Data Sources & Indicators

Every exposure profile is built from 11 indicators that describe the structural shape of a community: how people get around, what work is available, how far away critical supply chains are, and what buffers exist against economic pressure. None of these are predictions. They describe what is already true about a place.

Each community receives exposure weights across six domains: fuel, and transport, food and agriculture, electricity and grid, economic, housing, and emergency services. The weights are not machine-learned. They follow transparent rules: if a structural indicator crosses a threshold, the relevant domain’s weight increases.

The thresholds and cost relationships are drawn from published domain research: ACCC fuel monitoring (diesel as 30–40% of freight costs, regional price differentials), BITRE freight economics (distance-based cost modelling), ABARES commodity reports (freight share of food cost: 8–12%), RBA transmission research (rate-to-mortgage lag), and IEA energy policy reviews (import dependency, stockholding obligations). Full citations on the references page.

Each domain receives a weight between 0 and 1, calculated from how many structural factors contribute and how far each factor exceeds its threshold. A postcode with high car dependency in a regional location will have a higher fuel and transport weight than one with only high car dependency. The algorithm is deterministic: the same inputs always produce the same weights.

Two indicators use the Shannon diversity index, a measure borrowed from ecology that quantifies how evenly distributed a set of categories is. Higher values mean more even spread across categories. Lower values mean concentration in a few.

The spectrum runs from concentrated (low diversity, high exposure) to diversified (high diversity, lower exposure). Concentration is not inherently bad. A mining town may be wealthy, but it creates structural dependence on a single system. If that system is disrupted, there is less to fall back on.

This is what we call the entrainment penalty in the action urgency scoring. When a community's economy or transport system is entrained to a single mode, the urgency of diversification actions increases, even if current conditions look stable. A system that appears resilient because it is performing well is not the same as a system that can absorb disruption.

The exposure profile does not just describe structure. It contextualises live signals. The signals layer tracks real-time data across six cascade layers, from upstream market pressure (crude oil prices, exchange rates) through wholesale and retail prices to emergency incidents.

Not every signal matters equally to every postcode. A community with high car dependency in a regional area will see fuel price signals ranked higher. A community with a large agricultural workforce will see farm input cost signals ranked higher. The ranking uses parameterised templates: each signal has a relevance formula tied to the exposure domain it belongs to.

This is template-based contextualisation, not LLM generation. The same structural profile always produces the same signal ranking. The text you see in each profile is filled from templates using the community's actual numbers, not generated by a language model.

The codebase includes a full BRIC (Baseline Resilience Indicators for Communities) and INFORM (Index for Risk Management) composite scoring engine. It has been validated across 3,193 postcodes with five tests passing, including external validation against the ABS SEIFA index (Pearson r = 0.909).

This engine is preserved but not the primary output. The core structural data is from the 2021 Census, and a composite score built on five-year-old data risks false precision. The exposure profile approach was adopted to give communities actionable intelligence now, using the structural indicators we have confidence in, rather than waiting for a complete dataset.

When August 2026 Census data arrives, the composite scoring engine will be re-activated with fresh indicators across all six capitals. Full validation methodology and results are on the validation page.

Not every postcode has data for all 11 indicators. Some postcodes are absent from the Census data. Others lack solar capacity records or remoteness classifications. The system handles this transparently:

1. NULL values are excluded: if an indicator is missing for a postcode, it is left out of the exposure calculation rather than estimated or imputed.
2. Confidence drops proportionally: a profile built from 8 of 11 indicators carries lower confidence than one built from all 11. This is shown to the user.
3. Partial profiles are shown honestly: the system never hides what it does not know. If a domain's weight is based on incomplete indicators, the profile says so.

This approach prefers honesty over completeness. A partial profile with clear provenance is more useful than a complete-looking profile that papers over gaps.