How we build a report
Verdari is independent, source-traced research on a single Singapore property — what it's really worth, what it could return, and the risks in the detail. We hold no listings and take no commissions; we're paid by the reader, to be right. This page shows the working: where every figure comes from, what we measure versus what we model, and the checks each report passes before you see it.
What you're getting — and what you're not
A pre-purchase investment research report: independent analysis to inform your own decision. We are not an estate agent (we hold no listings and negotiate for no one), not a valuer, and this is not financial advice. Nothing here is a valuation for mortgage, stamp-duty or legal use — for that, use a licensed valuer.
Four pillars, weighted not averaged
Every property gets one headline score out of 100, built from four output pillars:
- Return (45%) — capital growth and rental yield over your horizon.
- Risk (25%) — lease decay, interest rates, nearby supply, liquidity.
- Liquidity (15%) — how readily the unit resells.
- Value (15%) — price against comparable transactions.
The headline is a risk-adjusted composite — not a simple average — so a serious risk can't be averaged away by a strong return. Hard floors cap the verdict where a disqualifying flaw exists (for example a remaining lease under 40 years, or a negative real return). Bands: Avoid (under 40) · Hold-off (40–59) · Strong (60–84) · Exceptional (85+).
Measured vs modelled — and why we show the difference
Every figure carries its source, its as-of date and its kind:
- Measured data — facts from official records: transacted prices, rental contracts, distances, cooling-measure math.
- Model estimate — our calculation over that data, labelled and shown with a confidence band. Our fair-value figure is a size-adjusted estimate over comparable transactions, presented as a range with a confidence level — never a single false-precise number, and never a “valuation”.
- Inference — a read from the listing you provide (facing, light, layout), clearly marked as such.
Estimates are visually distinct from measured data everywhere they appear. Forecasts are ranges that widen with the horizon, not promises.
Every report is checked before you see it
Before any report is shown, it passes an automated verification gate — the final, deterministic stage of generation. It runs on every report, and it checks, mechanically:
- Reconciliation — every headline figure ties back to its cited source; no un-sourced number.
- Arithmetic & internal consistency — the return waterfall sums to the headline; the score matches its band; comparables, cost stack and stamp duties recompute.
- Provenance completeness — every figure carries a source, an as-of date and a data-vs-model tag.
- Sanity checks — yields, psf, growth and returns sit within plausible ranges; the floor rules fire when they should.
- Confidence completeness — no figure ships as a bare number without a confidence level.
If a critical check fails, the report is held — we don't publish figures we can't stand behind; you see an honest hold notice, not a broken number. Where a report passes, the “✓ Verified · N figures” mark on it reflects that report's own gate run.
Its honest limit, stated plainly: this gate guarantees that figures are faithfully sourced and internally coherent. It cannot certify that the market will do what a forecast projects — which is why forecasts are ranges, and why we guarantee source-integrity, not accuracy (see Corrections).
Where the data comes from
Verdari is built on official Singapore records — named on every figure, used under the Singapore Open Data Licence with attribution:
- URA (Urban Redevelopment Authority) — Private-property transactions, rental contracts, the Master Plan, and Government Land Sales sites.
- SLA (Singapore Land Authority) — Location and distances, via OneMap.
- LTA (Land Transport Authority) — The rail network and committed transport pipeline.
- MAS (Monetary Authority of Singapore) — The SORA reference rate behind indicative financing.
- MOE (Ministry of Education) — Primary-school locations for the P1-proximity read.
Several of these are published through data.gov.sg. We also read the listing you provide — for facts only: we extract what we need, then discard it, and we never re-display a portal's text or photos.
How we check a guide before it is published
Our written guides are held to the same standard as the reports. These are not editorial aspirations — each one is a build check, and a guide that fails any of them cannot be published.
- Every figure we take from a document names that document, and the date we read it. A guide cannot be published with a sourced figure we cannot point at. (Figures that are our own arithmetic show their working instead — see below.)
- Every source is cited where you are reading, not only in a list at the bottom. A page cannot carry a source row that no claim on it points at.
- Every source carries the date it was published and the date we read it, so you can tell how old a fact is.
- Nothing inside quotation marks is a paraphrase. Where a quotation leaves something out, we say what it left out.
- Anything we calculated rather than quoted shows its working, and anything we assumed says so where it is used.
- A definition that states a rule cites the rule it states — we do not gloss a term into an obligation no source carries.
- No guide is published until an independent check has read it and its report has been written down. A check nobody can read afterwards is not a check.
- We never describe our own output as a valuation, and we guarantee source-integrity — that our figures faithfully reproduce their sources — not the accuracy of any forecast.
Two things that follow from doing it this way, and are worth stating plainly. Where a rule is ambiguous we say it is ambiguous rather than resolving it for you — and where a primary source contradicts itself, we show both wordings and tell you to confirm it directly. Nothing on a guide is a valuation or advice; a guide explains a rule, it does not price a property.