What this site refuses to tell you
15 things it has been asked for and turned down, each with the reason and the file that enforces it.
Every property site in Singapore has a longer feature list than this one. Some of that is work not done yet. Most of it is on this page, and none of it is here because it was difficult.
A refusal is only worth anything if you can check it, so each one names the file whose comment carries the argument. Nothing that is simply unbuilt is listed — a backlog is not a principle.
Numbers this site will not produce
Each of these is sold by somebody in this market. The objection is never that it is hard.
A valuation. One number for what your home is worth.
Asked for: The most requested thing there is, and the headline product of both comparable sites.
No public dataset can see your floor, your facing, your renovation or your lease. Valuation tools routinely disagree by tens of thousands on the same home, and the disagreement is not a bug in one of them — it is the honest width of the answer. This site shows the observed range and the transactions behind it, and lets you decide where inside it you sit.
Rule 2lib/blindspot/rubric.js
A projected launch price from what the developer paid for the land.
Asked for: Proposed as a "breakeven ladder": land price, plus construction, plus a margin.
Two of those three are published by nobody. Construction cost varies by site, by contractor and by year, and no developer files its margin — so the sum would be one fact carrying two guesses, presented as arithmetic. It is also a valuation of a building that does not exist yet.
Rule 2components/LandView.jsx
A project scorecard — one grade for whether a development is good.
Asked for: A competitor sells exactly this.
A score assembled by a language model is an opinion wearing a number’s clothes: it reads objective, it cannot be sourced, and it will not give the same answer twice. Blindspot does publish a score, and the difference is the whole point — its formula is printed on the page, a model never assigns a point, and the same inputs always produce the same result.
A model never assigns a numberlib/blindspot/rubric.js
Your purchase ranked against the stock market or gold.
Asked for: Proposed as an "opportunity cost engine" scoring a home against STI, the S&P 500 and gold.
None of those series is held here, each needs sourcing and licensing, and ranking a home against equities is investment advice, which nobody here is licensed to give. What survived is the CPF Ordinary Account rate — statutory, published, and where most of the money actually came from.
Not licensed for itlib/calc/ledger.js
Data this site will not use
Two of these are licensing. The third is about drawing things nobody published.
REALIS — the unit number on a private sale.
Asked for: It is what lets a competitor show the actual unit a transaction happened in, and they say so.
URA licenses REALIS for personal research, not commercial use. It is also the difference between a site whose every figure you can check against a free government source and one where part of the answer cannot be reproduced from public data — which that competitor’s own methodology states plainly.
Rule 1 · CEA PG 02-11 s6CLAUDE.md
Anyone else’s reporting.
Asked for: The obvious way to have a news section.
Nothing here reproduces the Straits Times, Business Times, EdgeProp or Stacked. The archive indexes primary sources — government announcements and public datasets — and links to them.
Rule 9app/archive
Geometry the data does not contain.
Asked for: A rail line drawn between stations. A boundary drawn around a point. Future MRT opening years.
A list of station coordinates is not a route, and a point is not a parcel. If a line would come from memory rather than from a published shapefile, it does not get drawn — which is why the future-rail file in this repo is an empty array rather than a plausible guess.
Rule 13data/sources/rail-future.json
Things the data cannot actually support
These were attempted, and the attempt is the record. Each one looked buildable until it was built.
What a specific home actually returned when it resold.
Asked for: The question every owner has, and a competitor sells a version of it.
Pairing filed sales needs a unit identifier, and neither HDB nor URA publishes one. The closest available match is address plus floor area plus floor band, and it is not a unit: Blk 362C Sembawang Crescent filed fifteen 4-room 93 sqm sales on storeys 7 to 9 inside seventeen months, two of them in the same month. A first build of this paired those and produced a confident median holding period out of fifteen different families’ homes. It was deleted rather than shipped.
What is there instead: What homes of your size have sold for here, year by year, beside the same figure for every size at that address — because a project’s headline moves when the mix of sizes sold moves, not only when prices do.
Deleted after building itlib/blindspot/measure.js
How long it takes to walk to the station.
Asked for: On every property site in the market.
What sits between two points — a canal, an expressway, a fence — is in no dataset held here, so a walking time would be a guess presented as a measurement. Distances are straight-line and say so on every line.
Rule 10lib/geo.js
A shaded 1km circle around a school.
Asked for: The standard way every portal shows primary-school priority.
MOE measures to the school’s land boundary, not to a pin in the middle of the campus. On a large site that is a couple of hundred metres, which is the entire width of the decision at a band edge. A circle drawn from one coordinate is not MOE’s measure and would imply a place nobody has been offered.
Rules 11 and 12lib/geo.js
A renovation-age negotiation anchor.
Asked for: Proposed as a way to price how tired a unit is.
Renovation is not in the filed data at all. Inferring it from a building’s age would dress an assumption as a comparable.
Refused at the auditNEXT.md
Advice this site will not give
The line is between a fact about your money and a decision about your life.
When to sell before the lease decays further.
Asked for: Proposed as an "inflection point" — the year to exit a leasehold.
The exit year rested on an invented 2% line that nobody publishes, and telling somebody when to sell their home is advice, not arithmetic. The State’s own leasehold table is published here in full, all ninety-nine years of it, and what you do with it is yours.
Sell advice on an invented linelib/calc/lease.js
A verdict on a floor plan.
Asked for: Proposed as a badge library — good layout, poor layout, scored.
That is a model assigning a verdict, and the plans themselves are the developers’ copyright. What the tool does instead is measure how much of the area is usable and list the questions to put to your designer and an engineer.
A model never assigns a numberapp/floorplan
That anything is undervalued, a good deal, or a buy.
Asked for: Standard vocabulary across this market, including on both comparable sites.
Those are opinions dressed as findings. CEA PG 02-11 s3.1 requires a market claim to be substantiated, and none of those words can be. What you get instead is disclosed arithmetic against a cited source, with the arithmetic on the page.
Rules 6 and 7CLAUDE.md
Your phone number, for anything.
Asked for: Every lead form in the industry asks for one.
There is no number field on this site. It was removed rather than reworded, because collecting a number with no purpose behind it is its own problem. Consent is per channel, never bundled, and an enquiry is not consent to be called.
Rules 3 and 4 · PDPA s14(2)lib/consent.js
What is there instead
- Six checks against filed recordsScored by a formula printed on the page, and it says so when a check could not run
- What the land costEvery government land sale since 1993, the winning tender and every losing bid — the floor under a launch price, without projecting one
- What owning it costsEvery duty, interest and commission, against what the same home actually lets for
- Who publishes this, and under what registrationEverything here goes out under one licensed agent’s number