survey photographs for urban planning purposes. Those images landed in at least three separate repositories — ones held by UN-Habitat's Nairobi Urban Lab on Gigiri's United Nations Avenue, the county's own Urban Renewal Secretariat, and the Kenya National Bureau of Statistics digital archive in Upper Hill. Cross-referencing never happened.

The Cost of Redundancy

Storage is not free. Kenya's government cloud contracts, several of which run through the Konza Technopolis Development Authority, bill on a per-gigabyte model. Industry benchmarks for enterprise cloud storage in East Africa currently sit around Ksh 8 to Ksh 12 per gigabyte per month, depending on the provider tier. When duplicates inflate a database by even 30 percent — a conservative estimate for unmanaged public sector archives, based on comparable audits conducted in Lagos and Johannesburg — the monthly overrun compounds fast across a multi-terabyte estate.

Beyond cost, duplicates create workflow failures. When a Nairobi Metro commuter rail communications officer pulls an image of the Syokimau station for a public update, the wrong version — an outdated one showing old signage — surfaces because duplicates lack consistent metadata tagging. The correct, current image exists on the same server. Finding it takes time nobody has.

The Konza-anchored Ajira Digital programme, which has trained thousands of young Kenyans in digital work since its 2016 launch, has produced a generation of data workers who understand deduplication. The irony is that government has been slow to deploy that exact skill set against its own archives.

The fix being discussed inside City Hall involves a phased duplicate image replacement protocol: first, automated hashing to flag identical files; second, human review of near-duplicates; third, a canonical file library with enforced metadata standards going forward. Pilot work is expected to start in the third quarter of 2026 on the county's land registry image holdings in Lands House on Community. If it works there, the model travels to health and infrastructure records next. The question is whether the budget survives the next round of austerity cuts before the work is done.