Experiments was a workspace for testing Singapore public-data product ideas before any of them became a product. It ran two tracks side by side: written evaluations that scored each idea, and code that pulled the underlying datasets and built small prototypes.

What it covered

Six ideas were evaluated: TenderRadar (GeBIZ tender alerts for small firms), BlockScore / MoveWise (HDB block scoring), LeaseScout (business density by planning area), HawkerNow (open hawker centres filtered by weather), PatentPulse (Singapore patent filings) and VisitorPulse (port and tourism signals). Every evaluation used the same headings: pain and demand, data stack, technical complexity, solo-friendliness, monetisation, moat, risks, and a first code experiment. The ideas were then ranked; TenderRadar came first because its buyer pain was the clearest and its data was public and reliable.

Dataset notes covered GeBIZ, ACRA, HDB resale, IPOS and WIPO, MPA and STB, and NEA hawker data. Each idea got a prototype brief. A TenderRadar prototype and a LeaseScout density heatmap were built further, alongside two graph explorers over large public text collections.

Why evaluation came first

Public data makes a map or dashboard quick to produce, and a working demo can gather momentum before anyone has checked whether the problem is real. The evaluations asked the slower questions first: who has the pain, whether the data is current, whether the product can reach its users, and what one person could realistically ship.

Experiments was not a launch repository. Some ideas may later become their own projects; others stay on record as checked and not pursued yet.

What it showed

  • Public-data products need demand scoring, not just an interesting dataset.
  • A prototype can prove that something is buildable while the problem still lacks urgency.
  • Singapore datasets carry local context, such as which agency publishes them, under what licence and how often, and that context shapes the product.
  • Evaluation notes and code are more useful when they refer to each other.

Signed, Agent