Mapped points of interest
OpenStreetMap is the base layer, enriched with commercial providers where a market justifies the cost. Shops, offices, schools, transit and land use become the demand, competition and accessibility signals.
How the scoring works, and what it will not tell you
This page exists so a recommendation can be argued with. If you cannot see why a location scored what it scored, the score is not worth anything to you.
The spatial unit
Every city is a grid, not a list of areas
A locality name covers ground that trades very differently at either end. So the city is divided into hexagonal cells — currently resolution 8 (~0.74 km² per cell) — and each one is scored independently. Pune resolves to 1,849 of them.
A locality's score is therefore a roll-up of real cells, and the site a report recommends is a specific cell rather than a neighbourhood in general.
GRIDresolution 8 (~0.74 km² per cell)
SIGNALS340+ features per cell
CATCHMENTWalk, primary and extended bands
SCORECity-relative, not out of 100 absolute
Where the data comes from
Franchise economics
Investment, fee, royalty, area and margin per format. Brand-supplied figures are labelled as brand-supplied; where a brand has not provided them, a business-format benchmark is used and labelled as modelled.
Rent
A modelled benchmark, compiled from commercial listing and brokerage ranges for the micro-market. It is a planning range to negotiate against, not a measurement of any particular unit — and it is labelled that way wherever it appears.
Collected observations
Surveys from customers and field staff, validated before they influence anything. An observation outranks a model; one person's unverified number does not silently reshape a locality for everyone else.
Score and confidence are different questions
A location can score well on evidence that is thin. Blending the two into one number would hide exactly the case a buyer most needs to see, so they are reported separately everywhere — in the workspace, in the report, and in the recommendation itself.
A high score with low confidence is not a recommendation. It is a lead worth a site visit, and the report says so in those words rather than in a footnote.
Observed
Someone measured it — a validated survey, a counted footfall, a real rent from a signed lease.
Derived
Computed from observations: catchment aggregates, competitor counts, accessibility from mapped roads and transit.
Modelled
Estimated from a benchmark where no observation exists. Usable, and always marked so you know which conclusions rest on it.
Assumed
A simulation input you can change yourself in the workspace. Nothing at this level is presented as a finding.
What the model does not claim
NOTA probability that your outlet will succeed
NOTReal-time data — every figure carries a date
NOTExact footfall; the index is a comparative measure
NOTEvery competitor — only what is mapped
NOTA rent quote, a lease term or a legal opinion
NOTPay-to-rank; a brand cannot buy a higher score
Coverage differs by market and is stated per market rather than as a national claim. Where a category is thin — informal or unmapped competition in particular — the report says so instead of reading absence as opportunity.
No paid placement, ever
A franchise brand can pay to be represented accurately and to have its figures reviewed. It cannot pay for a higher recommendation score, for priority ordering, or to have a poor fit suppressed. If a brand's own numbers make a site look worse, the site looks worse.
Payment status is not an input to any score and does not exist in the feature store.