Safe System Assessment explained · crashstats.com.au

Scoring a road against zero deaths

A Safe System Assessment puts one question to every road design: how far is it from a system in which no foreseeable crash kills or seriously injures anyone? Seven crash types, three dimensions, one auditable number — here is the whole method, in one place.

The method comes from the Austroads Safe System Assessment Framework and is operationalised in state guidance such as the VicRoads SSA Guidelines. It is proactive — scored from the design and its context, not from waiting for crashes — and it is not a Road Safety Audit: an RSA checks a design against standards; an SSA measures its distance from the Safe System end-state.

Proactive · design data 7 crash types × E·L·S Lower /448 = closer to zero SSA ≠ Road Safety Audit
Why this page exists

Everyone audits roads. This scores them against zero.

Road safety practice grew up checking designs against standards: does the curve meet the guide, is the sign the right size, is the barrier warranted? Those checks matter — but a design can pass every one of them and still leave people dying in foreseeable crashes. The Safe System Assessment exists to expose exactly that gap.

It starts from the Safe System principles: no death or serious injury is an acceptable price for mobility; people make mistakes, so the system must forgive them; people are physically fragile, so impact energies must stay below what a body can survive; and responsibility is shared between those who design, operate and use the network. An SSA turns those principles into a number a project team can be held to.

Scope: this page covers the assessment method

It explains how a road or design is scored against Safe System alignment — the matrix, the anchors, the thresholds and the treatment hierarchy. It deliberately stays at method level: it is not site-specific engineering advice, and it does not rank particular projects. For how network-level risk is measured — collective, personal and systemic risk, star ratings and the IRR — see the companion explainer at ausrap.crashstats.com.au.

The engine

Risk = Exposure × Likelihood × Severity

Every cell of an SSA reduces to three questions. Exposure — how many road users, of what kind, are present to be involved in this crash type? Likelihood — how strongly does the infrastructure invite or forgive the human error that triggers it? Severity — if it happens, do the impact energies stay below what a human body survives? Because the three multiply, a zero anywhere zeroes the cell: no exposure, or no plausible mechanism, or fully survivable energy each individually close out a crash type.

The Safe System Assessment pipeline A design and its context feed a matrix of seven crash types, each scored zero to four for exposure, likelihood and severity. The three scores multiply to a product out of 64 per crash type; the seven products sum to a total out of 448. Options are compared and treatments classified as Primary or Supporting. One goal — eliminate fatal & serious injury (FSI) one matrix, three multiplying dimensions, one comparable number Design + context function · speed environment · users · vehicle mix 7 foreseeable crash types run-off-road · head-on · intersection · other · ped · cyclist · m/c Exposure 0–4 AADT bands · VRU units/day movements · segment length Likelihood 0–4 geometry · conflict points · sight distance · separation · speed Severity 0–4 impact speeds & angles · hazards barriers · energy dissipation E × L × S = product /64 per crash type a zero in any dimension closes the crash type out Σ seven products = total /448 existing conditions vs each design option — lower is better then: classify treatments (Primary / Supporting) and recommend changes that move the score toward zero
The whole method on one canvas: context frames seven crash types; each is scored 0–4 three ways; products out of 64 sum to a comparable total out of 448.
The anchors

What 0–4 actually means

The scores are anchored, not vibes. Exposure follows traffic volumes; likelihood follows how strongly the layout invites error; severity follows whether impact energy stays survivable or is dissipated before it reaches a person.

ScoreExposureLikelihoodSeverity
0No exposure to the crash type — no side flow, no crossing pedestrians, no cyclistsOnly extreme behaviour or substantial vehicle failure could produce the crashImpact energies too low to cause FSI, or effectively redirected / dissipated first
1Very low — AADT < 1,000 veh/day, or < 10 pedestrians / cyclists / motorcycles per dayHighly unlikely for a given user given the infrastructure in placeFSI highly unlikely — energies fairly low or mostly dissipated
2Moderate — AADT 1,000–5,000, or 10–50 VRU units/dayUnlikelyFSI unlikely — moderate energies, mostly dissipated before reaching people
3High — AADT 5,000–10,000, or 50–100 VRU units/dayLikelyFSI likely — moderate energies, not effectively dissipated
4Very high — AADT > 10,000, or > 100 VRU units/day, or very long segmentUser error leading to this crash is likely given the layout — high approach speeds, priority control, filtering turns across opposing lanesFSI highly likely — energies high and unlikely to be dissipated before reaching the road user
0closed out — Safe System aligned
1very low residual risk
2moderate — work to do
3high — misaligned
4severe — furthest from zero

Practitioner conventions worth knowing: at an intersection site the Intersection and Other exposure scores take the same value as Head-on and Run-off-road — it is the same traffic stream. And a cell already scored 4 lists no reducing factors: 4 means nothing present mitigates it.

The physics

Severity is physics: the survivable-speed thresholds

Severity scoring rests on human crash tolerance. Above these approximate impact speeds, the probability of death or serious injury climbs steeply for each configuration:

Impact speed thresholds
  • ≈ 70 km/h — head-on, car to car
  • ≈ 50 km/h — right-angle intersection impact, car to car
  • ≈ 30–40 km/h — side impact into a pole or tree
  • ≈ 20–30 km/h — any conflict with a pedestrian, cyclist or motorcyclist

For people on foot or on bikes: roughly a 10% severe-injury risk at ~20 km/h impact — and death is almost certain at 50 km/h.

