Methodology

Ordus is built on primary data — NSW Valuer General bulk sales records and SQM Research listings — not aggregator estimates. This page explains exactly how every number on the site is derived, so you can judge it for yourself.

Last updated 2026-10-19.

Data sources

• Sales & prices: NSW Valuer General bulk property sales dataset, every recorded sale 1990–present (~8.1M transactions).
• Listings & days-on-market proxy: SQM Research, at postcode level.
• Demographics: ABS Census 2021; socio-economic ranking from ABS SEIFA 2021.
• Boundaries & LGA mapping: NSW Spatial Services suburb–LGA concordance.
• Macro overlays: RBA cash rate, ABS CPI / unemployment / GDP / migration, ASX 200, ABS building approvals (all 1990–present).

Median prices

Each suburb’s monthly median is estimated from VG sales using nearest-month estimation: for a target month we take qualifying sales within a 6-month window, require at least 3 sales (single-sale months are skipped to avoid outliers), smooth across the 3 closest qualifying values, and reject values more than 3× off the trailing-12-month baseline. The headline “median sale price” shown across the terminal, the screener, My Property and these pages is a pooled trailing-12-month median: every qualifying sale in the 12 calendar months to the suburb’s latest settled month, one median, with the number of sales shown beside it. This is the same definition CoreLogic, Domain and REA use for a suburb median, so the figures are directly comparable. Growth rates and the chart line are computed from the monthly series, which is more stable than any single calendar month.

House vs unit split

VG records a unit/strata number when a sale is a unit. Sales with a unit number are classed as units; the rest as houses. Every median, growth rate and screener metric can be viewed for houses, units, or all sales.

Settlement-lag handling

VG records appear after settlement, so the most recent months are always under-counted and can look like a sudden price cliff. Ordus nulls the trailing months of each series when the sale count falls below a noise floor (roughly the lower of 3 sales or 30% of the trailing-12-month average), and the charts draw a “data settling” band over that region. This removes artefacts that affected ~13% of suburbs in raw published data. See the insights write-up for the full story.

Growth rates (CAGR)

1, 3, 5 and 10-year compound annual growth rates are computed from the pooled trailing-12-month median at each anchor point (every sale in the 12 months to that date, one median, not single-month figures), so they reflect sustained movement rather than monthly noise. The same pooled series draws each chart’s annual and quarterly points, so a year-on-year figure is always the ratio of two points on the line. Growth is reported per suburb across all sales; very thin suburbs (fewer than ~1 sale/year) are excluded from ranked surfaces.

Opportunity score (screener)

The 0–100 score in the Market Screener is a weighted blend of momentum and supply signals, normalised to each metric’s sensible range:

• Year-on-year price momentum — weight 25
• 5-year price trend (CAGR) — weight 20
• Days-on-market tightening (falling = better) — weight 20
• Relative value vs the national median — weight 15
• Sales-volume trend — weight 10
• 10-year long-run trend — weight 10

A suburb with no usable price history, or with less than 40% of the total signal weight available, is left unscored rather than flattered by a partial signal.

Property value estimate (My Property)

Given a property’s last sale, Ordus estimates today’s value as:

estimate = individual-factor × current class median × street premium × land adjustment

The individual factor is how the last sale compared to its suburb’s house/unit median at the time, on the same street. As the prior sale ages, that factor is decayed toward 1.0 (so a decade-old premium regresses to the suburb mean) and clamped to a sane range so outlier priors can’t dominate. The street premium is how that street historically trades versus the suburb median; the land adjustment nudges houses up or down for unusually large or small blocks (capped at ±30%).

Each estimate carries a confidence band based on how recent and how typical the prior sale is. Backtested against ~660 fresh NSW sales, the median absolute error is ~12% and the mean absolute percentage error ~15–16%. VG data carries no bedroom, bathroom or floor-area information, which is the main ceiling on estimate accuracy — the estimate is a market-data model, not a physical inspection.

Listing age (days-on-market proxy)

SQM publishes listing counts bucketed by age. Ordus collapses those buckets (using their midpoints) into a single median listing age per postcode — a proxy for days-on-market. It is postcode-level, not per-suburb.

Coverage & limitations

• Coverage: 8,321 NSW suburbs in the data; ~4,700 have enough sales history to warrant a standalone page. National rollout is in progress.
• Supply data is postcode-level, not suburb-level.
• No dwelling attributes (beds/baths/floor area) — VG doesn’t publish them.
• Recent months settle late — see settlement-lag handling above.
• Estimates and scores are decision-support, not financial advice.