Week 12: From evidence to a proposal

How to structure a scoped, evaluable intervention proposal — and what separates a proposal a marker can act on from one they can only admire.

10 May 2027

Open the slides

Twelve weeks of physics, ecology, detection, engineering choice, liability, process, communication, and evidence transfer all point at one document: the final project. This lecture is about the structure that document needs, using a fictional worked example to show what “arguing from three strands” actually looks like written down, not just described.

The six moves a proposal has to make#

A marker reading your proposal is checking for six things, in this order, and a gap in any one of them is a gap a marker will find:

1. Problem — what specific road, road class, or site type, and what about it makes wildlife-vehicle collision a genuine issue there (not “kangaroos are a problem in Australia” — that’s too broad to evaluate).
2. Evidence — at least three of this unit’s strands (physics/injury severity, ecology/behaviour, detection/reporting bias, engineering effectiveness, liability, process, communication, evidence transfer), argued specifically, not gestured at.
3. Mechanism — why the evidence implies this particular intervention works, in causal terms.
4. Intervention — one primary measure. Not a bundle of five different things “for thoroughness” — a marker reading a five-measure bundle cannot tell which measure you actually think does the work.
5. Implementation — a realistic jurisdiction, an approval pathway through that jurisdiction’s actual process, and separated initial versus ongoing costs.
6. Evaluation — what would count as evidence the intervention worked, distinguishing a process metric (did the fence get built, is the underpass being used) from an outcome metric (did collisions actually fall, adjusted for exposure per Week 3).

A word on the synthetic dataset used throughout this unit: it is a teaching tool, not real-world data. Present a pattern found in it as “in this teaching dataset…” — never as a real-world finding, and never as evidence supporting a real jurisdiction’s numbers.

A worked example: weak version#

The fictional example below concerns Kestrel Pass Road — a fictional road, used here only as a worked teaching example, not a real place.

This proposal recommends fencing, an underpass, a rope bridge canopy crossing, dynamic LED warning signs, and a public awareness campaign along Kestrel Pass Road. Together, these measures will dramatically reduce kangaroo-vehicle collisions and set a new standard for wildlife road safety. Implementation should proceed as soon as possible.

What’s wrong, move by move: no site detail beyond a name (Problem — too thin); no strand cited at all (Evidence — missing entirely); “dramatically reduce” is asserted, not derived from a mechanism (Mechanism — missing); five measures bundled together (Intervention — not scoped); no jurisdiction, no approval pathway, no cost figures at all (Implementation — missing); no evaluation plan of any kind (Evaluation — missing). This version would fail nearly every one of the final project’s requirements, not just one.

The same example, improved#

Problem. Kestrel Pass Road (fictional) is a 4.2km rural two-lane stretch in NSW — the jurisdiction used throughout this proposal for cost and approval reference — bordered by continuous eucalypt scrub on both sides, with a traffic volume comparable to Coorabin Road in this unit’s teaching dataset.

Evidence. (i) Physics — Week 4’s kinetic-energy relationship shows that higher speed means more energy involved in a collision, which is why this stretch’s traffic speed is worth flagging at all; it does not, on its own, predict how severe any actual collision will be, since the real injury outcome also depends on the animal’s mass, the geometry of the impact, and the vehicle’s structure — the model explains why speed belongs in this discussion, not what any one collision will look like. (ii) Ecology — continuous roadside scrub without a break is the kind of pattern that makes it worth checking whether animals are actually crossing here in numbers. On its own it is a reason to survey movement on this stretch, not proof that this specific site has high crossing demand. (iii) Engineering effectiveness — Rytwinski et al. (2016), the meta-analysis introduced in Week 7, found fencing combined with a crossing structure reduced large-mammal road-kill by 83%, against roughly 1% for roadside reflectors and no detectable effect for crossing structures installed without fencing.

Mechanism. Fencing physically excludes animals from the carriageway, and the crossing structure preserves the movement route the fencing would otherwise cut off. Running the fence the full 4.2km length of the continuous-scrub section, rather than stopping short partway along it, reduces — but does not by itself eliminate — the Week 7 “end of fence” displacement problem: the fence still has to end somewhere, so each endpoint needs its own check for animal movement and collision risk, with a plan (an angled return wing, extra signage, or a monitoring commitment) if that check finds a live risk there.

Intervention. Exclusion fencing along the full 4.2km, plus one crossing structure, as the single primary measure. The midpoint is this proposal’s candidate site for the crossing structure, not a confirmed one — final siting depends on an animal-movement survey, terrain, and engineering conditions that haven’t been checked yet; the midpoint is where the investigation would start, not where construction is guaranteed to happen.

Implementation. The figures below reuse Week 7’s fictional planning-estimate assumptions for this teaching example — they are not a real NSW quote, and reusing them alone would not satisfy the final project’s requirement for a realistic cost basis, which needs figures sourced from the actual road authority or a comparable real project. Illustrative only: approximately $756,000 initial cost and $25,200 annual maintenance. On approval: which authority signs off depends on how this stretch is classified — Transport for NSW is the roads authority for the state-road network, while regional and local roads sit with the relevant council — and that classification hasn’t been confirmed for this fictional road. Fencing and vegetation work along a continuous scrub corridor may also need a Part 5 environmental assessment and a threatened-species check if the corridor turns out to support a listed species; both would need to be confirmed with the relevant authority before this pathway is anything more than “to be confirmed.”

Evaluation. A process metric (crossing-structure usage recorded via the same camera-trap method used in this unit’s detection weeks) and an outcome metric (exposure-adjusted collision rate on this stretch, compared pre- and post-construction using the Week 3 adjustment method, not a raw count).

Notice what this version does that the weak version didn’t: it uses only one primary measure, names a real jurisdiction for cost/approval purposes while keeping the road itself clearly fictional, separates “did people build it and do animals use it” from “did collisions actually fall” — and says out loud, next to each claim, which parts are backed by evidence and which are still open questions (the exact crossing site, the responsible authority, whether an environmental assessment is triggered, a real-world cost). Writing “further investigation needed” where that’s genuinely true is stronger evidence practice than dressing an unconfirmed detail up as settled just to make the proposal look finished.

Handling a question you didn’t anticipate#

If asked something your proposal doesn’t cover, say what your proposal does and doesn’t claim, rather than improvising an answer that oversteps your evidence — “my proposal doesn’t test that; here’s what would be needed to find out” is a stronger answer than a confident guess.

What the in-class session does and doesn’t do#

This week’s session gives you a chance to talk through your proposal with peers and get feedback before submission. It adds no separate grade of its own — marking is holistic, against the written proposal you submit by the real deadline, per the final project brief.

Reading#

Rytwinski, T., et al. (2016). “How Effective Is Road Mitigation at Reducing Road-Kill? A Meta-Analysis.” PLoS ONE, 11(11), e0166941. Already introduced in Week 7 — worth revisiting here as the evidence backbone of the worked example above.

Transport for NSW, “Road classification” (accessed 2026) — the state/regional/local split behind the “which authority signs off” question in the worked example above. Transport for NSW, “Biodiversity” (accessed 2026) — background on when road and vegetation work needs an environmental or threatened-species check. Neither page tells you what Kestrel Pass Road’s own answer is; that would still need confirming with whichever authority actually owns the stretch.