Reading transaction data without false conclusions

Know what one row represents

Begin with the source’s definition of a record. A transaction period reported by month does not contain an exact contract day. Keep that precision in the display. An ingestion date tells you when the website received a record, not when a buyer purchased a home.

Separate new sale, subsale and resale. Keep asking prices outside transaction tables. These categories answer different questions, and pooling them can make a change in the mix look like a price movement. A developer’s current stock requires a different source again.

Supporting references: URA: property data and REALIS ↗

Coverage comes before the chart

Ask which dates and properties the dataset covers, when it was last successfully updated and whether the licence permits redistribution. URA offers property data and a subscription service; access to a service does not automatically grant a website permission to republish all of its records. HomeVantage does not display synthetic transactions while permission is pending.

A stale or incomplete source should remain visible as a limitation. The date shown must come from a successful data fetch, not from today’s page request. If the last import partially failed, distinguish its coverage from the last complete run.

Supporting references: URA: property data and REALIS ↗

Match comparable observations

A median depends on the observations included. Show sample count, period, sale category and area band with any statistic. Do not describe a change in two unlike samples as investment performance. Floor bands, unit types and tenure may also explain differences.

Small samples are particularly sensitive to one unusual unit. Inspect the distribution and individual records where permitted, rather than treating one number as a valuation. Avoid extrapolating from a development-wide figure to a particular stack or floor.

Supporting references: URA: property data and REALIS ↗

Corrections should be traceable

The import workflow identifies records using provider and source-record ID, retains revisions and requires approved project aliases. Reimporting the same record should not increase the count. A correction should be auditable. Ambiguous mappings belong in quarantine, not in a public chart.

Use transactions to improve the questions you ask, not to manufacture certainty. Take a specific home’s attributes and current condition to an appropriately qualified valuation or financing discussion.

• Confirm licence, coverage and last successful update.
• Keep monthly date precision and sale categories intact.
• Check area units and explicit project mappings.
• Show sample size beside a summary.
• Reconcile corrections and avoid unsupported return claims.

Supporting references: URA: property data and REALIS ↗

Sources & review

This is original research guidance, not an individual recommendation, legal opinion or loan approval. Hypothetical examples are not client results or available properties.

• URA: property data and REALIS ↗