Methodology

• Transactions: URA caveat data for private residential sales — the official record lodged on every transaction. We surface recent caveats (trailing 12 months) and counts.
• Listings: live for-sale and for-rent listings aggregated across PropertyGuru, 99.co and EdgeProp, de-duplicated so the same unit isn't counted twice.
• Project facts: developer, tenure, TOP/completion year, total units and blocks — public development facts.
• Locations: MRT/LRT stations and schools with their coordinates and (for schools) level, used for the proximity pages.
• Commercial & industrial: URA commercial statistics and JTC industrial statistics, transaction records, commercial-portal listings, URA Master Plan zoning and ACRA registrations — detailed in the commercial & industrial section below.

• PSF = price ÷ strata floor area. Median is the middle value (less skewed by outliers than an average).
• Recent vs all-time: we lead with the recent median (trailing ~12 months) because a blended all-time median understates today's price for an appreciating project.
• “Within 1 km of a school” uses the straight-line (great-circle) distance — which is exactly how MOE defines the 1 km boundary for Primary 1 registration priority. P1 priority is claimed for primary schools only; for secondary schools and JCs, 1 km is framed as proximity, since admission there is not distance-based.
• “Near an MRT” = condo projects within 1.2 km straight-line of the station.
• Gross rental yield = annualised median asking rent ÷ the median price, shown only when both are available and the result is plausible. It is indicative, before costs.
• Cost-to-own figures (stamp duty, monthly instalment, income required) use current Buyer's/Additional Buyer's Stamp Duty tiers and TDSR rules at the stress-test rate.
• Remaining lease is derived from the tenure commencement/completion year and shown as an approximation (“≈ N years”), not an exact figure.

• Source — public records. Singapore's agent, agency and transaction data is public information published by the Council for Estate Agencies (CEA) and data.gov.sg under an open licence. We aggregate each agent's and agency's recorded transactions from those public records.
• Ranked by activity, not earnings. Every “top” and “best” agent or agency list is ordered by the number of deals on record — by property type, by town, or by the side represented (seller, buyer, landlord, tenant). We do not rank or publish any individual's earnings or commission.
• What a “deal” counts. A deal is one recorded transaction an agent was registered on. Counts are lifetime unless a period is stated; “active in the last 12 months” reflects deals completed in that window.
• Activity is experience, not a rating. A high deal count signals depth of experience — a factual count, not a quality score, endorsement or review. We show no star ratings and collect no reviews.
• Thin pages aren't indexed. A ranking with fewer than three agents on record is shown for visitors but marked noindex.
• Corrections & removal. If an agent believes a figure is wrong, or would prefer their profile not appear in search results, contact us — we will correct the record against the public register or remove the page from indexing.

• Land-area basis. For a landed house on its own land, the URA caveat records the land area — so landed prices and PSF on PropKaki are quoted per square foot of land, the way landed is actually priced. Strata-landed (cluster) homes price on strata floor area instead, and are excluded from our landed street/area medians to keep the land-PSF signal clean.
• Street and area medians come from URA caveats grouped by street and URA Master Plan 2025 planning area, leading with the trailing-12-month window (all-time shown for depth). Freehold share counts 999-year tenure as freehold — the market convention.
• Plot facts: surveyed land size and lot boundaries come from SLA's cadastral land-parcel map; frontage and corner/intermediate position are derived from lot adjacency and are best-effort estimates.
• Planning flags (Good Class Bungalow Areas, landed housing zones and storey control, conservation, tree conservation areas) are point-in-polygon checks against URA Master Plan 2025 / SDCP and NParks layers — reliable for orientation, but always verify against URA/SLA records before transacting.
• Distances on landed near-MRT/school pages are straight-line measures from each street's centre point; a specific house's measured distances are in its Landed House Checker report.
• We don't have a house's built-up floor area (caveats record land only), per-house rental history (URA landed rentals carry no house numbers), road reserve / Road Line Plans, or rebuild cost benchmarks — and we say so rather than guess.

