By the Numbers: Singapore's Duplicate Image Problem Is Bigger Than You Think

Singapore's digital storage bill is carrying a hidden tax. Across public-sector portals, e-commerce platforms and property listing services, duplicate images — identical or near-identical files stored multiple times — now account for a measurable and growing share of cloud infrastructure costs. Industry estimates drawn from regional cloud audit reports suggest that duplicate and redundant image files routinely represent between 20 and 35 percent of total unstructured data stored on enterprise servers, a proportion that compounds annually as upload volumes grow.

The timing matters. Singapore's Infocomm Media Development Authority has been pushing agencies and platform operators toward leaner, more efficient digital infrastructure under its ongoing Digital Government Blueprint. With the government targeting full cloud migration for eligible systems, every redundant file stored is a direct charge on public or corporate accounts. Cloud storage in the Asia-Pacific region is priced in US dollars, and major providers charge in the range of USD 0.02 to USD 0.025 per gigabyte per month for standard-tier object storage — a figure that adds up fast when duplicate image libraries run into tens of terabytes.

Where the Clutter Accumulates

Property listings are among the worst offenders. On platforms serving the Singaporean market, individual HDB resale flat listings in estates like Tampines and Jurong West frequently carry the same floor-plan graphic uploaded multiple times under different filenames — an artefact of how agents batch-upload listing packages. The Housing and Development Board's own resale portal and ancillary services managed by the CPF Board for grant documentation processing each maintain image repositories that, without automated deduplication routines, accumulate redundant files over successive submission cycles.

Retail and logistics are not exempt. At Changi Business Park and one-north — both home to regional headquarters of major tech and retail firms — internal IT teams have flagged image deduplication as a recurring item in cloud cost-optimisation reviews. Product catalogue images, which undergo minor edits for seasonal promotions and are then re-uploaded rather than replaced, are a primary driver. A single SKU in a mid-sized retailer's catalogue can exist in a dozen near-identical versions across staging, production and archival buckets.

The National Library Board, which digitises print and photographic archives at its repository facilities including Lee Kong Chian Reference Library on Victoria Street, uses hash-based deduplication tools as part of its archival ingest workflow. That approach — generating a unique fingerprint for each file and rejecting exact matches — is standard in heritage archiving but has been slower to migrate into commercial content management systems.

The Deduplication Gap

The core problem is not technical. Perceptual hashing tools, which can detect near-duplicate images even when filenames or minor pixel values differ, have been commercially available since the mid-2010s. Services built on these methods can process thousands of images per second and are integrated into platforms from Google Cloud Vision to open-source libraries maintained on GitHub. The gap is operational: organisations do not run deduplication audits regularly, and without a triggering event — a storage bill spike, a migration project — the problem compounds quietly.

For Singapore specifically, the financial calculus sharpens at the enterprise level. A medium-sized e-commerce operator storing 50 terabytes of product imagery and carrying a 30 percent duplication rate is paying for roughly 15 terabytes of files it does not need. At current regional cloud pricing, that is a recurring monthly charge in the range of SGD 400 to SGD 500 just for redundant image storage — before factoring in bandwidth and retrieval costs. Over a three-year contract cycle, the figure exceeds SGD 15,000 for a single storage bucket.

Organisations looking to close this gap have practical options available now. Running a one-time perceptual hash audit using tools compatible with Amazon S3 or Google Cloud Storage requires no vendor contract and can be completed over a single weekend for most mid-sized repositories. Setting upload policies that enforce filename normalisation and reject files above a similarity threshold of 95 percent prevents future accumulation. For HDB resale agents and property platform operators, standardising listing image packages at source — rather than leaving deduplication to the receiving platform — reduces the problem at its origin point. The Smart Nation and Digital Economy Office has published data management guidelines that touch on storage efficiency; those guidelines are worth revisiting with deduplication specifically in scope.