geodatamy Enter the observatory →

An open brief · Geodata MY · November 2026

What Malaysia’s road data actually looks like, measured.

Malaysia keeps returning to the same question: can we stand on our own geodata? Not licensed from abroad, not maintained elsewhere: ours, complete, current, and usable. This note is a contribution to that question, not a proposal. We measured the national road network end to end from public data, and every step is reproducible. No ask is attached. Take it, check it, use it, or tell us where it is wrong.

01
1,685,285road ways · 416,930 km

The baseline, today

38.3% of vehicle roads carry a name. 1.9% carry a speed limit. 49,508 places hold a complete 3-part address, against a 12M+ National Address System target. Measured across all 16 states and territories, plus 15 cities in depth.

02
5,139
1,392
6,982PPJ · MyGOS · OSM

Three records, three different maps

Three public layers each describe Putrajaya’s road network, and none agrees with the others. Even LLM’s public ArcGIS registry carries two corridors under one code (E1, E8, E19, E33). Merging custodian data without measuring it does not resolve that disagreement. It hides it.

03
1,094of 1,186 unnamed ways

A gap you can close cheaply

Unnamed OSM vehicle ways in Putrajaya within 25 m of a named PPJ segment: 92.2%. Candidate corrections recoverable from public data alone. ISPRS measured OSM road naming at 59% in 2018; our re-run says 61.4%. Eight years, +2.4 points. Attention has not moved it. Instrumentation might.

Who this is for

If you hold the dataCustodians and agencies get a neutral, reproducible baseline of the road fabric they own. Completeness, freshness, naming, per state and city, usable before any collation, verification or procurement cycle begins. Aggregate numbers only; your data stays yours.
If you build on itDevelopers and verifiers get 116,961 unnamed vehicle roads as edit-ready queues, plus the public JSONs and 10-minute spot-checks behind every number. Inspect it, export it, challenge it.
If you fund or govern itA concrete sizing of the physical metadata gap nation-usable geodata must close, set against the RM700m/yr spent overseas. The starting point is the measurement gap, not the politics.

Why the numbers matter

Sovereign geodata is not won by licensing. It is won when the data is complete, current, routable and verifiable by Malaysians, today. That is an engineering property, and engineering properties can be measured. If Malaysia is to stand on its own base map, the distance between “custodian-held” and “nation-usable” is exactly what these figures describe: per state, per city, per corridor, with the method attached.

Three ways in

The map16 states, 15 cities, 33 expressways & Pan Borneo trunks. Includes the LLM registry collisions and the Putrajaya divergence.
Open the interactive map →
The methodEvery number traces to a public endpoint or downloadable JSON/CSV. Run the 10-minute spot-checks yourself.
See Method & Sources →
The queues116,961 unnamed vehicle roads as CSVs with direct OSM links. The gap is actionable, not just observed.
Browse the queues →

This is a shared problem with a shared upside. Take any of it: the method, the queues, the scorecards. It exists to be used.

If you see something wrong, missing or worth doing differently, we want to hear it. Every number here has a reproducible trail, and corrections ship in the open. If you want to compare notes with the people behind this, the door is open: khanster.net.

Ground rules, plainly. OpenStreetMap data (ODbL) is the measurable baseline, not a substitute for custodian data. Edit recency is a mapping proxy, not ground truth. Custodian-versus-OSM comparisons need written permission (MyGDI SOP ss.8.1.3-8.1.4) and stay internal until then. We republish aggregate counts from PPJ’s public endpoint; PPJ geometry and per-feature records are not republished. Figures are snapshots from 15-16 Nov 2026 and drift with edits. Nothing here implies endorsement, commissioning, or agreement by PGN, MCMC, NAIO, JKR, LLM, PPJ or any agency. Corrections ship in the open, with a diff. The measurements, queues and downloads are free to use with attribution; the tooling and method internals are ours and come with an engagement. Want either, or want to talk about what this could look like inside your organisation? khanster.net is the door.