What a report measures, and how.
A location report is only as trustworthy as its sources and its honesty about its own limits. This page is the full account.
About Site Intelligence
Site Intelligence is a Singapore location-screening product: it turns the question “should I open an outlet here?” into an explainable, benchmarked scorecard for a specific address. Reports are built from government and licensed data, an explainable scoring engine, and automated commentary that is never presented as human analysis.
Signal categories
A report scores a site against roughly 18 signal categories. Each is benchmarked against published Singapore data and shown in the report with its weight and contribution — nothing is hidden behind a composite. A Site Snapshot reports these same signals without picking or ranking a business format; only a Feasibility Report or Healthcare Opportunity Scan weights them toward a specific format's fit.
- Population and demandResident population and density in the trade area; age structure (children, working-age, elderly); household composition.
- Affluence and spending powerHousehold income by planning area; HDB resale prices as corroboration; dwelling-type mix (HDB / condo / landed).
- Accessibility and transportBus-stop density; nearest rail station and its ridership; walking reach and walk-time bands; carpark supply.
- Land use and zoningThe site's zoning and the surrounding land-use mix — retail, residential, office and commercial potential.
- CompetitionCount of comparable businesses nearby from Google Places, where a clean category exists. Marked 'not assessed' for verticals with no clean type.
- Institutional proximityNearby schools, universities, childcare centres, hospitals, malls, parks and hotels — curated, licensed lists.
- Context and characterPlanning-area character and commercial context that shape how a site is used day to day.
Data sources
| Source | Vintage |
|---|---|
| Singapore Department of Statistics — Census of Population | 2020 (population, income, dwelling and household data) |
| Urban Redevelopment Authority — Master Plan 2025 land-use zones and plot ratios; subzone boundaries; MRT and LRT station locations | Gazetted 1 Dec 2025 |
| Housing & Development Board — resale transactions | Refreshed periodically; not tracked to a fixed release date |
| Land Transport Authority — bus-stop locations and rail ridership | Most recent available agency release |
| Ministry of Education and ECDA — school and childcare-centre registries | Refreshed periodically; not tracked to a fixed release date |
| OneMap (Singapore Land Authority) — geocoding and basemap tiles | Live at generation time |
| OpenRouteService on OpenStreetMap path data — modelled walk-time zones (a routing service, not government data) | Computed at generation time |
| Curated institutional lists — universities, international schools, hospitals, malls, parks and hotels | Compiled and maintained by us; reviewed periodically |
| Google Places — live competition counts and named nearby businesses | Live at generation time (a paid, per-report call) |
How scoring works
Each address is scored against roughly 18 signal categories. Every signal is measured against Singapore benchmarks (percentiles among published subzones or planning areas) and given a weighted contribution toward the concept-fit score. Weights come from the chosen business vertical's profile — a budget gym and a premium café weight the same signals differently. The result is an explainable score: you can see which signals helped, which dragged, and why. This weighting-and-scoring step is what a Feasibility Report and Healthcare Opportunity Scan do; a Site Snapshot reports the same underlying signals for a site without picking, weighting or ranking any business format.
Percentiles are always shown with their comparison population — a population score means "among all published Singapore subzones"; an affluence score means "among published planning areas." A suppressed figure is shown as unavailable, never guessed.
Peer comparisons
Some reports also show a peer percentile — for example "benchmarked above 75% of 269 comparable café sites." This is a second, independent figure, not a rescaling of the concept-fit score. The peer group is a fixed, dated sample of real Singapore locations inside commercially-zoned land, scored with the exact same engine as your report; it is not a live market, and not a list of actual named businesses. The sample is generated once per business concept and reused until it is refreshed, so its size and composition are stable between reports rather than recalculated live. It is currently only available for concepts with a large enough sample — shown as an honest "insufficient sample" note otherwise, never a fabricated number. Neither the concept-fit score nor the peer percentile is a probability of success.
Walking catchments and trade areas
Reports show three straight-line rings around the site (primary, secondary and fringe bands) with the residents, transit, carparks and land use in each. Where a walking-network isochrone is computed, it is shown as a separate, clearly-labelled view: modelled network reach within walk-time bands, with the explicit caveat that actual entrance routes, crossings and indoor mall circulation are unverified.
Competition
Competition is a live count of comparable businesses near the site from Google Places, where a clean category exists for the vertical. Two things to know:
- For verticals with no clean Google Places type, competition is marked not assessed rather than shown as a zero — the report is explicit when it is not measuring something.
- Named nearby businesses shown in the hosted view expire 30 days after generation and are then removed, per our data-use terms.
Data freshness and vintages
Sources carry their own official vintage — see the table above. Competition counts are live at generation time. Where a dataset's refresh date is not tracked, the report says so rather than inventing one.
Every report also carries its own generation date and a data-coverage indicator (high / partial / low) — so a reader can see how current the figures are and how much of the full signal set was computable for that site. These are two different clocks: a report generated yesterday can still be built on a Census that is several years old, or a land-use layer gazetted months ago — a recent generation date does not make every underlying source equally recent. Check the table above for each source's actual vintage.
Version
This methodology corresponds to Methodology version 1.0 — the same version number shown in every report's footer.
Measured versus modelled versus inferred
Measured: sourced facts from government or licensed datasets (population, transit stops, land-use zones, resale transactions). Modelled: values derived from those facts (area-weighted resident estimates per ring, density benchmarks, walk-network reach). Inferred: interpretive statements — always labelled as such (for example "this suggests…" or "a hypothesis worth verifying on site"). Footfall, customer behaviour and sales are never measured here, and the report never claims they are.
Automated commentary
Limitations
Built for early-stage site screening. It does not replace a site visit, lease review, financial model or professional advice.
This product is designed for early-stage screening and shortlisting. It does not provide primary market research, rent valuation, structural or lease review, or on-the-ground verification — those remain the work of a site visit and professional advice.