How sector capacity, need-adjusted density, peer metrics and AI analysis are calculated — with sources, caveats and known limitations.
For each combination of SA2 × cause group × financial year, we compute:
Organisations active: count of ACNC-registered charities with a service address in the SA2, active in that year, whose main_activity or sub_activity maps to the cause group.
Total revenue: sum of total_revenue from AIS for those organisations in that year.
FTE and volunteers: sum of full_time_employees + part_time_employees (converted to FTE at 0.6 ratio) and volunteer counts from AIS.
Need denominator: varies by cause (see Need-adjusted density below).
Density per 10k: organisations_active ÷ (estimated_resident_population ÷ 10,000).
The HHI measures market concentration. For each SA2 × cause × year, it is calculated as the sum of squared market shares (revenue share of each organisation), scaled to 0–10,000.
HHI > 2,500 indicates high concentration (one or two organisations dominate). Low HHI indicates a fragmented landscape. Percentile rank is calculated nationally within cause group.
Need-adjusted density replaces raw population with an estimated population in need as the denominator. This controls for the fact that not all residents equally require each type of service.
Formula: need_adjusted_density = organisations_active ÷ (need_denominator ÷ 10,000)
Disability: NDIS active participant count from NDIA quarterly data. Sourced quarterly. Not all people with disability are NDIS participants — this undercounts need but is the best available SA2-level proxy.
Mental Health: Estimated prevalence via indirect age-standardisation. National age-specific prevalence rates from ABS National Health Survey (DO003) are applied to each SA2's 2021 Census age structure. Rates: 0–14: 11.1%, 15–24: 25.6%, 25–34: 21.2%, 35–44: 20.6%, 45–54: 23.8%, 55–64: 23.0%, 65+: 20.0%.
Homelessness: AIHW Specialist Homelessness Services (SHS) clients per 10,000. State-level rates disaggregated to SA2 proportionally where SA2-level data is not available.
Aged Care: Population aged 75+, sourced from ABS Census 2021.
All other causes: Total estimated resident population (ERP). Effectively equivalent to raw density for these causes where no reliable SA2-level need proxy exists.
The ABS National Health Survey does not provide SA2-level estimates. We use indirect age-standardisation to estimate the proportion of each SA2's population likely to experience a mental health condition.
Method: For each SA2, multiply the 5-year age band population (from 2021 Census) by the national prevalence rate for that age band. Sum the results and divide by the SA2's total population.
Important caveat: This is an estimate of expected prevalence based on national rates and local age structure. It does not measure actual local mental health outcomes. It assumes the national age-specific rate applies uniformly within each age band, ignoring socioeconomic and geographic variation in true prevalence.
The result is stored as pct_mental_health in fact_abs_sa2_year.
Each organisation's financial and operating metrics are ranked within its peer group — organisations in the same primary cause group (from charity_cause_sa2_year) and the same financial year.
The following percentile columns are computed: revenue_percentile, fte_percentile, govt_dependence_percentile, funding_diversity_percentile, financial_resilience_percentile, revenue_growth_percentile, volunteer_intensity_percentile.
Revenue growth and volatility are capped at ±9,999% to avoid distortion from near-zero base-year values.
AI explanations, scores and recommendations are generated by Claude (Anthropic). Inputs are structured data from the Supabase database — financials, peer metrics, geographic context, sector scores. The model does not have access to real-time web search or any data outside the provided context.
Outputs include: financial health summary, governance risk flags, sector context narrative, peer comparison summary, and recommended due diligence questions.
Limitations: AI outputs reflect the quality and completeness of AIS data. Organisations that do not submit complete AIS returns will have limited financial context. AI analysis should be used as a starting point for investigation, not as a definitive assessment.
AIS data completeness: Not all charities file AIS returns. Small and medium charities may file simplified returns. Financial data for 2024 is incomplete as not all organisations have submitted by publication time.
SA2 boundary mismatch: 138 SA2s from 2016 ASGS boundaries used in older datasets do not map to 2021 ASGS boundaries. These are excluded from geographic analysis, resulting in 1,979 matched SA2s out of 2,310 total.
Service address vs operating area: Organisations are assigned to SA2s based on their registered service address, not their service delivery footprint. A statewide service provider may be recorded in a single metro SA2.
NDIS participant undercount: Not all people with disability are NDIS participants. The NDIA applies privacy suppression to SA2s with fewer than 5 participants. Need may be understated in thin markets.
AI analysis variability: AI-generated outputs may vary between calls for the same organisation. Use the structured data fields as the primary source of truth.