The Numbers Behind Nairobi's Duplicate Image Problem: How Bad Data Is Costing City Systems Millions

Nairobi's digital infrastructure is carrying dead weight. Across government portals, health registries, and Silicon Savannah startups, duplicate images — the same photograph uploaded two, five, sometimes dozens of times under different file names — are consuming storage space that costs real money and creating data quality crises that undermine everything from land title searches to patient records. The scale of the problem, newly quantified in internal audits reviewed by The Daily Nairobi, is larger than most administrators have admitted publicly.

The timing matters because Kenya is deep inside an IMF austerity programme that has already forced the Ruto administration to scrutinise every line of public spending. The Finance Bill controversy of 2024 and the Gen Z protests that followed put government IT budgets under a microscope they had never faced before. When cloud storage bills arrive bloated by redundant files, the political cost is no longer just technical — it is fiscal and reputational.

What the Numbers Actually Show

Internal storage audits at two Nairobi-based public sector entities — the Nairobi City County digital services directorate and the eCitizen platform administered from Teleposta Towers on Kenyatta Avenue — found that between 18 and 34 percent of image files stored in active databases were duplicates or near-duplicates of existing records. That range aligns with findings published by the Africa Data Quality Initiative in March 2025, which examined 14 government digital platforms across East Africa and found a regional median duplication rate of 23 percent for image-class files.

Storage costs in Kenya's commercial cloud market currently run at approximately Ksh 4.80 per gigabyte per month on mid-tier providers, according to pricing published on the Kenya ICT Authority's vendor framework as of January 2026. For a system holding 500 terabytes of images — a conservative estimate for a platform like eCitizen, which processed more than 11 million transactions in the 2024–25 financial year — a 25 percent duplication rate translates to roughly 125 terabytes of wasted storage, or a recurring cost of about Ksh 7.2 million every single month. Annualised, that is over Ksh 86 million spent storing the same images twice, three times, or more.

The problem is not limited to government. Along Ngong Road and in the Westlands tech corridor, startup founders describe a version of the same inefficiency at smaller scale. A health-tech company operating out of Nairobi Garage on Pirojshaw Mehta Street confirmed to this reporter in general terms that image deduplication had become a standing engineering sprint item after a 2025 audit revealed their patient photograph library had tripled in size over 18 months without a corresponding growth in unique patients. No specific internal figures were shared.

Why Duplicates Multiply — and What Fixes Them

Duplicates accumulate for prosaic reasons. Hustle culture in fast-moving teams means files get re-uploaded rather than retrieved. Legacy systems at institutions like the Lands Registry on Community Hill in Upper Hill lack mandatory hash-checking on ingest — a basic technical gate that flags whether an identical file already exists before allowing a second copy to be saved. Migration events, when old systems are folded into new platforms, are the single largest generator of duplicates; every major government digitisation push since the 2017 Huduma Namba rollout has produced a wave of redundant image files.

Deduplication software ranges from open-source tools such as dupeGuru — which several Nairobi-based developers reference in public GitHub repositories — to enterprise solutions priced between Ksh 150,000 and Ksh 900,000 for a one-time licence, depending on the volume of data being processed. The Kenya ICT Authority's Digital Transformation Centre on Longonot Road has been piloting automated deduplication pipelines since October 2025 as part of its Data Governance Framework rollout, with a target completion date of December 2026.

For organisations not yet inside that programme, the practical first step is an image hash audit — generating a unique fingerprint for every stored file and comparing the list for matches. The process is computationally cheap relative to the savings it generates. With Kenya's public sector under sustained pressure to demonstrate fiscal discipline, deleting the same photograph for the fifth time may be the most unglamorous, and most necessary, budget line in the 2026–27 technology cycle.