SafeFuture Lab — road-safety evidence that survives being questioned

At a glance

Built statewide in Victoria. Designed to run in any state.

Every pipeline runs on public open data. Victoria was the first full build; the same methods apply wherever a state publishes signal, volume and road-geometry data.

4,527 / 4,588volumed signalised intersections in Victoria with per-approach evidence (98.7%)

1,267,052Victorian road sections with directional traffic volume

34,220skewed Victorian intersections mapped against crash history

99reusable spatial models in our QGIS library (2022–2026), not tied to one state

Five flagship proofs

What we have actually done

Statewide signal evidence (Victoria)

About 4,870 signal-configuration sheets — many scanned, with no text layer — read with AI vision and checked against road geometry, giving per-approach evidence at 4,527 of 4,588 volumed signalised intersections in Victoria.

Directional traffic volume for every Victorian road

1,267,052 Victorian road sections in three evidence tiers (measured, intersection, modelled), with about 71 billion vehicle-kilometres closed to a documented envelope — and the source disagreements published, not hidden.

Before/after evaluation platform

Every study runs naive, comparison-group and Empirical Bayes methods side by side, and every result ships with its interval — or a stated reason it cannot be estimated. Nothing is suppressed for being inconvenient. Worked example: Barry Road, CMF 0.75 (95% CI 0.30–1.85), consistent with a reduction, not proof of one.

Speed-zone evaluation and aggregation bias (NSW)

Six NSW speed-zone changes evaluated with weighted difference-in-differences, and a connected-vehicle method that separates path-level speeds from aggregate averages.