Angle limits by speed
  • ≤ 40 km/h — any conflict angle acceptable
  • 50 km/h — up to 90°
  • 60 km/h — oblique only: 52° / 128°
  • 70 km/h — only 0° / 180° (head-on / rear-end geometry)
  • ≥ 80 km/h — no conflict angle is survivable by design

This is why geometry that trims impact angles — roundabout deflection, offset turns — is severity treatment, not decoration.

speed’s leverage: fatal crashes scale with ≈ the 4th power of relative speed change; serious-injury crashes with ≈ the 3rd power. A few km/h of mean speed is never “a few percent” of trauma.
The hierarchy

Primary, Supporting — and the trap between them

Primary — virtually eliminates FSI potential Supporting — improves safety, usually likelihood Non-Safe System — improves nothing, or blocks a future Primary

Primary treatments work on severity: they force impact energies below the survivable thresholds or remove the conflict outright — roundabouts with genuine speed-reducing geometry, raised safety platforms, grade separation, continuous flexible barrier, movement bans that delete a conflict. Supporting treatments trim likelihood — better delineation, sight lines, channelisation, signals — worth having, but nobody's kinetic energy changes because a lamp turned red.

Where assessments go wrong

“Primary” does not mean “preferred”. It is a physics claim about eliminating FSI, not a statement of project priority.

Signals are not severity treatments. Signalisation does not reduce conflict points or impact angles — roundabouts do. And it does little for cyclists unless cycle-specific infrastructure (lanes, advanced stop boxes, cycle signals) comes with it.

Exposure means removing movements. Left-in/left-out, closing a leg, redirecting traffic to higher-order roads reduce exposure. Channelising a turn improves likelihood — a frequent mislabel.

Severity claims must name the infrastructure. “Reduce speeds” is an intention; “raised platform at the conflict point” is a treatment.

Sibling confusion

SSA vs Road Safety Audit

Both are independent, proactive reviews — which is exactly why they get conflated. They answer different questions:

Road Safety AuditSafe System Assessment
QuestionIs the design safe against standards and known hazards?How far is the design from eliminating fatal and serious injury?
BenchmarkDesign guides, standards, audit experienceThe Safe System end-state — zero FSI
OutputFindings and corrective actionsScored matrix (/448), option comparison, Primary/Supporting treatment recommendations
A passing result meansNo defects found against current practiceQuantified residual distance from zero — even a “good” design gets a number

A design can clear an audit and still score badly on an SSA — a compliant 80 km/h undivided road with trees at the clear-zone edge is the classic case. The two reviews complement each other; neither substitutes for the other.

The fragile end of the matrix

Vulnerable road users: three of the seven rows

Pedestrians, cyclists and motorcyclists carry no protective shell, so their severity threshold — 20–30 km/h — is the strictest in the framework. The design choice is binary and honest: slow vehicles to survivable speeds where people mix, or separate them properly. Raised crossings and platforms, protected intersections and separated paths do the first two; barriers, medians and network routing do the third.

Motorcyclists get their own lens — the five S's: Sight (can they see and be seen), Surface (grip, debris, edges), Signs (roadside furniture as a strike hazard and a message), Strike (what they hit — severity), Speed (both likelihood and severity). Route function matters too: a road's Movement and Place classification shapes which users are present and which crash types are foreseeable in the first place.

Running one

From context to recommendation

  1. Frame the context. Why does the project exist, what is the road's function, what is the speed environment, who is present (elderly pedestrians, school children, cyclists), what is the vehicle mix (heavy-vehicle share, motorcyclists)? Context determines which crash types are live.
  2. Score existing conditions. Complete the full matrix with commentary: for every cell, the factors that push the score up and the factors that hold it down — auditable, not asserted.
  3. Score each design option the same way, and compare totals and per-crash-type products against existing conditions. The deltas show precisely where an option buys its safety.
  4. Classify treatments. Sort what the design already contains into Primary / Supporting / Non-Safe System — using the physics definitions, not preference.
  5. Recommend. New measures that would lift alignment further — recommendations are ideas beyond the assessed design, never a restatement of what it already includes.

Full or rapid

A full SSA convenes a team, meetings and site inspections; a rapid SSA is a leaner desktop application of the same matrix for smaller or earlier-stage projects — state guidance (e.g. VicRoads SSA Guidelines) sets when each applies. Either way the report records the assessment type and why, the team and its independence, the context, the scored matrices with commentary, and the treatment tables.

Primary sources

References

  1. Austroads — Safe System Assessment Framework and related guidance.
  2. Austroads AP-R560-18 — Towards Safe System Infrastructure: A Compendium of Current Knowledge.
  3. Austroads AP-R611-20 — Integrating Safe System with Movement and Place for Vulnerable Road Users.
  4. Austroads AP-R642-20 — Effectiveness and Implementation of Raised Safety Platforms.
  5. Austroads AGRS07 — Guide to Road Safety Part 7: Road Safety Strategy and Management.
  6. VicRoads — Safe System Assessment Guidelines (scoring factors and full-vs-rapid criteria).
  7. Department of Transport (Vic) — Road Design Notes RDN 03-07 (raised safety platforms) and RDN 0403 (compact roundabouts in rural high-speed environments).
  8. Monash University Accident Research Centre (2010) — Safe System intersection design principles (impact speed and angle limits).
  9. Elvik, R. et al. (2004) — speed–crash power model.
  10. Ding, C. et al. (2019) — motorcyclist injury risk as a function of crash speed, Accident Analysis & Prevention.
SSA Explainer AI Assistant
Answers only from this page and its cited primary sources - every answer independently verified before display.
G’day — ask me anything about the Safe System Assessment: the matrix, the 0–4 anchors, the /448 score, survivable impact speeds, or Primary vs Supporting treatments.
Educational explainer assistant by Sabour Khosravi (SafeFuture Lab - rCITI@UNSW). Independent of any government agency. Not engineering advice. Questions may be reviewed to improve accuracy.