How BuyerLens screens and compares suburb evidence.

Status and scope

Current versions: suburb-discovery/2.1.0 and suburb-scorecard/2.1.0.

BuyerLens uses deterministic multi-criteria screening. It does not use an opaque AI model to choose a suburb, and a rank is not a probability, valuation, forecast or instruction to buy. Research Coach explanations do not alter the calculation.

Validation status: research-informed and testable, but not yet empirically certified as a forecasting model. BuyerLens will not claim predictive superiority until the rolling-origin validation gate below is passed using a sufficiently broad point-in-time Australian dataset.

Candidate discovery

1. Begin with the user's own research constraints, including available funds, selected states, property type and preferred market characteristics.
2. Calculate an indicative affordability boundary using the configured lending, purchase-cost, duty and reserve assumptions. This is an educational estimate, not a lending decision.
3. Apply the user's price, yield and vacancy filters before ranking so an attractive evidence score cannot override a stated constraint.
4. Create a deterministic, geographically and price-diversified candidate sample. It is bounded by provider coverage and data budgets and is not an exhaustive census of Australian suburbs.
5. Enrich candidates with available demand, supply, rental and longer-horizon market evidence, then display coverage, confidence and observation dates with the result.

An absent suburb has not necessarily failed. Candidate scores are relative to the evidence and comparison sample available for that search; they are not universal suburb ratings.

Four discovery themes

Growth, balanced and cash-flow preferences change the relative emphasis of these themes. They do not change the underlying observations or turn a score into personal advice.

Scoring safeguards

The Research Coach can explain a result but cannot change the deterministic calculation.

Suburb Scorecard

The scorecard classifies each available observation as meeting the research target, requiring further attention, missing the target or remaining unknown. Version-controlled numerical bands are applied consistently across supported sources, but the public methodology does not publish the proprietary weighting and calibration recipe.

The demand-to-supply composite is displayed only as supporting context rather than being scored again. Short price movements do not substitute for longer-term evidence, and geographically approximate approvals data is labelled as context where its boundary does not match the suburb.

• A rating is withheld until there is sufficient coverage across several independent evidence themes.
• “Strong evidence match” requires broad, dated evidence across demand, rental return, longer-horizon price context and supply—not merely a high result in one area.
• Unknown themes are excluded and visible; they are never silently treated as positive.
• The score describes evidence alignment only. It does not assess an individual property or direct the user to buy.

Sources and traceability

BuyerLens combines licensed Australian property-market data with official statistics and state-government datasets where available. Each source has a defined role; no source is treated as complete merely because it is authoritative.

• Users can inspect the source, observation date, retrieval date and known limitation associated with surfaced evidence.
• Provider confidence and geographic coverage are kept separate from the score itself.
• Conflicting, stale or unavailable observations remain visible for further investigation.
• Material property, title, planning, condition, insurance and contract matters must still be checked independently with appropriately licensed professionals.

Current provider roles and refresh expectations are published in the BuyerLens Trust Centre.

Why these themes fit Australian evidence

• The RBA’s Australian housing-market model links prices, rents, vacancies, construction, population/income demand and interest rates, and finds a strong vacancy effect on rents. RBA RDP 2019-01
• RBA local-market research finds different responses across Australian locations and identifies supply conditions and average income among relevant factors. That supports local/state comparison rather than one national threshold-only league table. RBA RDP 2020-02
• Australian rental evidence associates falling vacancy with tighter rental conditions and rising advertised rents. RBA rental-market research
• Australian supply responsiveness varies with planning, land, topography and existing use, so approvals are one supply signal rather than a complete answer. AHURI Final Report 281
• ABS approvals measure authorised work and are subject to geography and revision limitations. ABS Building Approvals

Interest rates and credit conditions matter nationally, but they do not distinguish two suburbs observed at the same time. BuyerLens therefore handles them through affordability and deal stress-testing rather than awarding every suburb the same ranking points.

Backtesting and change control

Randomly mixing past and future housing observations would leak later market information into earlier decisions. BuyerLens therefore uses a rolling-origin design: each historical cohort may use only evidence that would have been available at its snapshot date, then evaluates later outcomes. This follows time-series cross-validation principles where training information precedes the test outcome. Forecasting: Principles and Practice

Before BuyerLens makes any predictive-performance claim, the methodology must be tested across multiple future horizons, sufficiently broad Australian geography and both return and downside measures. A candidate version must be compared with the prior version and reviewed on unseen holdout data. Until that evidence gate is passed, the methodology remains an evidence-ranking framework—not a validated return forecast.

A methodology version changes whenever the meaning of a saved result changes.

Version 2.1 changes

• State/property-type and available metro/regional peer comparisons.
• A material penalty where observation dates are missing.
• Known dates required for the highest scorecard evidence label.
• Validation expanded to observed total return and downside, with real-price outcomes reported when supplied.

Version 2 changes

• One central set of scorecard bands for every data source.
• Correlated discovery signals grouped into four themes.
• National plus within-state percentile comparison.
• Explicit missing-evidence, confidence and recency adjustments.
• Short 36-month price movement no longer substitutes for long-term growth.
• Automatic rent-growth classification no longer counts the same observation again as trend.
• Rolling-origin v1 versus v2 validation harness and minimum evidence gate.

Known limitations

• The fetched candidate set can be truncated by provider coverage, time and account data budgets.
• Percentile ranks can change when filters or the comparison sample change.
• Past growth does not guarantee future growth.
• Approvals do not equal completed dwellings, and geography mapping can be approximate.
• Census demographics are historical context. SEIFA is not currently used as a buy/no-buy score.
• The score does not assess an individual property, street, title, contract, condition, insurance, valuation or personal suitability.