Singapore's Duplicate Image Problem: The Key Decisions Ahead for Government Records and Digital Archives

Singapore's public sector is sitting on a growing backlog of duplicate images embedded across government databases, digital archives and public-facing portals — and the window for a clean, low-cost resolution is narrowing fast. The issue has been building quietly for years, accelerated by the rapid digitisation push that followed the Smart Nation initiative launched in 2014, but administrators now face concrete deadlines tied to infrastructure upgrades and storage contract renewals scheduled for late 2026 and into 2027.

The stakes are practical, not merely bureaucratic. Duplicate image files inflate storage costs, slow retrieval times, compromise data integrity and create real headaches for auditors verifying that public records are accurate. For a city-state that has staked considerable political capital on being a regional AI and digital hub — Singapore attracted more than S$9.4 billion in fixed asset investments in the info-communications sector in 2024, according to the Economic Development Board — a messy back-end is a problem that undermines the front-end pitch.

Where the Bottlenecks Are

The most acute pressure points are concentrated in agencies that process large volumes of visual documentation on a daily basis. The Housing Development Board, which manages records for more than one million flats across towns from Woodlands to Tampines, maintains image libraries covering estate inspections, renovation permits and resale flat assessments. The National Heritage Board, which oversees archives at the National Library Building on Victoria Street as well as the National Archives of Singapore in Canning Rise, holds digitised collections running into the tens of millions of image files. Both institutions are understood to be in active discussions with technology vendors about deduplication workflows, though neither has publicly confirmed a contract award.

The Government Technology Agency of Thailand — GovTech Singapore, rather — has been piloting AI-assisted deduplication tools under its data engineering programmes since at least 2023. The results of those pilots are expected to inform procurement decisions before the financial year closes on 31 March 2027. That timeline is tight. Agencies that miss the procurement window may find themselves locked into legacy storage arrangements for another two-year cycle, a delay that will compound costs as data volumes keep growing.

There is no single universally agreed standard for what counts as a true duplicate versus a legitimate variant — a resized image, a reprocessed scan, a file saved in a different format. That definitional gap is the first decision administrators must resolve. Without a shared taxonomy adopted across agencies, deduplication tools will produce inconsistent results, and the problem will simply reassert itself within a few budget cycles.

The Decisions That Will Define the Outcome

Three choices will determine whether Singapore's public sector emerges from this process with cleaner, leaner archives or simply shifts the problem sideways. First, agencies must decide whether to run deduplication as a one-time project or embed it as a continuous automated process — the latter costs more upfront but avoids the accumulation problem that created the current backlog. Second, the question of who holds authority over cross-agency data standards needs a clear answer; GovTech is the logical home for that function, but formal mandate matters. Third, there is the question of what gets deleted permanently versus what gets archived offline. Deletions that later prove premature can mean the irreversible loss of public records, a serious accountability risk.

For residents and businesses, the downstream effects are tangible. Slower image retrieval from HDB's resale portal, for instance, adds friction to transactions already under pressure from a market where five-room resale flat median prices in mature estates like Queenstown and Bishan have regularly crossed S$900,000 in recent quarters. Faster, cleaner systems would reduce processing time for renovation permit approvals and estate defect inspections — small gains that compound across a million-plus household base.

The next milestone to watch is GovTech's expected release of updated data management guidelines, flagged for the third quarter of 2026. How prescriptive those guidelines turn out to be will signal whether Singapore is serious about treating the duplicate image problem as a structural challenge or a box-ticking exercise. Agencies that move early and adopt common standards before the guidelines land will be better positioned. Those that wait may find the decisions made for them.