• Market series come from official quarterly statistics: URA for office, retail and shop space (price and rental indices, vacancy, stock and supply, history to 1975) and JTC for industrial (occupancy, price and rental indices on a 2012Q4 = 100 base, street rents and the supply pipeline). The two publishers use different geographies and bases and are never merged. Available and vacant stock is measured on nett floor area, pipeline supply on gross — the two are not comparable, and we keep them apart.
• Transactions are URA and JTC records from 1995 — that is where the data starts, not when a building was built. Commercial deals are thin (hundreds a year across ~14,500 buildings), so comparables default to a wider area and a five-year window, and a median is withheld when there are too few deals to support one.
• Listings aggregate CommercialGuru, 99.co and EdgeProp commercial feeds. Most C&I stock is open-listed — one landlord, many agencies — so we show one row per unit with a count of the agents marketing it, and name them all only on the unit's own view. Units are identified by building, type, size and price (the portals carry no unit numbers), so the count is a lower bound.
• Zoning (land-use zone, plot ratio, conservation, Central Area) is read from URA Master Plan data matched to each building. About a quarter of buildings carry a URA notation (such as EVA) instead of a numeric plot ratio — we show the notation, decoded, and never compute on it. Buildable headroom is shown only where BCA's built floor area verifies it.
• Rents are shown at the most specific level published, and always say which level they are. This building: JTC J-SPACE's rentals of strata-titled multiple-user factories, the 25th percentile, median and 75th percentile of the gross rent declared to IRAS for leases in that building each quarter. Its street: URA's rental statistics for office and retail space by street (IRAS-stamped contracts; URA publishes a median only where a street has at least two buildings and three contracts in the quarter), and JTC's industrial rents by street. A region band: URA REALIS's grid by postal district or office location. None of them is the rent of a particular unit, and a yield built from them is given as a range, never one figure. We also hold JTC's posted land rents (per square metre of land a year), JTC's landed-factory price tenders and SLA's tender and bidding results for state properties (data.gov.sg, Singapore Open Data Licence) — each one award for one site, not a market rent. Winning bidders without a company name are not shown.
• ACRA counts are registrations, not tenancy. Companies registered at an address may not occupy space there, and a low count does not mean a building is empty. We label them as registrations.
• Building facts (year completed, storeys, developer) carry their source on every page and answer, and an official record outranks a portal: HDB Property Information (data.gov.sg) for the year and top floor of HDB shop blocks; JTC's industrial completions for the first TOP of major industrial projects since 2013 (renovations never count); then the facts CommercialGuru and EdgeProp publish on their building pages and listings, labelled as theirs. A total-units figure is CommercialGuru's own count, never verified.
• Strata facts come from BCA's MCST register (data.gov.sg, Singapore Open Data Licence): the number of strata units, the managing agent and the year the MCST was formed, which is not the year the building was completed. A mixed development's MCST counts its homes too, so there we show its lots as the whole development's. Floors and year for about 65 of Singapore's tallest towers come from Wikipedia's list of tallest buildings (CC BY-SA).
• Where a building sits: the JTC estate from JTC's estate boundaries (data.gov.sg), and an office/retail submarket (Raffles Place, Marina Bay, Shenton Way / Tanjong Pagar, City Hall / Bugis, Orchard, city fringe, regional centres, suburban) from our own map of URA planning subzones. Brokers draw these lines differently, so we name it as ours.
• Who's registered here counts businesses by industry from ACRA's corporate entities (data.gov.sg, Singapore Open Data Licence): companies, LLPs and branches whose registered address is the building. These are registered businesses, not who occupies the space. Sole proprietors are not counted and no business is named; a building where most registrations are dead is marked as a registered-office address.
• Owners and valuations come from the owners' own filings on SGX: notices of valuation of real assets (Listing Rule 703), asset acquisition and disposal announcements, and annual reports, filed by Singapore REITs (CapitaLand Integrated Commercial Trust, CapitaLand Ascendas REIT, Mapletree Pan Asia Commercial Trust, Mapletree Industrial Trust, Mapletree Logistics Trust, Keppel REIT, Keppel DC REIT, Suntec REIT, Frasers Centrepoint Trust, Frasers Logistics & Commercial Trust, ESR REIT, AIMS APAC REIT, Alpha Integrated REIT, OUE REIT, Lendlease Global Commercial REIT, Starhill Global REIT) and listed developers (among them GuocoLand, UOL, Singapore Land Group, Hong Fok, Ho Bee Land and Frasers Property). We take facts only, each with its date and filing: a building's owner is the one whose latest filing lists it, and an owner of part is shown as such. A valuation is an independent valuer's opinion for the owner's accounts, not a sale price; occupancy and WALE are as of the date given. Buildings with a private owner show no owner — we never name one.
• Building specs (ceiling and floor-to-floor height, floor loading, power, lifts, loading bays, ramp-up, 40-foot container access, air-conditioning, facilities) come, in this order, from the developer's sales brochure — the building as designed at launch, so a unit may have been altered since — then JTC Find Space (JTC's listing of its own properties), the owner's annual report (car-park lots), CommercialGuru's facilities list, and what at least two listings in the building advertise. We keep a brochure figure only when it is printed on the page next to its own label, and never for the homes of a mixed development. A floor-to-floor height is slab to slab, not a clear ceiling height. A spec we don't hold for a building is shown as not on record.
