Methodology / Data inputs

Data inputs

Every number on this site starts as a real, published dataset — government-sourced timetable data, official boundary definitions, and census-based population counts.

Provenance

Vintage table

Some data is not up to date. This table pins every vintage in one place so you never have to hunt through prose to find out when a number was last refreshed.

InputSourceVintageRefreshCaveats
Timetables, stops, routesPTV GTFS, Data Vic (7 sub-feeds)processed 2026-07-12per buildPromise of timetable, not delivery. Cancellations and ghost services not reflected.
Suburb boundariesVicmap Admin localitycurrentper buildScope = Greater Melbourne + commuter corridor (70km of CBD, 657 localities).
Population totalsABS ERP per SA22024-25annualLatest available. ERP distributed dasymetrically within each SA2 by G-NAF addresses.
Population distributionG-NAF Core addressesG-NAF MAY 2026quarterly~1 quarter occupancy lead: addresses register before residents move in. Direction noted in assumptions.
Dwelling counts (route catchment)ABS Census mesh blocks20215-yearlySuperseded for suburb weighting by ERP+G-NAF model. Retained for route catchment only.
JobsCensus place-of-work by SA220215-yearlyDZN (Destination Zone) not available in free DataPacks — corrected from original plan. SA2 is suburb-scale, a reasonable match.
Roads (car comparison)OpenStreetMap extractper matrix buildirregularCovers 296 suburbs so far — only those inside the OSM/GTFS extract.
Congestion pendingDTP Bluetooth / Freeway Travel TimependingBlocked on API key; toggle built. Only free-flow driving comparisons are live.

Coverage

Scale of this build

This build scores 28,141 stops across 788 suburbs, with 5.3% unattributed (below the 10% build-failure threshold). Car competitiveness covers 296 suburbs.

The build reports its attribution failure rate honestly and fails if it exceeds 10%. A suburb with zero stops inside its exact locality boundary does not appear in the league table — that is not good service, it is the opposite, and has been confirmed for several growth-corridor localities.

Strength

Machine-readable inputs

Unlike some modelling projects whose inputs are hand-coded across multiple jurisdictions, every Transport Score input is machine-readable end to end. The GTFS feed, Vicmap boundaries, G-NAF address register, and ABS census data are all ingested programmatically with no manual scheme interpretation. This makes the pipeline reproducible by design: anyone with the same data and the published formulas can verify every number.