• Commercial and industrial floor plans come from the developer's sales brochure — as designed at launch, so a unit may have been merged, subdivided or altered since. Most are plans of a whole storey with every unit numbered. We link a plan to a unit with a recorded sale only where the plan prints that unit's number, and where it prints the unit's area, only when that area matches the area on record. A brochure stamped or re-made by a third party gives facts, never an image. Plans are shown only in conversation with the PropKaki assistant.
• Getting there, parking and foot traffic for commercial and industrial buildings: LTA DataMall passenger volumes by train station and bus stop (average daily fare-card taps in and out, by month and day type) and its development car parks; URA's car park and season parking rates (URA Data Service); HDB and JTC car park information on data.gov.sg, with HDB's published parking charges; workers by planning area of workplace from the Census of Population 2020; and residents by subzone from the Singapore Department of Statistics (June 2026) — all under the Singapore Open Data Licence. Taps are people passing a station's gates, not visitors to any building, and the building's own car park is not in the open data. The foot-traffic score is our own model: station taps (weighted down with walking distance), bus taps within 400 m, workers and residents around, and for retail the competing malls within 1 km, ranked against buildings of the same kind (retail, office or industrial). It reads “busier than N% of retail buildings” — a relative score, never a count of people.
• Market context for commercial and industrial buildings cites the brokers' own quarterly figures: CBRE Research's Singapore Figures and Cushman & Wakefield's Singapore MarketBeat (office, retail, business park and industrial rents, vacancy, capital values and yields). We take the figures only, each with its broker, quarter and a link to the report, and show each broker's figure under the broker's own area name (“Core CBD (Grade A)”) for the part of Singapore a building sits in, beside URA's and JTC's official series. A broker's submarket figure describes that market, not any one building's rent, and brokers draw their areas differently. Flexible workspace centres come from the operators' public location lists (JustCo, Industrious, WeWork, The Executive Centre, The Work Project), dated when taken. Collective sales of commercial and industrial buildings since 2024 are compiled by hand from marketing agents' announcements and news reports, each linked; a collective sale needs the owners' consent and may fail, and a closed tender with no reported result is shown as such. How fast units move is from listings we track ourselves: units listed now, new listings and listings that came off between our checks, and days on market only where we saw a listing both appear and disappear.
• OpenStreetMap (© OpenStreetMap contributors, ODbL) supplies a building's storeys only when no other source has them, and is credited wherever it is shown. It is kept apart from our own building records.
• From our own records: units with a recorded sale (a lower bound, never a building's total), price per square foot by floor (ground versus upper, office low, mid and high floors; a median only from 10 sales up), and what units have been listed as on the portals, which is an agent's description and not an approved use. Unit numbers are never shown.
• What's been approved here comes from URA's planning decisions (URA Data Service, every decision since 2000): changes of use, additions and alterations and new buildings, matched to a building by its address or land lot. It is the building's history, not today's approved use of any unit: a temporary permission lapses, and an approval belongs to the unit it names. Where URA does not approve new eating houses or bars, we say so from URA circular DC23-08 (2 October 2023, the streets where new eating houses in shophouses are not allowed) and URA's Use of Buildings in Conservation Areas (no new bars, pubs, nightclubs or karaoke lounges in the Historic Conservation Areas). Both are guidelines, not a decision on a unit. A unit's approved use is answered only by URA's Enquiry on Approved Use of Premises (S$54.50 a unit), which we point to and never use in bulk.
• We don't have completion years for most private C&I buildings (about a third of commercial and a tenth of industrial buildings carry one), building specs such as ramp-up access, ceiling height or floor loading, a unit count for buildings without an MCST (single-owner towers and factories), the approved use of a particular unit, and CEA's transaction records are residential-only — so there are no commercial agent rankings or leaderboards, and we say so rather than fabricate one.

• Counts are exact; some medians are withheld. When a location has more than ~1,000 listings or transactions, the median is computed over a sample that can be biased — so we show the exact count but suppress that median rather than publish a skewed number. Most locations are well under that threshold.
• Listing figures exclude new-launch placeholder prices and room rentals, to match what a buyer would actually shortlist.
• Thin pages aren't indexed. A location or project with fewer than three projects / transactions is shown for visitors but marked noindex — we don't manufacture pages where there's no real data.
• We don't have unit-level stack/facing/view, new-launch balance units, maintenance (MCST) fees, exact HDB grant amounts, or live mortgage package rates — and we say so rather than guess.

Figures are re-aggregated from the latest data each reporting period. Data-dependent pages show an “as of” / “last updated” date and carry a machine-readable dateModified so the visible and structured freshness signals agree. Transacted figures cover the trailing 12 months to the stated month.